ARM Community: ARM Powered Servers: 2013 is off to a great start & it is only March! - ARM Community

Open Compute Summit, Linley Data Center Conference, Univa Enterprise Workload Management and Baidu – The World’s First Commercial ARM Based Server Deployment!

Data centers are not what they used to be, and that’s a good thing! And the next generation of data centers will be quite different from today. That’s because we’re entering an era of unprecedented innovation and disruption focused on improving efficiency while reducing acquisition and operating costs. In parallel, new use cases are emerging that require server deployments and infrastructure on a scale that not too long ago was unimaginable. People are updating, and in some cases quite literally, rewriting the rule books on server, networking and data center design.

Let’s use Facebook as an example, now a household name providing a service (that is mission critical if you’re a teenager!) allowing friends, families, colleagues and acquaintances to connect and share as much or as little about their daily lives as they see fit. Sounds quite straight forward right? However dealing with all those simultaneous users, complex connections, frequent updates and the masses of associated data (recent photo stats: 350M/day, 7PB/month!) has led Facebook to be one of the most innovative and sophisticated data center operators in existence today.

Facebook: What if hardware were open?
The data center is at the heart of Facebook’s business and they realize there’s massive value in optimizing that resource for the task at hand. That last sentence is what makes the Open Compute Project (OCP) so important – in 2011 Facebook posed the audacious question “What if hardware were open?” and the Open Compute Project Foundation was formed to provide a structure in which individuals and organizations can share their intellectual property.

Open Compute Summit IV
With 2012 barely in the rear view mirror the Open Compute team were back at it putting the finishing touches to what would prove to be the largest and most comprehensive Open Compute Summit to date. Over 1500 people gathered at the Santa Clara Convention Center to hear about the latest best practices and innovations destined for the next generation data center. This Open Compute Summit was particularly important for ARM and our partners – during Frank Frankovsky’s keynote, ARM was announced (alongside Calxeda & Tilera) as a new members of the foundation! From there things got even more interesting with initial foundation contribution announcements and products showcased from multiple ARM Partners:

• Calxeda announced and contributed the design for their ‘Knockout’ board which adds fully fledged server capability to the Open Vault storage platform SAS expander. Calxeda also announced support for the OCP Common Slot
• AppliedMicro’s CEO Paramesh Gopi presented the X-Gene™ Server on a Chip™ Platform. AppliedMicro showcased their X-Compute™ and X-Memory™ platforms in addition to announcing the contribution of their ARM-based Common Slot Specification to the Open Compute Project.
• Dell’s Jimmy Pike presented a heterogeneous server management solution using a prototype server chassis featuring the Applied Micro X-Gene platform (which was also showcased at the Dell booth).
• Facebook published their Common Slot specification and showed a backplane design populated with 4 x ARM-based servers and 4 x ‘x86’-based servers (being careful to point out they were being fair using 4 of each).
• George Grey, CEO of Linaro presented an overview of Open Source challenges, how Linaro can help and touched on the previously announced Linaro Enterprise Group.

There were lots of great articles covering the Open Compute announcements – here are a select few for your reading pleasure:

• ARM friends Facebook at summit
• Facebook and Open Compute just blew up the server and disrupted a $55B market
• Facebook puts some brains in Open Vault JBOD storage
• Applied Micro ARM servers all over Open Compute Project

Linley Data Center Conference
Shortly after the Open Compute Summit Linley hosted their 2013 Data Center Conference in Santa Clara, ARM-based servers was a recurring theme among all the talk of new interconnects; disaggregation and software defined networking (SDN). The final session of the event titled “Designing Power-Efficient Servers” started with Linley Gwennap painting a picture of the current server processor landscape including current high performance and low power x86 offerings contrasted to the new entrants based on alternate architectures (covered by Rick Merritt of EETimes). This framed the presentations and panel discussion that would follow. Calxeda provided an update on their EnergyCore™ product and shared a collection of ‘performance-per-watt’ data across a variety of benchmarks and scale out workloads. AppliedMicro presented an overview of their X-Gene™ product and discussed the implications on future server and data center design. Then came the panel discussion – there’s clearly a lot of interest in ARM based servers based on the interaction, number of questions and comments from the audience. There was a strong alignment statement from another ARM Partner in attendance that lead Linley to quip “Maybe all the ARM server folks can stay behind for a Kum ba yah session at the end of the event” – we’ll take that as recognition of a healthy and active ecosystem, thanks Linley!

Baidu Rolls Out World’s First Commercial Deployment of Marvell’s ARM Processor-based Server
Initially reported a week earlier, Marvell officially announced first commercial deployments of ARM-based servers into the Baidu cloud. While we continue to believe 2014 is the year for volume shipments of ARM-based servers this is a major milestone in the journey towards realizing real world deployments. Some of you reading this may not know who Baidu is; well the simple way to describe them is as the Google of China. As you can imagine they have a large (and growing) data center infrastructure in order to handle all the indexing, search and cloud based services they offer to their customers (just as Google does). As the Marvell press release states “As the first organization in the world to leverage ARM servers for commercial use, Baidu is pioneering a new era of more cost-effective and environmentally friendly data centers that feature power-sipping consumption and greater performance. Marvell's implementation of ARM in its server SoC provides an easy to integrate, all-in-one solution that supports a new class of server demands in today's data center, especially given the growth of high-storage light-computing tasks for cloud-based services and applications, Web hosting and social sites. Baidu is using the new ARM servers in its personal cloud storage application named Baidu Cloud.” I chose to post this particular section of the press release as it does a good job of highlighting the key attributes of what ARM-based servers are all about: Pioneering, cost-effective, environmentally friendly, power-sipping & greater performance (for a specific workload). Congratulations to the Marvell and Baidu teams and everyone else involved – you are creating history here and contributing to the shift towards new data center thinking and ultimately reduced environmental impact which will be beneficial for generations to come, thank you!

Univa Announce Enterprise Workload Management for ARM based servers
Given all the cool hardware designs announced recently (pun intended) and the associated news it’s all too easy to forget that these platforms are all but useless without software! A large part of installing and managing these large data center installations revolves around workload management – Univa’s recently announced Grid Engine for ARM provides a distributed resource management platform that enables shared infrastructures. Sharing the infrastructure allows enterprises to more efficiently and rapidly deploy and operationalize mission critical distributed applications such as Hadoop (a Big Data software framework that supports data-intensive distributed applications). There’s a Beta release of Univa Grid Engine 8.1 available now – If you’d like to be one of the first to deploy Univa Grid Engine on an ARM-based cluster then check out the links below to download the beta and read more:

To download a beta version, go here.

More details on Univa Grid Engine can be found at www.univa.com.

As I stated up front – 2013 is off to a great start, and it’s only March! We’ve definitely entered a new era of unprecedented innovation and disruption in the data center and there’s so much more coming from the ARM ecosystem – things just got very interesting.

Jeff Underhill, Server Segment Marketing Manager, ARM, is based in Silicon Valley. After spending 10+ years working in the traditional server market Jeff saw an opportunity to revisit server design and redefine an industry. ARM’s business model enables innovative companies the freedom and choice to ask themselves “how would I architect highly efficient servers if I had a clean slate?” Consequently, he is helping drive ARM’s server program with a view to redefining the boundaries of traditional servers as opposed to simply replacing incumbent platforms.