Updates as of Dec 6, 2013 (8 months after the original post):

Martin Fink, CTO and Director of HP Labs, Hewlett-Packard Company [Oct 29, 2013]:

This Cloud, Social, Big Data and Mobile we are referring to as this “New Style of IT” [when talking about the slide shown above]

Through the Telescope: 3 Minutes on HP Moonshot [HewlettPackardVideos YouTube channel, July 24, 2013]

HIGHLY RECOMMENDED READING:
– HP Offers Exclusive Peek Inside Impending Moonshot Servers [Enterprise Tech, Nov 26, 2013]: “The company is getting ready to launch a bunch of new server nodes for Moonshot in a few weeks”.
– So far, the most simple and understandable info is serviced in Visual Configuration Moonshot diagram set: http://www.goldeneggs.fi/documents/GE-HP-MOONSHOT-A.pdf  This site includes also full visualisation for all x86 rack, desktop and blade servers.

From HP Launches Investment Solutions to Ease Organizations’ Transitions to “New Style of IT” [press release, Dec 6, 2013]

The HP accelerated migration program for cloud—helps …

The HP Pre-Provisioning Solution—lets …

New investment solutions for HP Moonshot servers and HP Converged Systems—provide customers and channel partners with quick access to the latest HP products through a simple, scalable and predictable monthly payment that aligns technology and financial requirements to business needs.   

Access the world’s first software defined server [HP offering, Nov 27, 2013]
With predictable and scalable monthly payments

HP Moonshot Financing
Cloud, Mobility, Security and Big Data require a different level of technology efficiency and scalability. Traditional systems may no longer be able to handle the increasing internet workloads with optimal performance. Having and investment strategy that gives you access to newer technology such as HP Moonshot allows you to meet the requirements for the New Style of IT.

A simple and flexible payment structure can help you access the latest technology on your terms.

Why leverage a predictable monthly payment?
• Provides financial flexibility to scale up your business
• May help mitigate the financial risk of your IT transformation
• Enables IT refresh cycles to keep up with latest technology
• May help improve your cash flow
• Offers predictable monthly payments which can help you stay within budget

How does it work?
• Talk to your HP Sales Rep about acquiring HP Moonshot using a predictable monthly payment
• Expand your capacity easily with a simple add-on payment
• Add spare capacity needed for even greater agility
• Set your payment terms based on your business needs
• After an agreed term, you’ll be able to refresh your technology

From The HP Moonshot team provides answers to your questions about the datacenter of the future [The HP Blog Hub, as of Aug 29, 2013]

Q: WHAT IS THE FUNDAMENTAL IDEA BEHIND THE HP MOONSHOT SYSTEM?

A: The idea is simple—use energy-efficient CPU’s attuned to a particular application to achieve radical power, space and cost savings. Stated another way; creating software defined servers for specific applications that run at scale.

Q: WHAT IS INNOVATIVE ABOUT THE HP MOONSHOT ARCHITECTURE?

A: The most innovative characteristic of HP Moonshot is the architecture. Everything that is a common resource in a traditional server has been converged into the chassis. The power, cooling, management, fabric, switches and uplinks are all shared across 45 hot-pluggable cartridges in a 4.3U chassis.

…

Q: EXPLAIN WHAT IS MEANT BY “SOFTWARE DEFINED” SERVER

A: Software defined servers achieve optimal useful work per watt by specializing for a given workload: matching a software application with available technology that can provide the most optimal performance. For example, the firstMoonshot server is tuned for the web front end LAMP (Linux/Apache/MySQL/PHP) stack. In the most extreme case of a future FPGA (Field Programmable Gate Array) cartridge, the hardware truly reflects the exact algorithm required.

Q: DESCRIBE THE FABRIC THAT HAS BEEN INTEGRATED INTO THE CHASSIS

A: The HP Moonshot 1500 Chassis has been built for future SOC designs that will require a range of network capabilities including cartridge to cartridge interconnect. Additionally, different workloads will have a range of storage needs. 

There are four separate and independent fabrics that support a range of current and future capabilities; 8 lanes of Ethernet; storage fabric (6Gb SATA) that enable shared storage amongst cartridges or storage expansion to a single cartridge; a dedicated iLO management network to manage all the servers as one; a cluster fabric with point to point connectivity and low latency interconnect between servers.

Martin Fink, CTO and Director of HP Labs, Hewlett-Packard Company [Oct 29, 2013]:

We’ve actually announced three ARM-based cartridges. These are available in our Discovery Labs now, and they’ll be shipping next year with new processor technology. [When talking about the slide shown above.]

Calxeda Midway in HP Moonshot [Janet Bartleson YouTube channel, Oct 28, 2013]

Details about the latest and future Calxeda SoCs see in the closing part of this Dec 7 update

@SC13: HP Moonshot ProLiant m800 Server Cartridge with Texas Instruments [Janet Bartleson YouTube channel, Nov 26, 2013]

Details about the latest Texas Instruments DSP+ARM SoCs see after the Calxeda section in the closing part of this Dec 7 update

The New Style of IT & HP Moonshot: Keynote by HP’s Martin Fink at ARM TechCon ’13 [ARMflix YouTube channel, recorded on Oct 29, published on Nov 11, 2013]

From Big Data and the future of computing – A conversation with John Sontag [HP Enterprise 20/20 Blog, October 28, 2013]

20/20 Team: Where is HP today in terms of helping everyone become a data scientist?

John Sontag: For that to happen we need a set of tools that allow us to be data scientists in more than the ad hoc way I just described. These tools should let us operate productively and repeatably, using vocabulary that we can share – so that each of us doesn’t have to learn the same lessons over and over again. Currently at HP, we’re building a software tool set that is helping people find value in the data they’re already surrounded by. We have HAVEn for data management, which includes the Vertica data store, and Autonomy for analysis. For enterprise security we have ArcSight and ThreatCentral. We have our work around StoreOnce to compress things, and Express Query to allow us to consume data in huge volumes. Then we have hardware initiatives like Moonshot, which is bringing different kinds of accelerators to bear so we can actually change how fast – and how effectively – we can chew on data.

20/20 Team: And how is HP Labs helping shape where we are going?

John Sontag: One thing we’re doing on the software front is creating new ways to interrogate data in real time through an interface that doesn’t require you to be a computer scientist.  We’re also looking at how we present the answers you get in a way that brings attention to the things you most need to be aware of. And then we’re thinking about how to let people who don’t have massive compute resources at their disposal also become data scientists.

20/20 Team: What’s the answer to that?

John Sontag: For that, we need to rethink the nature of the computer itself. If Moonshot is helping us make computers smaller and less energy-hungry, then our work on memristors will allow us to collapse the old processor/memory/storage hierarchy, and put processing right next to the data. Next, our work on photonics will help collapse the communication fabric and bring these very large scales into closer proximity. That lets us combine systems in new and interesting ways. And then we’re thinking about how to package these re-imagined computers into boxes of different sizes that match the needs of everyone from the individual to the massive, multinational entity. On top of all that, we need to reduce costs – if we tried to process all the data that we’re predicting we’ll want to at today’s prices, we’d collapse the world economy – and we need to think about how we secure and manage that data, and how we deliver algorithms that let us transform it fast enough so that you, your colleagues, and partners across the world can conduct experiments on this data literally as fast as we can think them up.

…

About John Sontag:

John Sontag is vice president and director of systems research at HP Labs. The systems research organization is responsible for research in memristor, photonics, physical and system architectures, storing data at high volume, velocity and variety, and operating systems. Together with HP business units and partners, the team reaches from basic research to advanced development of key technologies.

With more than 30 years of experience at HP in systems and operating system design and research, Sontag has had a variety of leadership roles in the development of HP-UX on PA-RISC and IPF, including 64-bit systems, support for multiple input/output systems, multi-system availability and Symmetric Multi-Processing scaling for OLTP and web servers.

Sontag received a bachelor of science degree in electrical engineering from Carnegie Mellon University.

Meet the Innovators [HewlettPackardVideos YouTube channel, May 23, 2013]

From Meet the innovators behind the design and development of Project Moonshot [The HP Blog Hub, June 6, 2013]

This video introduces you to key HP team members who were part of the team that brings you the innovative technology that fundamentally changes how hyperscale servers are built and operated such as:
• Chandrakant Patel – HP Senior Fellow and HP Labs Chief Engineer
• Paul Santeler  – Senior Vice President and General Manager of the HyperScale Business Unit
• Kelly Pracht – Moonshot Hardware Platform Manager, HyperScale Business Unit
• Dwight Barron – HP Fellow, Chief Technologist, HyperScale Business Unit

From Six IT technologies to watch [HP Enterprise 20/20 Blog, Sept 5, 2013]

1. Software-defined everything

Over the last couple of years we have heard a lot about software defined networks (SDN) and more recently, software defined data center (SDDC). There are fundamentally two ways to implement a cloud. Either you take the approach of the major public cloud providers, combining low-cost skinless servers with commodity storage, linked through cheap networking. You establish racks and racks of them. It’s probably the cheapest solution, but you have to implement all the management and optimization yourself. You can use software tools to do so, but you will have to develop the policies, the workflows and the automation.

Alternatively you can use what is becoming known as “converged infrastructure,” a term originally coined by HP, but now used by all our competitors. Servers, storage and networking are integrated in a single rack, or a series of interconnected ones, and the management and orchestration software included in the offering, provides an optimal use of the environment. You get increased flexibility and are able to respond faster to requests and opportunities.

We all know that different workloads require different characteristics. Infrastructures are typically implemented using general purpose configurations that have been optimized to address a very large variety of workloads. So, they do an average job for each. What if we could change the configuration automatically whenever the workload changes to ensure optimal usage of the infrastructure for each workload? This is precisely the concept of software defined environments. Configurations are no longer stored in the hardware, but adapted as and when required. Obviously this requires more advanced software that is capable of reconfiguring the resources.

A software-defined data center is described as a data center where the infrastructure is virtualized and also delivered as a service. Control of the data center is automated by software – meaning hardware configuration is maintained through intelligent software systems. Three core components comprise the SDDC, server virtualization, network virtualization and storage virtualization. It remains to be said that some workloads still require physical systems (often referred to as bare metal), hence the importance of projects such as OpenStack’s Ironic which could be defined as a hypervisor for physical environments.

2. Specialized servers

As I mentioned, all workloads are not equal, but run on the same, general purpose servers (typically x86). What if we create servers that are optimized for specific workloads? In particular, when developing cloud environments delivering multi-tenant SaaS services, one could well envisage the use of servers specialized for a specific task, for example video manipulation, dynamic web service management. Developing efficient, low energy specialized servers that can be configured through software is what HP’s Project Moonshot is all about. Today, although still in its infancy, there is much more to come. Imagine about 45 server/storage cartridges linked through three fabrics (for networking, storage and high speed cartridge to cartridge interconnections), sharing common elements such as network controllers, management functions and power management. If you then build the cartridges using low energy servers, you reduce energy consumption by nearly 90%. If you build SaaS type environments, using multi-tenant application modules, do you still need virtualization? This simplifies the environment, reduces the cost of running it and optimizes the use of server technology for every workload.

Particularly for environments that constantly run certain types of workloads, such as analyzing social or sensor data, the use of specialized servers can make the difference. This is definitely an evolution to watch.

3. Photonics

Let’s now complement those specialized servers with photonic based connections enabling flat, hyper-efficient networks boosting bandwidth, and we have an environment that is optimized to deliver the complex tasks of analyzing and acting upon signals provided by the environment in its largest sense.

But technology is going even further. I talked about the three fabrics, over time; why not use photonics to improve the speed of the fabrics themselves, increasing the overall compute speed. We are not there yet, but early experiments with photonic backplanes for blade systems have shown overall compute speed increased up to a factor seven. That should be the second step.

The third step takes things further. The specialized servers I talked about are typically system on a chip (SoC) servers, in other words, complete computers on a single chip. Why not use photonics to link those chips with their outside world? On-chip lasers have been developed in prototypes, so we are not that far out. We could even bring things one step further and use photonics within the chip itself, but that is still a little further out. I can’t tell you the increase in compute power that such evolutions will provide you, but I would expect it to be huge.

4. Storage

Storage is at a crossroads. On the one hand, hard disk drives (HDD) have improved drastically over the last 20 years, both in reading speed and in density. I still remember the 20MB hard disk drive, weighing 125Kg of the early 80’s. When I compare that with the 3TB drive I bought a couple months ago for my home PC, I can easily depict this evolution. But then the SSD (solid state disk) has appeared. Where a HDD read will take you 4 ms, the SDD read is down at 0.05 ms.

Using nanotechnologies, HP Labs did develop prototypes of the Memristor, a new approach to data storage, faster than Flash memory and consumes way less energy. Such a device could store up to 1 petabit of information per square centimeter and could replace both memory and storage, speeding up access to data and allowing order of magnitude increase in the amount of data stored. Since HP has been busy preparing production of these devices. First production units should be available towards the end of 2013 or early in 2014. It will transform our storage approaches completely.

Details about the latest and future Calxeda SoCs:

Calxeda EnergyCore ECX-2000 family – ARM TechCon ’13 [ARMflix YouTube channel, recorded on Oct 30, 2013]

From ECX-2000 Product Brief [October, 2013]

The Calxeda EnergyCore ECX-2000 Series is a family of SoC (Server-on-Chip) products that delivers the power efficiency of ARM® processors, and the OpenStack, Linux, and virtualization software needed for modern cloud infrastructures. Using the ARM Cortex A15 quad-core processor, the ECX-2000 delivers roughly twice the performance, three times the memory bandwidth, and four times the memory capacity of the ground-breaking ECX-1000. It is extremely scalable due to the integrated Fleet Fabric Switch, while the embedded Fleet Engine simultaneously provides out-of-band control and intelligence for autonomic operation.

In addition to enhanced performance, the ECX-2000 provides hardware virtualization support via KVM and Xen hypervisors. Coupled with certified support for Ubuntu 13.10 and the Havana Openstack release, this marks the first time an ARM SoC is ready for Cloud computing. The Fleet Fabric enables the highest network and interconnect bandwidth in the MicroServer space, making this an ideal platform for streaming media and network-intensive applications.

The net result of the EnergyCore SoC architecture is a dramatic reduction in power and space requirements, allowing rapidly growing data centers to quickly realize operating and capital cost savings.

Scalability you can grow into. An integrated EnergyCore Fabric Switch within every SoC provides up to five 10 Gigabit lanes for connecting thousands of ECX-2000 server nodes into clusters capable of handling distributed applications at extreme scale. Completely topology agnostic, each SoC can be deployed to work in a variety of mesh, grid, or tree network structures, providing opportunities to find the right balance of network throughput and fault resiliency for any given workload.

…

Fleet Fabric Switch
• Integrated 80Gb (8×8) crossbar switch with through-traffic support
• Five (5) 10Gb external channels, three (3) 10Gb internal channels
• Configurable topology capable of connecting up to 4096 nodes
• Dynamic Link Speed Control from 1Gb to 10Gb to minimize power and maximize performance
• Network Proxy Support maintains network presence even with node powered off
• In-order flow delivery
• MAC learning provider support for virtualization
…

ARM Servers and Xen — Hypervisor Support at Hyperscale – Larry Wikelius, [Co-Founder of] Calxeda [TheLinuxFoundation YouTube channel, Oct 1, 2013]

Calxeda Launches Midway ARM Server Chips, Extends Roadmap [EnterpriseTech, Oct 28, 2013]

ARM server chip supplier Calxeda is just about to ship its second generation of EnergyCore processors for hyperscale systems and most of its competitors are still working on their first products. Calxeda is also tweaking its roadmap to add a new chip to its lineup, which will bridge between the current 32-bit ARM chips and its future 64-bit processors.

There is going to be a lot of talk about server-class ARM processors this week, particularly with ARM Holdings hosting its TechCon conference in Santa Clara.

A month ago, EnterpriseTech told you about the “Midway” chip that Calxeda had in the works and as well as its roadmap to get beefier 64-bit cores and extend its Fleet Services fabric to allow for more than 100,000 nodes to be linked together.

The details were a little thin on the Midway chip, but we now know that it will be commercialized as the ECX-2000, and that Calxeda is sending out samples to server makers right now. The plan is to have the ECX-2000 generally available by the end of the year, and that is why company is ready to talk about some feeds and speeds. Karl Freund, vice president of marketing at Calxeda, walked EnterpriseTech through the details.

The Midway chip is fabricated in the same 40 nanometer process as the existing “High Bank” ECX-1000 chip that Calxeda first put into the field in November 2011 in the experimental “Redstone” hyperscale servers from Hewlett-Packard. That 32-bit chip, based on the ARM Cortex-A9 core, was subsequently adopted in systems from Penguin Computing, Boston, and a number of other hyperscale datacenter operators who did proofs of concept with the chips. The ECX-1000 has four cores and was somewhat limited in its performance and was definitely limited in its main memory, which topped out at 4 GB across the four-core processor. But the ECX-2000 addresses these issues.

The ECX-2000 is based on ARM Holding’s Cortex-A15 core and has the 40-bit physical memory extensions, which allows for up to 16 GB of memory to be physically attached to each socket. With the 40-bit physical addressing added with the Cortex-A15, the memory controller can, in theory, address up to 1 TB of main memory; this is called Large Physical Address Extension (LPAE) in the ARM lingo, and it maps the 32-bit physical addressing on the core to a 40-bit virtual address space. Each core on the ECX-2000 has 32 KB of L1 instruction cache and 32 KB of L1 data cache, and ARM licensees are allowed to scale the L2 cache as they see fit. The ECX-2000 has 4 MB of L2 cache shared across the four cores on the die. These are exactly the same L1 and L2 cache sizes as used in the prior ECX-1000 chips.

The Cortex-A15 design was created to scale to 2.5 GHz, but as you crank up the clocks on any chip, the amount of energy consumed and heat radiated grows progressively larger as clock speeds go up. At a certain point, it just doesn’t make sense to push clock speeds. Moreover, every drop in clock speed gives a proportionately larger increase in thermal efficiency, and this is why, says Freund, Calxeda is making its implementation of the Cortex-A15 top out at 1.8 GHz. The company will offer lower-speed parts running at 1.1 GHz and 1.4 GHz for customers that need an even better thermal profile or a cheaper part where low cost is more important than raw performance or thermals.

What Calxeda and its server and storage array customers are focused on is the fact that the Midway chip running at 1.8 GHz has twice the integer, floating point, and Java performance of a 1.1 GHz High Bank chip. That is possible, in part, because the new chip has four times the main memory and three times the memory bandwidth as the old chip in addition to a 64 percent boost in clock speed. Calxeda is not yet done benchmarking systems using the chips to get a measure of their thermal efficiency, but is saying that there is as much as a 33 percent boost in performance per watt comparing old to new ECX chips.

The new ECX-2000 chip has a dual-core Cortex-A7 chip on the die that is used as a controller for the system BIOS as well as a baseboard management controller and a power management controller for the servers that use them. These Fleet Engines, as Calxeda calls them, eliminate yet another set of components, and therefore their cost, in the system. These engines also control the topology of the Fleet Services fabric, which can be set up in 2D torus, mesh, butterfly tree, and fat tree network configurations.

The Fleet Services fabric has 80 Gb/sec of aggregate bandwidth and offers multiple 10 Gb/sec Ethernet links coming off the die to interconnect server nodes on a single card, multiple cards in an enclosure, multiple enclosures in a rack, and multiple racks in a data center. The Ethernet links are also used to allow users to get to applications running on the machines.

Freund says that the ECX-2000 chip is aimed at distributed, stateless server workloads, such as web server front ends, caching servers, and content distribution. It is also suitable for analytics workloads like Hadoop and distributed NoSQL data stores like Cassandra, all of which tend to run on Linux. Both Red Hat and Canonical are cooking up commercial-grade Linuxes for the Calxeda chips, and SUSE Linux is probably not going to be far behind. The new chips are also expected to see action in scale-out storage systems such as OpenStack Swift object storage or the more elaborate Gluster and Ceph clustered file systems. The OpenStack cloud controller embedded in the just-announced Ubuntu Server 13.10 is also certified to run on the Midway chip.

Hewlett-Packard has confirmed that it is creating a quad-node server cartridge for its “Moonshot” hyperscale servers, which should ship to customers sometime in the first or second quarter of 2014. (It all depends on how long HP takes to certify the system board.) Penguin Computing, Foxconn, Aaeon, and Boston are expected to get beta systems out the door this year using the Midway chip and will have them in production in the first half of next year. Yes, that’s pretty vague, but that is the server business, and vagueness is to be expected in such a young market as the ARM server market is.

Looking ahead, Calxeda is adding a new processor to its roadmap, code-named “Sarita.” Here’s what the latest system-on-chip roadmap looks like now:

The future “Lago” chip is the first 64-bit chip that will come out of Calxeda, and it is based on the Cortex-A57 design from ARM Holdings –one of several ARMv8 designs, in fact. (The existing Calxeda chips are based on the ARMv7 architecture.)

Both Sarita and Lago will be implemented in TSMC’s 28 nanometer processes, and that shrink from the current 40 nanometer to 28 nanometer processes is going to allow for a lot more cores and other features to be added to the die and also likely a decent jump in clock speed, too. Freund is not saying at the moment which way it will go.

But what Freund will confirm is that Sarita will be pin-compatible with the existing Midway chip, meaning that server makers who adopt Midway will have a processor bump they can offer in a relatively easy fashion. It will also be based on the Cortex-A57 cores from ARM Holdings, and will sport four cores on a die that deliver about a 50 percent performance increase compared to the Midway chips.

The Lago chips, we now know, will scale to eight cores on a die and deliver about twice the performance of the Midway chips. Both Lago and Sarita are on the same schedule, in fact, and they are expected to tape out this quarter. Calxeda expects to start sampling them to customers in the second quarter of 2014, with production quantities being available at the end of 2014.

Not Just Compute, But Networking, Too

As important as the processing is to a system, the Fleet Services fabric interconnect is perhaps the key differentiator in its design. The current iteration of that interconnect, which is a distributed Layer 2 switch fabric that is spread across each chip in a cluster, can scale across 4,096 nodes without requiring top-of-rack and aggregation switches.

Both of the Lago and Sarita chips will be using the Fleet Services 2.0 intehttp://www.ti.com/product/66ak2h12rconnect that is now being launched with Midway. This iteration of the interconnect has all kinds of tweaks and nips and tucks but no scalability enhancements beyond the 4,096 nodes in the original fabric.

Freund says that the Fleet Services 3.0 fabric, which allows the distributed switch architecture to scale above 100,000 nodes in a flat network, will probably now come with the “Ratamosa” chips in 2015. It was originally – and loosely – scheduled for Lago next year. The circuits that do the fabric interconnect is not substantially different, says Freund, but the scalability is enabled through software. It could be that customers are not going to need such scalability as rapidly as Calxeda originally thought.

The “Navarro” kicker to the Ratamosa chip is presumably based on the ARMv9 architecture, and Calxeda is not saying anything about when we might see that and what properties it might have. All that it has said thus far is that it is aimed at the “enterprise server era.”

Details about the latest Texas Instruments DSP+ARM SoCs:

A Better Way to Cloud [MultiVuOnlineVideo YouTube channel, Nov 13, 2012]

Infinite Scalability in Multicore Processors [Texas Instruments YouTube channel, Aug 27, 2012]

From Imagine the impact…TI’s KeyStone SoC + HP Moonshot [TI’s Multicore Mix Blog, April 19, 2013]

TI’s participation in HP’s Pathfinder Innovation Ecosystem is the first step towards arming HP’s customers with optimized server systems that are ideally suited for workloads such as oil and gas exploration, Cloud Radio Access Networks (C-RAN), voice over LTE and video transcoding. This collaboration between TI and HP is a bold step forward, enabling flexible, optimized servers to bring differentiated technologies, such as TI’s DSPs, to a broader set of application providers. TI’s KeyStone II-based SoCs, which integrate fixed- and floating- point DSP cores with multiple ARM® Cortex™A-15 MPCore processors, packet and security processing, and high speed interconnect, give customers the performance, scalability and programmability needed to build software-defined servers. HP’s Moonshot system integrates storage, networking and compute cards with a flexible interconnect, allowing customers to choose the optimized ratio enabling the industry’s first software-defined server platform. Bringing TI’s KeyStone II-based SoCs into HP’s Moonshot system opens up several tantalizing possibilities for the future. Let’s look at a few examples:

Think about the number of voice conversations happening over mobile devices every day. These conversations are independent of each other, and each will need transcoding from one voice format to another as voice travels from one mobile device, through the network infrastructure and to the other mobile device. The sheer number of such conversations demand that the servers used for voice transcoding be optimized for this function. Voice is just one example. Now think about video and music, and you can imagine the vast amount of processing required. Using TI’s KeyStone II-based SoCs with DSP technology provides optimized server architecture for these applications because our SoCs are specifically tuned for signal processing workloads.

Another example can be with C-RAN. We have seen a huge push for mobile operators to move most of the mobile radio processing to the data center. There are several approaches to achieve this goal, and each has pros and cons associated with them. But one thing is certain – each approach has to do wireless symbol processing to achieve optimum 3G or 4G communications with smart mobile devices. TI’s KeyStone II-based SoCs are leading the wireless communication infrastructure market and combine key accelerators such as BCP (Bit Rate Co-Processor), VCP (Viturbi Co-Processor) and others to enable 3G/4G standards compliant for wireless processing. These key accelerators offload standard-based wireless processing from the ARM and/or DSP cores, freeing the cores for value-added processing. The combination of ARM/DSP with these accelerators provides an optimum SoC for 3G/4G wireless processing. By combining TI’s KeyStone II-based SoC with HP’s Moonshot system, operators and network equipment providers can now build customized servers for C-RAN to achieve higher performance systems at lower cost and ultimately provide better experiences to their customers.

A better way to cloud: TI’s new KeyStone multicore SoCs [embeddednewstv YouTube channel, published on Jan 12,2013 (YouTube: Oct 21, 2013)]

Texas Instruments Offers System on a Chip for HPC Applications [RichReport YouTube channel, Nov 20, 2012]

A better way to cloud: TI’s new KeyStone multicore SoCs revitalize cloud applications, enabling new capabilities and a quantum leap in performance at significantly reduced power consumption

• Industry’s first implementation of quad ARM® Cortex™-A15 MPCore™ processors in infrastructure-class embedded SoC offers developers exceptional capacity & performance at significantly reduced power for networking, high performance computing and more
• Unmatched combination of Cortex-A15 processors, C66x DSPs, packet processing, security processing and Ethernet switching, transforms the real-time cloud into an optimized high performance, power efficient processing platform
• Scalable KeyStone architecture now features 20+ software compatible devices, enabling customers to more easily design integrated, power and cost-efficient products for high-performance markets from a range of devices

ELECTRONICA – MUNICH (Nov.13, 2012) /PRNewswire/ — To most technologists, cloud computing is about applications, servers, storage and connectivity. To Texas Instruments Incorporated (TI) (NASDAQ: TXN) it means much more. Today, TI is unveiling a BETTER way to cloud with six new multicore System-on-Chips (SoCs). Based on its award winning KeyStone architecture, TI’s SoCs are designed to revitalize cloud computing, inject new verve and excitement into pivotal infrastructure systems and, despite their feature rich specifications and superior performance, actually reduce energy consumption.

To TI, a BETTER way to cloud means:

• Safer communities thanks to enhanced weather modeling;
• Higher returns from time sensitive financial analysis;
• Improved productivity and safety in energy exploration;
• Faster commuting on safer highways in safer cars;
• Exceptional video on any screen, anywhere, any time;
• More productive and environmentally friendly factories; and
• An overall reduction in energy consumption for a greener planet.

TI’s new KeyStone multicore SoCs are enabling this – and much more. These 28-nm devices integrate TI’s fixed-and floating-point TMS320C66x digital signal processor (DSP) generation cores – yielding the best performance per watt ratio in the DSP industry – with multiple ARM® Cortex™-A15 MPCore™ processors – delivering unprecedented processing capability combined with low power consumption – facilitating the development of a wide-range of infrastructure applications that can enable more efficient cloud experiences. The unique combination of Cortex-A15 processors and C66x DSPcores, with built-in packet processing and Ethernet switching, is designed to efficiently offload and enhance the cloud’s first generation general purpose servers; servers that struggle with big data applications like high performance computing and video processing.

“Using multicore DSPs in a cloud environment enables significant performance and operational advantages with accelerated compute intensive cloud applications,” said Rob Sherrard, VP of Service Delivery, Nimbix. “When selecting DSP technology for our accelerated cloud compute environment, TI’s KeyStone multicore SoCs were the obvious choice. TI’s multicore software enables easy integration for a variety of high performance cloud workloads like video, imaging, analytics and computing and we look forward to working with TI to help bring significant OPEX savings to high performance compute users.”

TI’s six new high-performance SoCs include the 66AK2E02, 66AK2E05, 66AK2H06, 66AK2H12, AM5K2E02 and AM5K2E04, all based on the KeyStone multicore architecture. With KeyStone’s low latency high bandwidth multicore shared memory controller (MSMC), these new SoCs yield 50 percent higher memory throughput when compared to other RISC-based SoCs. Together, these processing elements, with the integration of security processing, networking and switching, reduce system cost and power consumption, allowing developers to support the development of more cost-efficient, green applications and workloads, including high performance computing, video delivery and media and image processing. With the matchless combination TI has integrated into its newest multicore SoCs, developers of media and image processing applications will also create highly dense media solutions.

“Visionary and innovative are two words that come to mind when working with TI’s KeyStone devices,” said Joe Ye, CEO, CyWee. “Our goal is to offer solutions that merge the digital and physical worlds, and with TI’s new SoCs we are one step closer to making this a reality by pushing state-of-the-art video to virtualized server environments. Our collaboration with TI should enable developers to deliver richer multimedia experiences in a variety of cloud-based markets, including cloud gaming, virtual office, video conferencing and remote education.”

Simplified development with complete tools and support

TI continues to ease development with its scalable KeyStone architecture, comprehensive software platform and low-cost tools. In the past two years, TI has developed over 20 software compatible multicore devices, including variations of DSP-based solutions, ARM-based solutions and hybrid solutions with both DSP and ARM-based processing, all based on two generations of the KeyStone architecture. With compatible platforms across TI’s multicore DSPs and SoCs, customers can more easily design integrated, power and cost-efficient products for high-performance markets from a range of devices, starting at just $30 and operating at a clock rate of 850MHz all the way to 15GHz of total processing power.

TI is also making it easier for developers to quickly get started with its KeyStone multicore solutions by offering easy-to-use, evaluation modules (EVMs) for less than $1K, reducing developers’ programming burdens and speeding development time with a robust ecosystem of multicore tools and software.

In addition, TI’s Design Network features a worldwide community of respected and well established companies offering products and services that support TI multicore solutions. Companies offering supporting solutions to TI’s newest KeyStone-based multicore SoCs include 3L Ltd., 6WIND, Advantech, Aricent, Azcom Technology, Canonical, CriticalBlue Enea, Ittiam Systems, Mentor Graphics, mimoOn, MontaVista Software, Nash Technologies, PolyCore Software and Wind River.

Availability and pricing

TI’s 66AK2Hx SoCs are currently available for sampling, with broader device availability in 1Q13 and EVM availability in 2Q13. AM5K2Ex and 66AK2Ex samples and EVMs will be available in the second half of 2013. Pricing for these devices will start at $49 for 1 KU.

66AK2H14 (ACTIVE) Multicore DSP+ARM KeyStone II System-on-Chip (SoC) [TI.com, Nov 10, 2013]
The same as below for 66AK2H12 SoC with addition of:

More Literature:

• The Case for 10G Ethernet in Embedded Processing 
  (PDF , 189KB) 13 Nov 2013

From that the below excerpt is essential to understand the added value above 66AK2H12 SoC:

Figure 1. TI’s KeyStone™ 66AK2H14 SoC

The 66AK2H14 SoC shown in Figure 1, with the raw computing power of eight C66x processors and quad ARM Cortex-A15s at over 1GHz performance, enables applications such as very large fast fourier transforms (FFT) in radar and multiple camera image analytics where a 10Gbit/s networking connection is needed. There are, and have been, several sophisticated technologies that have offered the bandwidth and additional features to fill this role. Some such as Serial RapidIO® and Infiniband have been successful in application domains that Gigabit Ethernet could not address, and continue to make sense, but 10Gbit/s Ethernet will challenge their existence.

66AK2H12 (ACTIVE) Multicore DSP+ARM KeyStone II System-on-Chip (SoC) [TI.com, created on Nov 8, 2012]

Datasheet manual [351 pages]:

• 66AK2H14/12/06 Multicore DSP+ARM KeyStone II System-on-Chip (SoC) (Rev. E) (PDF , 8763 KB)   14 Nov 2013

More Literature:

• [Datasheet:] 66AK2Hx KeyStone Multicore DSP+ARM System-on-chips (Rev. A) (PDF , 193KB) 8 Nov 2013
• Multicore DSPs for High-Performance Video Coding 
  (PDF , 245KB) 22 Jan 2013
• Video Infrastructure – Applications of the K2E, K2H platforms 
  (PDF , 199KB) 9 Nov 2012
• Industrial Imaging: Applications of the K2H and K2E platforms 
  (PDF , 515) 9 Nov 2012

Description

The 66AK2Hx platform is TI’s first to combine the quad ARM® Cortex™-A15 MPCore™ processors with up to eight TMS320C66x high-performance DSPs using the KeyStone II architecture. Unlike previous ARM Cortex-A15 devices that were designed for consumer products, the 66AK2Hx platform provides up to 5.6 GHz of ARM and 11.2 GHz of DSP processing coupled with security and packet processing and Ethernet switching, all at lower power than multi-chip solutions making it optimal for embedded infrastructure applications like cloud computing, media processing, high-performance computing, transcoding, security, gaming, analytics and virtual desktop. Using TI’s heterogeneous programming runtime software and tools, customers can easily develop differentiated products with 66AK2Hx SoCs.

Taking Multicore to the Next Level: KeyStone II Architecture [Texas Instruments YouTube channel, Feb 26, 2012]

Introducing the EVMK2H [Texas Instruments YouTube channel, Nov 15, 2013]

Kick start development of high performance compute systems with TI’s new KeyStone™ SoC and evaluation module [TI press release, Nov 14, 2013]

Combination of DSP + ARM® cores and high-speed peripherals offer developers an optimal compute solution at low power consumption

DALLAS, Nov. 14, 2013 /PRNewswire/ — Further easing the development of processing-intensive applications, Texas Instruments (TI) (NASDAQ: TXN) is unveiling a new system-on-chip (SoC), the 66AK2H14, and evaluation module (EVM) for its KeyStone^TM-based 66AK2Hx family of SoCs. With the new 66AK2H14 device, developers designing high-performance compute systems now have access to a 10Gbps Ethernet switch-on-chip. The inclusion of the 10GigE switch, along with the other high-speed, on-chip interfaces, saves overall board space, reduces chip count and ultimately lowers system cost and power. The EVM enables developers to evaluate and benchmark faster and easier. The 66AK2H14 SoC provides industry-leading computational DSP performance at 307 GMACS/153 GFLOPS and 19600 DMIPS of ARM performance, making it ideal for a wide variety of applications such as video surveillance, radar processing, medical imaging, machine vision and geological exploration.

“Customers today require increased performance to process compute-intensive workloads using less energy in a smaller footprint,” said Paul Santeler, vice president and general manager, Hyperscale Business, HP. “As a partner in HP’s Moonshot ecosystem dedicated to the rapid development of new Moonshot servers, we believe TI’s KeyStone design will provide new capabilities across multiple disciplines to accelerate the pace of telecommunication innovations and geological exploration.”

Meet TI’s new 10Gbps Ethernet DSP + ARM SoC

TI’s newest silicon variant, the 66AK2H14, is the latest addition to its high-performance 66AK2Hx SoC family which integrates multiple ARM Cortex™-A15 MPCore™ processors and TI’s fixed- and floating-point TMS320C66x digital signal processor (DSP) generation cores. The 66AK2H14 offers developers exceptional capacity and performance (up to 9.6 GHz of cumulative DSP processing) at industry-leading size, weight, and power. In addition, the new SoC features a wide array of unique high-speed interfaces, including PCIe, RapidIO, Hyperlink, 1Gbps and 10Gbps Ethernet, achieving total I/O throughput of up to 154Gbps. These interfaces are all distinct and not multiplexed, allowing designers tremendous flexibility with uncompromising performance in their designs.

Ease development and debugging with TI’s tools and software

TI helps simplify the design process by offering developers highly optimized software for embedded HPC systems along with development and debugging tools for the EVMK2H – all for under $1,000. The EVMK2H features a single 66AK2H14 SoC, a status LCD, two 1Gbps Ethernet RJ-45 interfaces and on-board emulation. An optional EVM breakout card (available separately) also provides two 10Gbps Ethernet optical interfaces for 20Gbps backplane connectivity and optional wire rate switching in high density systems.

The EVMK2H is bundled with TI’s Multicore Software Development Kit (MCSDK), enabling faster development with production ready foundational software. The MCSDK eases development and reduces time to market by providing highly-optimized bundles of foundational, platform-specific drivers, optimized libraries and demos.

Complementary analog products to increase system performance

TI offers a wide range of power management and analog signal chain components to increase the system performance of 66AK2H14 SoC-based designs. For example, the TPS53xx integrated FET DC/DC converters provide the highest level of power conversion efficiency even at light loads, while the LM10011 VID converter with dynamic voltage control helps reduce system power consumption. The CDCM6208 low-jitter clock generator also eliminates the need for external buffers, jitter cleaners and level translators.

Availability and pricing

TI’s EVMK2H is available now through TI distribution partners or TI.com for $995. In addition to TI’s Linux distribution provided in the MCSDK, Wind River^® Linux is available now for the 66AK2Hxx family of SoCs. Green Hills^® INTEGRITY^® RTOS and Wind River VxWorks^® RTOS support will each be available before the end of the year. Pricing for the 66AK2H14 SoC will start at $330 for 1 KU. The 10Gbps Ethernet breakout card will be available from Mistral.

Ask the Expert: How can developers accelerate scientific computing with TI’s multicore DSPs? [Texas Instruments YouTube channel, Feb 7, 2012]

Ask the Expert: Arnon Friedmann [Texas Instruments YouTube channel, Sept 6, 2012]

End of Updates as of Dec 6, 2013

---

The original post (8 months ago):

HP Moonshot: Designed for the Data Center, Built for the Planet [HP press kit, April 8, 2013]

On April 8, 2013, HP unveiled the world’s first commercially available HP Moonshot system, delivering compelling new infrastructure economics by using up to 89 percent less energy, 80 percent less space and costing 77 percent less, compared to traditional servers. Today’s mega data centers are nearing a breaking point where further growth is restricted due to the current economics of traditional infrastructure. HP Moonshot servers are a first step organizations can take to address these constraints.

For more details on the disruptive potential of HP Moonshot, visit TheDisruption.com

• Press Release: HP Launches New Class of Server for Social, Mobile, Cloud and Big Data

Introducing HP Moonshot [HewlettPackardVideos April 11, 2013]

HP’s Cutting Edge Data Center Innovation [Ramón Baez, Senior Vice President and Chief Information Officer (CIO) of HP, HP Next [launched on April 2], April 10, 2013]

This is an exciting time to be in the IT industry right now. For those of you who have been around for a while — as I have — there have been dramatic shifts that have changed how businesses operate.

From the early days of the mainframes, to the explosion of the Internet and now social networks, every so often very important game-changing innovation comes along. We’re in the midst of another sea change in technology.

Inside HP IT, we are testing the company’s Moonshot servers. With these servers running the same chips found in smart phones and tablets, they are using incredibly less power, require considerably less cooling and have a smaller footprint.

We currently are running some of our intensive hp.com applications on Moonshot and are seeing very encouraging results. Over half a billion people will visit hp.com this year, and the new Moonshot technology will run at a fraction of the space, power and cost – basically we expect to run HP.com off of the same amount of energy needed for a dozen 60-watt light bulbs.

This technology will revolutionize data centers.

Within HP IT, we are fortunate in that over the past several years we have built a solid data center foundation to run our company. Like many companies, we were a victim of IT sprawl — with more than 85 data centers in 29 countries. We decided to make a change and took on a total network redesign, cutting our principle worldwide data centers down to six and housing all of them in the United States.

With the addition of four new EcoPODs to our infrastructure and these new Moonshot servers, we are in the perfect position to build out our private cloud and provide our businesses with the speed and quality of innovation they need.

Moonshot is just the beginning.The product roadmap for Moonshot is extremely promising and I am excited to see what we can do with it within HP IT, and what benefits our customers will see.

What Calxeda is saying about HP Moonshot [HewlettPackardVideos YouTube channel, April 8, 2013] which is best to start with for its simple and efficient message, as well as what Intel targeting ARM based microservers: the Calxeda case [‘Experiencing the Cloud’ blog, Dec 14, 2012] already contained on this blog earlier:

Then we can turn to the Moonshot product launch by HP 2 days ago:

Note that the first three videos following here were released 3 days later, so don’t be surpised by YouTube dates, in fact the same 3 videos (as well as the “Introducing HP Moonshot” embedded above) were delivered on April 8 live webcast, see the first 18 minutes of that, and then follow according HP’s flow of the presentation if you like. I would certainly recommend my own presentation compiled here.

HP president and CEO Meg Whitman on the emergence of a new style of IT [HewlettPackardVideos YouTube channel, April 11, 2013]

EVP and GM of HP’s Enterprise Group Dave Donatelli discusses HP Moonshot [HewlettPackardVideos YouTube channel, April 11, 2013]

Tour the Houston Discovery Lab — where the next generation of innovation is created [HewlettPackardVideos YouTube channel, April 11, 2013]

A new era of accelerated innovation [HP Moonshot minisite, April 8, 2013]

Cloud, Mobility, Security, and Big Data are transforming what the business expects from IT resulting in a “New Style of IT.” The result of alternative thinking from a proven industry leader, HP Moonshot is the world’s first software defined server that will accelerate innovation while delivering breakthrough efficiency and scale.

Watch the unveiling [link to HP Moonshot – The Disruption [HP Event registration page at ‘thedisruption.com’]

• Products
• Services & Support
• Partners
• Customers
• Videos

On the right is the Moonshot System with the very first Moonshot servers (“microservers/server appliances” as called by the industry) based on Intel® Atom S1200 processors and for supporting web-hosting workloads (see also on right part  of the image below). Currently there is also a storage cartridge (on the left of the below image) and a multinode for highly dense computing solutions (see in the hands of presenter on the image below). Many more are to come later on.

With up to a 180 servers inside the box (45 now) it was necessary to integrate network switching. There are two sockets (see left) for the network switch so you can configure for redundancy. The downlink module which talks to the cartridges is on left of the below image. This module is paired with an uplink module (see on the middle of the below image as taken out, and then shown with the uplink module on the right) that is in the back of the server. There will be more options available.

More information:
– Enterprise Information Library for Moonshot
– HP Moonshot System [Technical white paper from HP, April 5, 2013] from which I will include here the following excerpts for more information:

HP Moonshot 1500 Chassis

The HP Moonshot 1500 Chassis is a 4.3U form factor and slides out of the rack on a set of rails like a file cabinet drawer. It supports 45 HP ProLiant Moonshot Servers and an HP Moonshot-45G Switch Module that are serviceable from the top.

It is a modern architecture engineered for the new style of IT that can support server cartridges, server and storage cartridges, storage only cartridges and a range of x86, ARM or accelerator based processor technologies.

As an initial offering, the HP Moonshot 1500 Chassis is fully populated 45 HP ProLiant Moonshot Servers and one HP Moonshot-45G Switch Module and a second HP Moonshot-45G Switch Module can be purchased as an option. Future offerings will include quad server cartridges and will result in up to 180 servers per chassis. The 4.3U form factor allows for 10 chassis per rack, which with the quad server cartridge amounts to 1800 servers in a single rack.

The Moonshot 1500 Chassis simplifies management with four iLO processors that share management responsibility for the 45 servers, power, cooling, and switches.

…

Highly flexible fabric

Built into the HP Moonshot 1500 Chassis architecture are four separate and independent fabrics that support a range of current and future capabilities:

• Network fabric
• Storage fabric
• Management fabric
• Integrated cluster fabric

Network fabric

The Network fabric provides the primary external communication path for the HP Moonshot 1500 Chassis.

For communication within the chassis, the network switch has four communication channels to each of the 45 servers. Each channel supports a 1-GbE or 10-GbE interface. Each HP Moonshot-45G Switch Module supports 6 channels of 10GbE interface to the HP Moonshot-6SFP network uplink modules located in the rear of the chassis.

Storage fabric

The Storage fabric provides dedicated SAS lanes between server and storage cartridges. We utilize HP Smart Storage firmware found in the ProLiant family of servers to enable multiple core to spindle ratios for specific solutions. A hard drive can be shared among multiple server cartridges to enable low cost boot, logging, or attached to a node to provide storage expansion.

The current HP Moonshot System configuration targets light scale-out applications. To provide the best operating environment for these applications, it includes HP ProLiant Moonshot Servers with a hard disk drive (HDD) as part of the server architecture. Shared storage is not an advantage for these environments. Future releases of the servers thattarget different solutions will take advantage of the storage fabric.

Management fabric

We utilize the Integrated Lights-Out (iLO) application-specific integrated circuit (ASIC) standard in the HP ProLiant family of servers to provide the innovative management features in the HP Moonshot System. To handle the range of extreme low energy processors we provide a device neutral approach to management, which can be easily consumed by data center operators to deploy at scale.

The Management fabric enables management of the HP Moonshot System components as one platform with a dedicated iLO network. Benefits of the management fabric include:

• The iLO Chassis Manager aggregates data to a common set of management interfaces.
• The HP Moonshot 1500 Chassis has a single Ethernet port gateway that is the single point of access for the Moonshot Chassis manager.
• Intelligent Platform Management Interface (IPMI) and Serial Console for each server
• True out-of-band firmware update services
• SL-APM Rack Management spans rack or multiple racks

Integrated Cluster fabric

The Integrated Cluster fabric provides a high-speed interface among future server cartridge technologies that will benefit from high bandwidth node-to-node communication. North, south, east, and west lanes are provided between individual server cartridges.

The current HP ProLiant Moonshot Servertargets light scale-out applications. These applications do not benefit from the node-to-node communications, so the Integrated Cluster fabric is not utilized. Future releases of the cartridges that target different workloads that require low latency interconnects will take advantage of the Integrated Cluster fabric.

HP ProLiant Moonshot Server

HP will bring a growing library of cartridges, utilizing cutting-edge technology from industry leading partners. Each server will target specific solutions that support emerging Web, Cloud, and Massive-Scale Environments, as well as Analytics and Telecommunications. We are continuing server development for other applications, including Big Data, High-Performance Computing, Gaming, Financial Services, Genomics, Facial Recognition, Video Analysis, and more.

Figure 4. Cartridges target specific solutions

The first server cartridge now available is HP ProLiant Moonshot Server, which includes the Intel® Atom Processor S1260. This is a low power processor that is right-sized for the light workloads. It has dedicated memory and storage, with discrete resources. This server design is idealfor light scale-out applications. Light scale-out applications require relatively little processing but moderately high I/O and include environments that perform the following functions:

• Dedicated web hosting
• Simple content delivery

The HP ProLiant Moonshot Server can hot plug in the HP Moonshot 1500 Chassis. If service is necessary, it can be removed without affecting the other servers in the chassis. Table 1 defines the HP ProLiant Moonshot Server specifications.

Table 1. HP ProLiant Moonshot Server specifications

Processor	One Intel® Atom Processor S1260
Memory	8 GB DDR3 ECC 1333 MHz
Networking	Integrated dual-port 1Gb Ethernet NIC
Storage	500 GB or 1 TB HDD or SSD, non-hot-plug, small form factor
Operating systems	• Canonical Ubuntu 12.04 • Red Hat Enterprise Linux 6.4 • SUSE Linux Enterprise Server 11 SP2

With that HP CEO Seeks Turnaround Unveiling ‘Moonshot’ Super-Server: Tech [Bloomberg, April, 2013] as well as HP Moonshot: Say Goodbye to the Vanilla Server [Forbes, April 8, 2013]. HP however is much more eyeing the ARM based Moonshot servers which are expected to come later, because of the trends reflected on the left (source: HP). The software defined server concept is very general.

There are a number of quite different server cartridges expected to come, all specialised by server software installed on it. Typical specialised servers, for example, are the ones on which CyWee from Taiwan is working on with Texas Instruments’ new KeyStone II architecture featuring both ARM Cortex-A15 CPU cores and TI’s own C66x DSP cores for a mixture of up to 32 DSP and RISC cores in TI’s new 66AK2Hx family of SoCs, first of which is the TMS320TCI6636 implemented in 28nm foundry technology. Based on that CyWee will deliver multimedia Moonshot server cartridges for cloud gaming, virtual office, video conferencing and remote education (see even the first Keystone announcement). This CyWee involvement in HP Moonshot effort is part of HP’s Pathfinder Partner Program which Texas Instruments also joined recently to exploit a larger opportunity as:

TI’s 66AK2Hx family and its integrated c66x multicore DSPs are applicable for workloads ranging from high performance computing, media processing, video conferencing, off-line image processing & analytics, video recorders (DVR/NVR), gaming, virtual desktop infrastructure and medical imaging.

But Intel was able to win the central piece of the Moonshot System launch (originally initiated by HP as the “Moonshot Project” in November 2011 for disruption in terms of power and TCO for servers, actually with a Calxeda board used for research and development with other partners), at least as it was productized just two days ago:
Raejeanne Skillern from Intel – HP Moonshot 2013 – theCUBE [siliconangle YouTube channel]

However ARM was not left out either just relegated in the beginning to highly advanced and/or specialised server roles with its SoC partners, and coming later in the year:

• Applied Micro with networking and connectivity background having now the X-Gene ARM 64-bit Server on a Chip platform as well which features 8 ARM 64-bit high-performance cores developed from scratch according to an architecture license (i.e. not ARM’s own Cortex-A50 series core), clocked at up to 2.4GHz and also has 4 smaller cores for network and storage offloads (see AppliedMicro on the X-Gene ARM Server Platform and HP Moonshot [SiliconANGLE blog [April 9, 2013]). Sample reference boards to key customers were shipped in March (see Applied Micro’s cloud chip is an ARM-based, switch-killing machine [GigaOM, April 3, 2013]). In the latest X-Gene Arrives in Silicon [Open Compute Summit Winter 2013 presentation, Jan 16, 2013] video you can have the most recent strategic details (upto 2014 with FinFET implementation of a “Software defined X-Gene based data center components”, should be assumed that at 16nm). Here I will include a more product-oriented AppliedMicro Shows ARM 64-bit X-Gene Server on a Chip Hardware and Software [Charbax YouTube channel, Nov 3, 2012] overview video:
  
  Vinay Ravuri, Vice President and General Manager, Server Products at AppliedMicro gives an update on the 64bit ARM X-Gene Server Platform. At ARM Techcon 2012, AppliedMicro, ARM and several open-source software providers gave updates on their support of the ARM 64-bit X-Gene Server on a Chip Platform.

  More information: A 2013 Resolution for the Data Center [Applied Micro on Smart Connected Devices blog from ARM, Feb 4, 2013] about “plans from Oracle, Red Hat, Citrix and Cloudera to support this revolutionary architecture … Dell’s “Iron” server concept with X-Gene … an X-Gene based ARM server managed by the Dell DCS Software suite …” etc.

• Texas Instruments with digital signal processing (DSP) background, as it was already presented above. 
• Calxeda with integration of storage fabric and Internet switching background, with details coming later, etc.:

This is what is empasized by Lakshmi Mandyam from ARM – HP Moonshot 2013 – theCUBE [siliconangle YouTube channel, April 8, 2013]

She is also mentioning in the talk the achievements which could put ARM and its SoC partners into a role which Intel now has with its general Atom S1200 based server cartridge product fitting into the Moonshot system. Perspective information on that is already available on my ‘Experiencing the Cloud’ blog here:
– The state of big.LITTLE processing [April 7, 2013]
– The future of mobile gaming at GDC 2013 and elsewhere [April 6, 2013]
– TSMC’s 16nm FinFET process to be further optimised with Imagination’s PowerVR Series6 GPUs and Cadence design infrastructure [April 8, 2013]
– With 28nm non-exclusive in 2013 TSMC tested first tape-out of an ARM Cortex™-A57 processor on 16nm FinFET process technology [April 3, 2013]

The absence of Microsoft is even more interesting as AMD is also on this Moonshot bandwagon: Suresh Gopalakrishnan from AMD – HP Moonshot 2013 – theCUBE [siliconangle YouTube channel, April 8, 2013]

already showing a Moonshot fitting server cartridge with AMD’s four next-generation SoCs (while Intel’s already productized cartridge is not yet at an SoC level). We know from CES 2013 that AMD Unveils Innovative New APUs and SoCs that Give Consumers a More Exciting and Immersive Experience [press release, Jan 7, 2013] with the:

“Temash” … elite low-power mobility processor for Windows 8 tablets and hybrids … to be the highest-performance SoC for tablets in the market, with 100 percent more graphics processing performance² than its predecessor (codenamed “Hondo.”)

“Kabini” [SoC which] targets ultrathin notebooks with exceptional battery life and offers impressive levels of performance in both dual- and quad-core options. “Kabini” is expected to deliver an increase of more than 50 percent in performance³ over the previous generation of AMD essential computing APUs (codenamed “Brazos 2.0.”)

Both APUs are scheduled to ship in the first half of 2013

so AMD is really close to a server SoC to be delivered soon as well.

The “more information” sections which follow her are:

1. The Announcement
2. Software Partners
3. Hardware Partners

---

1. The Announcement

HP Moonshot [MultiVuOnlineVideo YouTube channel, April 8, 2013]

HP Launches New Class of Server for Social, Mobile, Cloud and Big Data [press release, April 8, 2013]

Software defined servers designed for the data center and built for the planet

… Built from HP’s industry-leading server intellectual property (IP) and 10 years of extensive research from HP Labs, the company’s central research arm, HP Moonshot delivers a significant improvement in energy, space, cost and simplicity. …

…

The HP Moonshot system consists of the HP Moonshot 1500 enclosure and application-optimized HP ProLiant Moonshot servers. These servers will offer processors from multiple HP partners, each targeting a specific workload.

With support for up to 1,800 servers per rack, HP Moonshot servers occupy one-eighth of the space required by traditional servers. This offers a compelling solution to the problem of physical data center space.⁽³⁾ Each chassis shares traditional components including the fabric, HP Integrated Lights-Out (iLo) management, power supply and cooling fans. These shared components reduce complexity as well as add to the reduction in energy use and space.  

The first HP ProLiant Moonshot server is available with the Intel® Atom S1200 processor and supports web-hosting workloads. HP Moonshot 1500, a 4.3u server enclosure, is fully equipped with 45 Intel-based servers, one network switch and supporting components.

HP also announced a comprehensive roadmap of workload-optimized HP ProLiant Moonshot servers incorporating processors from a broad ecosystem of HP partners including AMD, AppliedMicro, Calxeda, Intel and Texas Instruments Incorporated.

Scheduled to be released in the second half of 2013, the new HP ProLiant Moonshot servers will support emerging web, cloud and massive scale environments, as well as analytics and telecommunications. Future servers will be delivered for big data, high-performance computing, gaming, financial services, genomics, facial recognition, video analysis and other applications.

…

The HP Moonshot system is immediately available in the United States and Canada and will be available in Europe, Asia and Latin America beginning next month.

Pricing begins at $61,875 for the enclosure, 45 HP ProLiant Moonshot servers and an integrated switch.⁽⁴⁾

…

⁽⁴⁾ Estimated U.S. street prices. Actual prices may vary.

More information:
– HP Moonshot System [Family data sheet, April 8, 2013]
– HP Moonshot – The Disruption [HP Event registration page at ‘thedisruption.com’ with embedded video gallery, press kit and more, originally created on April 12, 2010, obviously updated for the April 8, 2013 event]

Moonshot 101 [HewlettPackardVideos YouTube channel, April 8, 2013]

Alert for Microsoft:

[4:42] We defined the industry standard server market [reference to HP’s Compaq heritage] and we’ve been the leader for years. With Moonshot we bring to find the market and taking it to the next level. [4:53]

People Behind HP Moonshot [HP YouTube channel, April 10, 2013]

HP Moonshot System Tour [HewlettPackardVideos YouTube channel, April 8, 2013]

HP Moonshot System is Hot Pluggable [HewlettPackardVideos YouTube channel, April 8, 2013]

Alert for Microsoft: how and when will you have a system like this with all the bells and whistles as presented above, as well as the rich ecosystem of hardware and software partners given below 

HP Pathfinder Innovation Ecosystem [HewlettPackardVideos YouTube channel, April 8, 2013]

Software partners:

What Linaro is saying about HP Moonshot [HewlettPackardVideos YouTube channel, April 8, 2013]

Alert for Microsoft:

[0:11] In HP approach Linaro is about forming an enterprise group. What they were hoping for, what’s happened is to get a bunch of companies who are interested in taking the ARM architecture into the server space. [0:26]

Canonical joins Linaro Enterprise Group (LEG) and commits Ubuntu Hyperscale Availability for ARM V8 in 2013 [press release, Nov 1, 2012]

• Canonical continues its leadership of commercial deployment for ARM-based servers through membership of Linaro Enterprise Group (LEG)
• Ubuntu, the only commercially supported OS for ARM v7 today, commits to support ARM v8 server next year
• Ubuntu extends its position as the natural choice for hyperscale  server computing with long term support

… “Canonical has been supporting our work optimising and consolidating the Linux kernel since our founding in June 2010”, said George Grey, CEO of Linaro. “We’re very happy to welcome them as a member of the Linaro Enterprise Group, building on our relationship to help accelerate development of the ARM server software ecosystem.” …

… “Calxeda has been thrilled with Canonical’s leadership in developing the ARM ecosystem”,  said Karl Freund, VP marketing at Calxeda. “These guys get it. They are driving hard and fast, already delivering enterprise-class code and support for Calxeda’s 32-bit product today to our mutual clients.  Working together in LEG will enable us to continue to build on the momentum we have already created.” …

What Canonical is saying about HP Moonshot [HewlettPackardVideos YouTube channel, April 8, 2013]

HP Moonshot and Ubuntu work together [Ubuntu partner site, April 9, 2013]

… Ubuntu, as the lead operating system platform for x86 and ARM-based HP Moonshot Systems, featured extensively at the launch of the program in April 2013. …

… Ubuntu Server is the only OS fully operational today across HP Moonshot x86 and ARM servers, launched in April 2013.

Ubuntu is recognised as the leader in scale out and Hyperscale. Together, Canonical and HP are delivering massive reductions in data-center energy, space and costs. …

“Canonical has been working with HP for the past two years
on HP Moonshot, and with Ubuntu, customers can achieve higher performance with greater manageability across both x86 and ARM chip sets” Paul Santeler, VP & GM, Hyperscale Business Unit, HP

…

Ubuntu & HP’s project Moonshot [Canonical blog, Nov 2, 2011]

Today HP announced Project Moonshot  – a programme to accelerate the use of low power processors in the data centre.

The three elements of the announcement are the launch of Redstone – a development platform that harnesses low-power processors (both ARM & x86),  the opening of the HP Discovery lab in Houston and the Pathfinder partnership programme.

Canonical is delighted to be involved in all three elements of HP’s Moonshot programme to reduce both power and complexity in data centres.

The HP Redstone platform unveiled in Palo Alto showcases HP’s thinking around highly federated environments and Calxeda’s EnergyCore ARM processors. The Calxeda system on chip (SoC) design is powered by Calxeda’s own ARM based processor and combines mobile phone like power consumption with the attributes required to run a tangible proportion of hyperscale data centre workloads.

The promise of server grade SoC’s running at less than 5W and achieving per rack density of 2800+ nodes is impressive, but what about the software stacks that are used to run the web and analyse big data – when will they be ready for this new architecture?

Ubuntu Server is increasingly the operating system of choice for web, big data and cloud infrastructure workloads. Films like Avatar are rendered on Ubuntu, Hadoop is run on it and companies like Rackspace and HP are using Ubuntu Server as the foundation of their public cloud offerings.

The good news is that Canonical has been working with ARM and Calxeda for several years now and we released the first version of Ubuntu Server ported for ARM Cortex A9 class  processors last month.

The Ubuntu 11.10 release (download) is an functioning port and over the next six months and we will be working hard to benchmark and optimize Ubuntu Server and the workloads that our users prioritize on ARM.  This work, by us and by upstream open source projects is going to be accelerated by today’s announcement and access to hardware in the HP Discovery lab.

As HP stated today, this is beginning of a journey to re-inventing a power efficient and less complex data center. We look forward to working with HP and Calxeda on that journey.

The biggest enterprise alert for Microsoft because of what was discussed in Will Microsoft Stand Out In the Big Data Fray? [Redmondmag.com, March 22, 2013]: What NuoDB is saying about HP Moonshot [HewlettPackardVideos YouTube channel, April 9, 2013] especially as it is a brand new offering, see NuoDB Announces General Availability of Industry’s First & Only Cloud Data Management System at Live-Streamed Event [press release, Jan 15, 2013] now available in archive at this link: http://go.nuodb.com/cdms-2013-register-e.html

Extreme density on HP’s Project Moonshot [NuoDB Techblog, April 9, 2013]

A few months ago HP came to us with something very cool. It’s called Project Moonshot, and it’s a new way of thinking about how you design infrastructure. Essentially, it’s a composable system that gives you serious flexibility and density.

A single Moonshot System is 4.3u tall and holds 45 independent servers connected to each other via 1-Gig Ethernet. There’s a 10-Gig Ethernet interface to the system as a whole, and management interfaces for the system and each individual server. The long-term design is to have servers that provide specific capabilities (compute, storage, memory, etc.) and can scale to up to 180 nodes in a single 4.3u chassis.

The initial system, announced this week, comes with a single server configuration: an Intel Atom S1260 processor, 8 Gigabytes of memory and either a 200GB SSD or a 500GB HDD. On its own, that’s not a powerful server, but when you put 45 of these into a 4.3 rack-unit space you get something in aggregate that has a lot of capacity while still drawing very little power (see below). The challenge, then, is how to really take advantage of this collection of servers.

NuoDB on Project Moonshot: Density and Efficiency

We’ve shown how NuoDB can scale a single database to large transaction rates. For this new system, however, we decided to try a different approach. Rather than make a single database scale to large volume we decided to see how many individual, smaller databases we could support at the same time. Essentially, could we take a fully-configured HP Project Moonshot System and turn it into a high-density, low-power, easy to manage hosting appliance.

To put this in context, think about a web site that hosts blogs. Typically, each blog is going to have a single database supporting it (just like this blog you’re reading). The problem is that while a few blogs will be active all the time, most of them see relatively light traffic. This is known as a long-tail pattern. Still, because the blogs always need to be available, so too the backing databases always need to be running.

This leads to a design trade-off. Do you map the blogs to a single database (breaking isolation and making management harder) or somehow try to juggle multiple database instances (which is hard to automate, expensive in resource-usage and makes migration difficult)? And what happens when a blog suddenly takes off in popularity? In other words, how do you make it easy to manage the databases and make resource-utilization as efficient as possible so you don’t over-spend on hardware?

As I’ve discussed on this blog NuoDB is a multi-tenant system that manages individual databases dynamically and efficiently. That should mean that we’re a perfect fit for this very cool (pun intended) new system from HP.

The Design

After some initial profiling on a single server, we came up with a goal: support 7,200 active databases. You can read all about how we did the math, but essentially this was a balance between available CPU, Memory, Disk and bandwidth. In this case a “database” is a single Transaction Engine and Storage Manager pair, running on one of the 45 available servers.

When we need to start a database, we pick the server that’s least-utilized. We choose this based on local monitoring at each server that is rolled up through the management tier to the Connection Brokers. It’s simple to do given all that NuoDB already provides, and because we know what each server supports it lets us calculate a single capacity percentage.

It gets better. Because a NuoDB database is made of an agile collection of processes, it’s very inexpensive to start or stop a database. So, in addition to monitoring for server capacity we also watch what’s going on inside each database, and if we think it’s been idle long enough that something else could use the associated resources more effectively we shut it down. In other words, if a database isn’t doing anything active we stop it to make room for other databases.

When an SQL client needs to access that database, we simply re-start it where there are available resources. We call this mechanism hibernating and waking a database. This on-demand resource management means that while there are some number of databases actively running, we can really support a much larger in total (remember, we’re talking about applications that exhibit a long-tail access pattern). With this capability, our original goal of 7,200 active databases translates into 72,000 total supported databases. On a single 4.3u System.

The final piece we added is what we call database bursting. If a single database gets really popular it will start to take up too many resources on a single server. If you provision another server, separate from the Moonshot System, then we’ll temporarily “burst” a high-activity database to that new host until activity dies down. It’s automatic, quick and gives you on-demand capacity support when something gets suddenly hot.

The Tests

I’m not going to repeat too much here about how we drove our tests. That’s already covered in the discussion on how we’re trying to design a new kind of benchmark focused on density and efficiency. You should go check that out … it’s pretty neat. Suffice it say, the really critical thing to us in all of this was that we were demonstrating something that solves a real-world problem under real-world load.

You should also go read about how we setup and ran on a Moonshot System. The bottom-line is that the system worked just like you’d expect, and gave us the kinds of management and monitoring features to go beyond basic load testing.

The Results

We were really lucky to be given access to a full Moonshot System. It gave us a chance to test out our ideas, and we actually were able to do better than our target. You can see this in the view from our management interface running against a real system under our benchmark load. You can see there that when we hit 7200 active databases we were only at about 70% utilization, so there was a lot more room to grow. Huge thanks to HP for giving us time on a real Moonshot System to see all those idea work!

Something that’s easy to lose track of in all this discussion is the question of power. Part of the value proposition from Project Moonshot is in energy efficiency, and we saw that in spades. Under load a single server only draws 18 Watts, and the system infrastructure is closer to 250 Watts. Taken together, that’s a seriously dense system that is using very little energy for each database.

Bottom Line

We were psyched to have the chance to test on a Moonshot System. It gave us the chance to prove out ideas around automation and efficiency that we’ll be folding into NuoDB over the next few releases. It also gave us the perfect platform to put our architecture through its paces and validate a lot about the flexibility of our core architecture.

We’re also seriously impressed by what we experienced from Project Moonshot itself. We were able to create something self-contained and easy to manage that solves a real-world problem. Couple that with the fact that a Moonshot System draws so little power, the Total Cost of Ownership is impressively low.  That’s probably the last point to make about all this: the combination of our two technologies gave us something where we could talk concretely about capacity and TCO, something that’s usually hard to do in such clear terms.

In case it’s not obvious, we’re excited. We’ve already been posting this week about some ideas that came out of this work, and we’ll keep posting as the week goes on. Look for the moonshot tag and please follow-up with comments if you’re curious about anything specific and would like to hear more!

Project Moonshot by the Numbers [NuoDB Techblog, April 9, 2013]

To really understand the value from HP Project Moonshot you need to think beyond the list price of one system and focus instead on the Total Cost of Ownership. Figuring out the TCO for a server running arbitrary software is often a hard (and thankless?) task, so one of the things we’ve tried to do is not just demonstrate great technology but something that naturally lets you think about TCO in a simple way. We think the final metrics are pretty simple, but to get there requires a little math.

Executive Summary

If you’re a CIO, and just want to know the bottom line, then we’ll ruin the suspense and cut to the chase. It will cost you about $70,500 up-front, $1,800 in your first year’s electricity bills and take 8.3 rack-units to support the web-front end and database back-end for 72,000 blogs under real-world load.

Cost of a Single Database

Recall that we set the goal at 72,000 databases within a single system. At launch the list price for a fully-configured Moonshot System is around $60,000, so we start out at 83 cents per-database. In practice were seeing much higher capacity in our tests, but let’s start with this conservative number.

Now consider the power used by the system. From what we’ve measured through the iLO interfaces a single server draws no more than 18 Watts at peak load (measured against CPU and IO activity). The System itself (fans, switches etc.) draws around 250 Watts in our tests. That means that under full load each database is drawing about .015 Watts.

NuoDB is a commercial software offering, which means that you pay up-front to deploy the software (and get support as part of that fee). For anyone who wants to run a Moonshot System in production as a super-dense NuoDB appliance we’ll offer you a flat-rate license.

Put together, we can say that the cost per database-watt is 1.22 cents. That’s on a 4.3 rack-unit system. Awesome.

Quantify the Supported Load

As we discussed in our post on benchmarking, we’re trying to test under real-world load. As a simple starting-point we chose a profile based on WordPress because it’s fairly ubiquitous and has somewhat serious transactional requirements. In our benchmarking discussion we explain that a typical application action (post, read, comment) does around 20 SQL operations.

Given 72,000 databases most of these are fairly inactive, so on average we’ll say that each database gets about 250 hits a day (generous by most reports I’ve seen). That’s 18,000,000 hits a day or 208 hits per-second. 4,166 SQL statements a second isn’t much for a single database, but it’s pretty significant given that we’re spreading it across many databases some of which might have to be “woken” on-demand.

HP was generous enough not only to give us time on a Moonshot System but also access to some co-located servers for driving our load tests. In this case, 16 lower-powered ARM-based Calxeda systems that all went through the same 1-Gig ethernet connection to our Moonshot System. These came from HP’s Discovery Lab; check out our post about working with the Moonshot System for more details.

From these load-drivers we able to run our benchmark application with up to 16 threads per server, simulating 128 simultaneous clients. In this case a typical “client” would be a web server trying to respond to a web client request. We averaged around 320 hits per-second, well above the target of 208. From what we could observe, we expect that given more capable network and client drivers we would be able to get 3 or 4 times that rate easily.

Tangible Cost

We have the cost of the Moonshot System itself. We also know that it can support expected load from a fairly small collection of low-end servers. In our own labs we use systems that cost around $10,000, fit in 3 rack-units and would be able to drive at least the same kind of load we’re citing here. Add a single switch at around $500 and you have a full system ready to serve blogs. That’s $70,500 total in 8.3 rack units, still under $1 per database.

I don’t know what power costs you have in your data center, but I’ve seen numbers ranging from 2.5 to 25 cents per Kilowatt-Hour. In our tests, where we saw .015 Watts per-database, if you assume an average rate of 13.75 cents per KwH that comes out to .00020625 cents per-hour per-database in energy costs. In one year, with no down-time, that would cost you $1,276.77 in total electricity fees.

Just as an aside, according to the New York Times, Facebook uses around 60,000,000 Watts a year!

One of the great things about a Moonshot System is that the 45 servers are already being switched inside the chassis. This means that you don’t need to buy switches & cabling, and you don’t need to allocate all the associated space in your racks. For our systems administrator that alone would make him very happy.

Intangible Cost

What I haven’t been talking about in all of this are the intangible costs. This is where figuring out TCO becomes harder.

For instance, one of the value-propositions here is that the Moonshot System is a self-contained, automated component. That means that systems administrators are freed up from the tasks of figuring out how to allocate and monitor databases, and how to size the data-center for growth. Database developers can focus more easily on their target applications. CIOs can spend less time staring at spreadsheets … or, at least, can allocate more time to spreadsheets on different topics.

Providing a single number in terms of capacity makes it easy to figure out what you need in your datacenter. When a single server within a Moonshot System fails you can simply replace it, and in the meantime you know that the system will still run smoothly just with slightly lower capacity. From a provisioning point of view, all you need to figure out is where your ceiling is and how much stand-by capacity you need to have at the ready.

NuoDB by its nature is dynamic, even when you’re doing upgrades. This means that you can roll through a running Moonshot System applying patches or new versions with no down-time. I don’t know how you calculate the value in saved cost here, but you probably do!

Comparisons and Planned Optimizations

It’s hard to do an “apples-to-apples” comparison against other database software here. Mostly, this is because other databases aren’t designed to be dynamic enough to support hibernation, bursting and capacity-based automated balancing. So, you can’t really get the same levels of density, and a lot of the “intangible” cost benefits would go away.

Still, to be fair, we tried running MySQL on the same system and under the same benchmarks. We could indeed run 7200 instances, although that was already hitting the upper-bounds of memory/swap. In order to get the same density you would need 10 Moonshot Systems, or you would need larger-powered expensive servers. Either way, the power, density, automation and efficiency savings go out the window, and obviously there’s no support for bursting to more capable systems on-demand.

Unsurprisingly, the response time was faster on-average (about half the time) from MySQL instances. I say “unsurprisingly” for two reasons. First, we tried to use schema/queries directly from WordPress to be fair in our comparison, and these are doing things that are still known to be less-optimized in NuoDB. They’re also in the path of what we’re currently optimizing and expect to be much faster in the near-term.

The second is that NuoDB clients were originally designed assuming longer-running connections (or pooled connections) to databases that always run with security & encryption enabled. We ran all of our tests in our default modes to be fair. That means we’re spending more time on each action setting up & tearing down a connection. We’ve already been working on optimizations here that would shrink the gap pretty substantially.

In the end, however, our response time is still on the order of a few hundred milliseconds worst-case, and is less important than the overall density and efficiency metrics that we proved out. We think the value in terms of ease of use, density, flexibility on load spikes and low-cost speaks for itself. This setup is inexpensive by comparison to deploying multiple servers and supports what we believe is real-world load. Just wait until the next generation of HP Project Moonshot servers roll out and we can start scaling out individual databases at the same time!

More information:
– Benchmarking Density & Efficiency [NuoDB Techblog, April 9, 2013]
– Database Hibernation and Bursting [NuoDB Techblog, April 8, 2013]
– An Enterprise Management UI for Project Moonshot [NuoDB Techblog, April 9, 2013]Regarding the cloud based version of NuoDB see:
– NuoDB Partners with Amazon [press release, March 26, 2013]
– NuoDB Extends Database Leadership in Scalability & Performance on a Private Cloud [press release, March 14, 2013] “… the industry’s first and only patented, elastically scalable Cloud Data Management System (CDMS), announced performance of 1.84 million transactions per second (TPS) running on 32 machines. … With NuoDB Starlings release 1.0.1, available as of March 1, 2013, the company has made advancements in performance and scalability and customers can now experience 26% improvement in TPS per machine.”
– Google Compute Engine: interview with NuoDB [GoogleDevelopers YouTube channel, March 21, 2013]

Actually Calxeda was best to explain the preeminence of software over the SoC itself:
Karl Freund from Calxeda – HP Moonshot 2013 – theCUBE [siliconangle YouTube channel, April 8, 2013], see also HP Moonshot: It’s a lot closer than it looks! [Calxeda’s ‘ARM Servers, Now!’ blog, April 8, 2013]

as well as ending with Calxeda’s very practical, gradual approach to ARM based served market with things like:

[16.03] Our 2nd generation platform called Midway, which will be out later this year [in the 2nd half of the year], that’s probably the target for Big Data. Our current product is great for web serving, it’s great for media serving, it’s great for storage. It doesn’t have enough memory for Big Data … in a large. So we’ll getting that 2nd generation product out, and that should be a really good Big Data platform. Why? Because it’s low power, it’s low cost, but it’s also got a lot of I/O. Big Data is all that moving a lot of data around. And if you do that more cost effectively you save a lot of money. [16:38]

mentioning also that their strategy is using standard ARM cores like the Cortex-A57 for their H1 2014 product, and focus on things like the fabric and the management, which actually allows them to work with a streamlined staff of around 150 people.

Detailed background about Calxeda in a concise form:
– Redefining Datacenter Efficiency: An Overview of Calxeda’s architecture and early performance measurements [Karl Freund, Nov 12, 2012] from where the core info is:

• Founded in 2008   
• $103M Funding       
• 1st Product Announced with HP,  Nov  2011   
• Initial Shipments in Q2 2012   
• Volume production in Q4 2012

* The power consumed under normal operating conditions
under full application load (ie, 100% CPU utilization)

A small Calxeda Cluster: a Simple Example
• Start with four ServerNodes
• Consumes only 20W total power   
• Connected via distributed fabric switches   
• Connect up to 4 SATA drives per node   
• Then scale this to thousands of ServerNodes

EnergyCard: a Quad-Node Reference Design

• Four-node reference platform from Calxeda
• Available as product and/or design
• Plugs into OEM system board with passive fabric, no additional switch HW
  EnergyCard delivers 80Gb Bandwidth to the system board. (8 x 10Gb links)

It is also important to have a look at what were the Open Source Software Packages for Initial Calxeda Shipments [Calxeda’s ‘ARM Servers, Now!’ blog, May 24, 2012]

We are often asked what open-source software packages are available for initial shipments of Calxeda-based servers.

Here’s the current list (changing frequently).  Let us know what else you need!

Then Perspectives From Linaro Connect [Calxeda’s ‘ARM Servers, Now!’ blog, March 20, 2013] sheds more light on the recent software alliances which make Calxeda to deliver:

– From Larry Wikelius,   Co-Founder and VP Ecosystems,  Calxeda:

The most recent Linaro Connect (Linaro Connect Asia 2013 – LCA), held in Hong Kong the first week of March, really put a spotlight on the incredible momentum around ARM based technology and products moving into the Data Center.  Yes – you read that correctly – the DATA CENTER!

When Linaro was originally launched almost three years ago the focus was exclusively on the mobile and client market – where ARM has and continues to be dominant.  However, as Calxeda has demonstrated, the opportunity for the ARM architecture goes well beyond devices that you carry in your pocket.  Calxeda was a key driver in the formation of the Linaro Enterprise Group (LEG), which was publicly launched at the previous LinaroConnect event in Copenhagen in early November, 2012.

LEG has been an exciting development for Linaro and now has 13 member companies that include server vendors such as Calxeda, Linux distribution companies Red Hat and Canonical, OEM representation from HP and even Hyperscale Data Center end user Facebook.  There were many sessions throughout the week that focused on Server specific topics such as UEFI, ACPI, Virtualization, Hyperscale Testing with LAVA and Distributed Storage.  Calxeda was very active throughout the week with the team participating directly in a number of roadmap definition sessions, presenting on Server RAS and providing guidance in key areas such as application optimization and compiler focus for Servers.

Linaro Connect is proving to be a tremendous catalyst for the the growing eco-system around the ARM software community as a whole and the server segment in particular.  A great example of this was the keynote presentation given jointly by Mark Heath and Lars Kurth from Citrix on Tuesday morning.  Mark is the VP of XenServer at Citirix and Lars is well know in the OpenSource community for his work with Xen.  The most exciting announcement coming out of Mark’s presentation is that Citrix will be joining Linaro as a member of LEG.  Citrix will be certainly prove to be another valuable member of the Linaro team and during the week attendees were able to appreciate how serious Citrix is about supporting ARM servers.  The Xen team has not only added full support for ARM V7 systems in the Xen 4.3 release but they have accomplished some very impressive optimizations for the ARM platform.  The Xen team has leveraged Device Tree for optimal device discovery.  Combined with a number of other code optimizations they showed a dramatically smaller code base for the ARM platform.  We at Calxeda are thrilled to welcome Citrix into LEG!

As an indication of the draw that the Linaro Connect conference is already having on the broader industry the Open Compute Project (OCP) held their first International Event co-incident with LCA at the same venue.  The synergy between Linaro and OCP is significant with the emphasis on both organizations around Open Source development (one software and one hardware) along with the dramatically changing design points for today’s Hyperscale Data Center.  In fact the keynote at LCA on Wednesday morning really put a spotlight on how significant this is likely to be.  Jason Taylor, Director of Capacity Engineering and Analysis at Facebook, presented on Facebook’s approach to ARM based servers.   Facebook’s consumption of Data Center equipment is quite stunning – Jason quoted from Facebook’s 10-Q filed in October 2012 which stated that “The first nine months of 2012 … $1.0 billion for capital expenditures” related to data center equipment and infrastructure.  Clearly with this level of investment Facebook is extremely motivated to optimize where possible.  Jason focused on the strategic opportunity for ARM based severs in a disaggregated Data Center of the future to provide lower cost computing capabilities with much greater flexibility.

Calxeda has been very active in building the Server Eco-System for ARM based servers.  This week in Hong Kong really underscored how important that investment has become – not just for Calxeda but for the industry as a whole. Our commitment to Open Source software development in general and Linaro in particular has resulted in a thriving Linux Infrastructure for ARM servers that allows Calxeda to leverage and focus on key differentiation for our end users.  The Open Compute Project, which we are an active member in and have contributed to key projects such as the Knockout Storage design as well as the Open Slot Specification, demonstrates how the combination of an Open Source approach for both Software and Hardware can compliment each other and can drive Data Center innovation.  We are early in this journey but it is very exciting!

Calxeda will continue to invest aggressively in forums and industry groups such as these to drive the ARM based server market.  We look forward to continue to work with the incredibly innovative partners that are members in these groups and we are confident that more will join this exciting revolution.  If you are interested in more information on these events and activities please reach out to us directly at info@calxeda.com.

The next Linaro Connnect is scheduled for early July in Dublin. We expect more exciting events and topics there and hope to see you there!

They are also referring on their blog to Mobile, cloud computing spur tripling of micro server shipments this year [IHS iSuppli press release, Feb 6, 2013] which showing the general market situation well into the future as:

Driven by booming demand for new data center services for mobile platforms and cloud computing, shipments of micro servers are expected to more than triple this year, according to an IHS iSuppli Compute Platforms Topical Report from information and analytics provider IHS (NYSE: IHS).

Shipments this year of micro servers are forecast to reach 291,000 units, up 230 percent from 88,000 units in 2012. Shipments of micro servers commenced in 2011 with just 19,000 units. However, shipments by the end of 2016 will rise to some 1.2 million units, as shown in the attached figure.

The penetration of micro servers compared to total server shipments amounted to a negligible 0.2 percent in 2011. But by 2016, the machines will claim a penetration rate of more than 10 percent—a stunning fiftyfold jump.

Micro servers are general-purpose computers, housing single or multiple low-power microprocessors and usually consuming less than 45 watts in a single motherboard. The machines employ shared infrastructure such as power, cooling and cabling with other similar devices, allowing for an extremely dense configuration when micro servers are cascaded together.

“Micro servers provide a solution to the challenge of increasing data-center usage driven by mobile platforms,” said Peter Lin, senior analyst for compute platforms at IHS. “With cloud computing and data centers in high demand in order to serve more smartphones, tablets and mobile PCs online, specific aspects of server design are becoming increasingly important, including maintenance, expandability, energy efficiency and low cost. Such factors are among the advantages delivered by micro servers compared to higher-end machines like mainframes, supercomputers and enterprise servers—all of which emphasize performance and reliability instead.”

Server Salad Days

Micro servers are not the only type of server that will experience rapid expansion in 2013 and the years to come. Other high-growth segments of the server market are cloud servers, blade servers and virtualization servers.

The distinction of fastest-growing server segment, however, belongs solely to micro servers.

The compound annual growth rate for micro servers from 2011 to 2016 stands at a remarkable 130 percent—higher than that of the entire server market by a factor of 26. Shipments will rise by double- and even triple-digit percentages for each year during the period.

Key Players Stand to Benefit

Given the dazzling outlook for micro servers, makers with strong product portfolios of the machines will be well-positioned during the next five years—as will their component suppliers and contract manufacturers.

A slew of hardware providers are in line to reap benefits, including microprocessor vendors like Intel, ARM and AMD; server original equipment manufacturers such as Dell and Hewlett-Packard; and server original development manufacturers including Taiwanese firms Quanta Computer and Wistron.

Among software providers, the list of potential beneficiaries from the micro server boom extends to Microsoft, Red Hat, Citrix and Oracle. For the group of application or service providers that offer micro servers to the public, entities like Amazon, eBay, Google and Yahoo are foremost.

The most aggressive bid for the micro server space comes from Intel and ARM.

Intel first unveiled the micro server concept and reference design in 2009, ostensibly to block rival ARM from entering the field.

ARM, the leader for many years in the mobile world with smartphone and tablet chips because of the low-power design of its central processing units, has been just as eager to enter the server arena—dominated by x86 chip architecture from the likes of Intel and a third chip player, AMD. ARM faces an uphill battle, as the majority of server software is written for x86 architecture. Shifting from x86 to ARM will also be difficult for legacy products.

ARM, however, is gaining greater support from software and OS vendors, which could potentially put pressure on Intel in the coming years.

Read More > Micro Servers: When Small is the Next Big Thing

Then there are a number of Intel competitive posts on Calxeda’s ‘ARM Servers, Now!’ blog:
– What is a “Server-Class” SOC? [Dec 12, 2012]
– Comparing Calxeda ECX1000 to Intel’s new S1200 Centerton chip [Dec 11, 2012]
which you can also find in my Intel targeting ARM based microservers: the Calxeda case [‘Experiencing the Cloud’ blog, Dec 14, 2012] with significantly wider additional information upto binary translation from x86 to ARM with Linux

See also:
– ARM Powered Servers: 2013 is off to a great start & it is only March! [Smart Connected Devices blog of ARM, March 6, 2013]
– Moonshot – a shot in the ARM for the 21st century data center [Smart Connected Devices blog of ARM, April 9, 2013]
– Are you running out of data center space? It may be time for a new server architecture: HP Moonshot [Hyperscale Computing Blog of HP, April 8, 2013]
– HP Moonshot: the HP Labs team that did some of the groundbreaking research [Innovation @ HP Labs blog of HP, April 9, 2013]
– HP Moonshot: An Accelerator for Hyperscale Workloads [Moor Insights White Paper, April 8, 2013]
– Comparing Pattern Mining on a Billion Records with HP Vertica and Hadoop [HP Vertica blog, April 9, 2013] by team of HP Labs researchers show how the Vertica Analytics Platform can be used to find patterns from a billion records in a couple of minutes, about 9x faster than Hadoop.
– PCs and cloud clients are not parts of Hewlett-Packard’s strategy anymore [‘Experiencing the Cloud’, Aug 11, 2011 – Jan 17, 2012] see the Autonomy IDOL related content there
– ENCO Systems Selects HP Autonomy for Audio and Video Processing [HP Autonomy press release, April 8, 2013]

HP Autonomy today announced that ENCO Systems, a global provider of radio automation and live television audio solutions, has selected Autonomy’s Intelligent Data Operating Layer (IDOL) to upgrade ENCO’s latest-generation enCaption product.

ENCO Systems provides live automated captioning solutions to the broadcast industry, leveraging technology to deliver closed captioning by taking live audio data and turning it into text. ENCO Systems is capitalizing on IDOL’s unique ability to understand meaning, concepts and patterns within massive volumes of spoken and visual content to deliver more accurate speech analytics as part of enCaption3.

“Many television stations count on ENCO to provide real-time closed captioning so that all of their viewers get news and information as it happens, regardless of their auditory limitations,” said Ken Frommert, director, Marketing, ENCO Systems. “Autonomy IDOL helps us provide industry-leading automated closed captioning for a fraction of the cost of traditional services.”

enCaption3 is the only fully automated speech recognition-based closed captioning system for live television that does not require speaker training. It gives broadcasters the ability to caption their programming, including breaking news and weather, any time, day or night, since it is always on and always available. enCaption3 provides captioning in near real time-with only a 3 to 6 second delay-in nearly 30 languages.

“Television networks are under increasing pressure to provide real-time closed captioning services-they face fines if they don’t, and their growing and diverse viewers demand it,” said Rohit de Souza, general manager, Power, HP Autonomy. “This is another example of a technology company integrating Autonomy IDOL to create a stronger, faster and more accurate product offering, and demonstrates yet another powerful way in which IDOL can be applied to help organizations succeed in the human information era.”

– Using Big Data to change the game in the Energy industry [Enterprise Services Blog of HP, Oct 24, 2012]

… Tools like HP’s Autonomy that analyzes the unstructured data found in call recordings, survey responses, chat logs, e-mails, social media posts and more. Autonomy’s Intelligent Data Operating Layer (IDOL) technology uses sophisticated pattern-matching techniques and probabilistic modeling to interpret information in much the same way that humans do. …

– Stouffer Egan turns the tables on computers in keynote address at HP Discover [Enterprise Services Blog of HP, June 8, 2012]

For decades now, the human mind has adjusted itself to computers by providing and retrieving structured data in two-dimensional worksheets with constraints on format, data types, list of values, etc. But, this is not the way the human mind has been architected to work. Our minds have the uncanny ability to capture the essence of what is being conveyed in a facial expression in a photograph, the tone of voice or inflection in an audio and the body language in a video. At the HP Discover conference, Autonomy VP for United States, Stouffer Egan showed the audience how software can begin to do what the human mind has being doing since the dawn of time. In a demonstration where Iron Man came live out of a two-dimensional photograph, Egan turned the tables on computers. It is about time computers started thinking like us rather than us forcing us to think like them.

Egan states that the “I” in IT is where the change is happening. We have a newfound wealth of data through various channels including video, social, click stream, audio, etc. However, data unprocessed without any analysis is just that — raw data. For enterprises to realize business value from this unstructured data, we need tools that can process it across multiple media. Imagine software that recognizes the picture in a photograph and searches for a video matching the person in the picture. The cover page of a newspaper showing a basketball star doing a slam dunk suddenly turns live pulling up the video of this superstar’s winning shot in last night’s game. …

---

2. Software Partners

HP Moonshot is setting the roadmap for next generation data centers by changing the model for density, power, cost and innovation. Ubuntu has been designed to meet the needs of Hyperscale customers and, combined with its management tools, is ideally suited be the operating system platform for HP Moonshot. Canonical has been working with HP since the beginning of the Moonshot Project, and Ubuntu is the only OS integrated and fully operational across the complete Moonshot System covering x86 and ARM chip technologies. What Canonical is saying about HP Moonshot	As mobile workstyles become the norm, the scalability needs of today’s applications and devices are increasingly challenging what traditional infrastructures can support. With HP’s Moonshot System, customers will be able to rapidly deploy, scale, and manage any workload with dramatically lower space and energy constraints. The HP Pathfinder Innovation Ecosystem is a prime opportunity for Citrix to help accelerate the development of innovative solutions that will benefit our enterprise cloud, virtualization and mobility customers.
We’re committed to helping enterprises achieve the most from their Big Data initiatives. Our partnership with HP enables joint customers to keep and query their data at scale so they can ask bigger questions and get bigger answers. By using HP’s Moonshot System, our customers can benefit from the improved resource utilization of next generation data center solutions that are workload optimized for specific applications.
Today’s interactive applications are accessed 24×365 by millions of web and mobile users, and the volume and velocity of data they generate is growing at an unprecedented rate. Traditional technologies are hard pressed to keep up with the scalability and performance demands of these new applications. Couchbase NoSQL database technology combined with HP’s Moonshot System is a powerful offering for customers who want to easily develop interactive web and mobile applications and run them reliably at scale.	Our partnership with HP facilitates CyWee’s goal of offering solutions that merge the digital and physical worlds. With TI’s new SoCs, we are one step closer to making this a reality by pushing state-of-the-art video to specialized server environments. Together, CyWee and HP will deliver richer multimedia experiences in a variety of cloud-based markets, including cloud gaming, virtual office, video conferencing and remote education.
HP’s new Moonshot System will enable organizations to increase the energy efficiency of their data centers while reducing costs. Our Cassandra-based database platform provides the massive scalability and multi-datacenter capabilities that are a perfect complement to this initiative, and we are excited to be working with HP to bring this solution to a wide range of customers.	Big data comes in a wide range for formats and types and is a result of the connected everything world we live in. Through Project Moonshot, HP has enabled a new class of infrastructure to run more efficient workloads, like Apache Hadoop, and meet the market demand of more performance for less.
The unprecedented volume and variety of data introduces unique challenges to organizations today… By combining the HP Moonshot system with Autonomy IDOL’s unique ability to understand concepts in information, organizations can dramatically reduce the cost, space, and energy requirements for their big data initiatives, and at the same time gain insights that grow revenue, reduce risk, and increase their overall Return on Information.	Big Data is not just for Big Companies – or Big Servers – anymore – it’s affecting all sectors of the market. At HP Vertica we’re very excited about the work we’ve been doing with the Moonshot team on innovative configurations and types of analytic appliances which will allow us to bring the benefits of real-time Big Data analytics to new segments of the market. The combination of the HP Vertica Analytics Platform and Moonshot is going to be a game-changer for many.
HP worked closely with Linaro to establish the Linaro Enterprise Group (LEG). This will help accelerate the development of the software ecosystem around ARM Powered servers. HP’s Moonshot System is a great platform for innovation – encouraging a wide range of silicon vendors to offer competing ‘plug-and-play’ server solutions, which will give end users maximum choice for all their different workloads. What Linaro is saying about HP Moonshot[HewlettPackardVideos YouTube channel, April 8, 2013]	Organizations are looking for ways to rapidly deploy, scale, and manage their infrastructure, with an architecture that is optimized for today’s application workloads. HP Moonshot System is an energy efficient, space saving, workload-optimized solution to meet these needs, and HP has partnered with MapR Technologies, a Hadoop technology leader, to accelerate innovation and deployment of Big Data solutions.
NuoDB and HP are shattering the scalability and density barriers of a traditional database server. NuoDB on the HP Moonshot System delivers unparalleled database density, where customers can now run their applications across thousands of databases on a single box, significantly reducing the total cost across hardware, software, and power consumption. The flexible architecture of HP Moonshot coupled with NuoDB’s hyper-pluggable database design and its innovative “database hibernation” technology makes it possible to bring this unprecedented hardware and software combination to market. What NuoDB is saying about HP Moonshot [HewlettPackardVideos YouTube channel, April 9, 2013]	As the leading solution provider for the hosting market, Parallels is excited to be collaborating in the HP Pathfinder Innovation Ecosystem. The HP Moonshot System in concert with Parallels Plesk Panel and Parallels Containers provides a flexible and efficient solution for cloud computing and hosting.
Red Hat Enterprise Linux on HP’s converged infrastructure means predictability, consistency and stability. Companies around the globe rely on these attributes when deploying applications every day, and our value proposition is just as important in the Hyperscale segment. When customers require a standard operating environment based on Red Hat Enterprise Linux, I believe they will look to the HP Moonshot System as a strong platform for high-density Hyperscale implementations. What Red Hat is saying about HP Moonshot [HewlettPackardVideos YouTube channel, April 8, 2013]	HP Project Moonshot’s promise of extreme low-energy servers is a game changer, and SUSE is pleased to partner with HP to bring this new innovation to market. For more than twenty years, SUSE has adapted its enterprise-grade Linux operating system to achieve ever-increasing performance needs that succeed both today and tomorrow in areas such as Big Data and cloud computing. What SUSE is saying about HP Moonshot [HewlettPackardVideos YouTube channel, April 8, 2013]

---

3. Hardware Partners

AMD is excited to continue our deep collaboration with HP to bring extreme low-energy, ultra dense, specialized server solutions to the market. Both companies share a passion to bring innovative workload optimized solutions to the market, enabling customers to scale-out to new levels within existing energy and space constraints. The new low-power x86 AMD Opteron™ APU is optimized in the HP Moonshot System to dramatically lower TCO in quickly emerging media oriented workloads. What AMD is saying about HP Moonshot	It is exciting to see HP take the lead in innovating low-energy servers for the cloud. Applied Micro’s ARM 64-bit X-Gene Server on a Chip will enable performance levels seen in today’s deployments while offering higher densities, greatly improved I/O, and substantial reductions in the total cost of ownership. Together, we will unleash innovation unlike anything we’ve seen in the server market for decades. What Applied Micro is saying about HP Moonshot
In the current economic and power realities, today’s server infrastructure cannot meet the needs of the next billion data users, or the evolving needs of currently supported users. Customers need innovative SoC solutions which deliver more integration and optimization than has historically been required by traditional enterprise workloads. HP’s Moonshot System is a departure from the one size fits all approach of traditional enterprise and embraces a range of ARM partner solutions that address different performance, workloads and cost points. What ARM is saying HP Moonshot	Calxeda and HP’s new Moonshot System are a powerful combination, and sets a new standard for ultra-efficient web and application serving. Fulfilling a journey started together in November 2011, Project Moonshot creates the foundation for the new age of application-specific computing. What Calxeda is saying about HP Moonshot
HP Moonshot System is a game changer for delivering optimized server solutions. It beautifully balances the need for mixing different processor solutions optimized for different workloads under a standard hardware and software framework. Cavium’s Project Thunder will provide a family of 64-bit ARM v8 processors with dense and scalable sever class performance at extremely attractive power and cost metrics. We are doing this by blending performance and power efficient compute, high performance memory and networking into a single, highly integrated SoC. What Cavium is saying about HP Moonshot	Intel is proud to deliver the only server class, 64-bit SoC technology that powers the first and only production shipping HP ProLiant Moonshot Server today. 64-bit Intel Atom processor S1200 family features extreme low power combined with required datacenter class capabilities for lightweight web scale workloads, such as low end dedicated hosting and static web serving. In collaboration with HP, we have a strong roadmap of additional server solutions shipping later this year, including Intel’s 2nd generation 64-bit SoC, “Avoton” based on leading 22nm manufacturing technology, that will deliver best in class energy efficiency and density for HP Moonshot System. What Intel is saying about HP Moonshot
	What Marvell is saying about HP Moonshot
HP Moonshot System’s high density packaging coupled with integrated network capability provides the perfect platform to enable HP Pathfinder Innovation Ecosystem partners to deliver cutting edge technology to the hyper-scale market. SRC Computers is excited to bring its history of delivering paradigm shifting high-performance, low-power, reconfigurable processors to HP Project Moonshot’s vision of optimizing hardware for maximum application performance at lowest TCO. What SRC Computers is saying about HP Moonshot	The scalability and high performance at low power offered through HP’s Moonshot System gives customers an unmatched ability to adapt their solutions to the ever-changing and demanding market needs in the high performance computing, cloud computing and communications infrastructure markets. The strong collaboration efforts between HP and TI through the HP Pathfinder Innovation Ecosystem ensure that customers understand and get the most benefit from the processors at a system-level. What TI is saying about HP Moonshot