Not logged in » Login
Jan 14 2014

Next-generation Scale-out Servers: the Fujitsu PRIMERGY CX400 S2


As the digital revolution proceeds, concepts like High Performance Computing (HPC), Big Data, and "the Cloud" continue to get more popular on a daily basis. For most observers, however, these concepts are associated with giant data centers set up in some greenfield rather than with services that could be deployed within existing infrastructures. Fujitsu's second-generation, multi-node scale-out system PRIMERGY CX400 S2 is about to radically change that belief.

Although concepts such as Big Data, HPC, and Cloud Computing have rapidly gained popularity over the past ten years, many users still have a somewhat limited understanding of the infrastructures that "loom large" behind these application scenarios. Their general idea seems to be one of giant cluster servers residing in secluded spots that have developed intelligence worthy of the infamous HAL 9000 in Kubrick's 2001: A Space Odyssey. Little do they realize that nowadays such clusters can be built almost anywhere, and that it doesn't take the proverbial evil genius to come up with the idea. After all, the steady increase in computing power and storage capacities, coupled with a constant decrease in form factors, has led to a point where many small and medium enterprises process more data than some large corporations did a decade or two ago. In other words, the above scenarios literally start "at home" and usually serve rational purposes. Moreover, with hardware extensions like GPGPUs even comparatively pint-sized machines can carry out the most demanding tasks.

Fujitsu PRIMERGY Scale-out Servers
A popular example for this combination of maximum performance and small footprint was Fujitsu's first generation of PRIMERGY CX400 Scale-out Servers. First launched at the CeBIT 2012, they already showed most of the main characteristics that still define this server category today: increased platform density, improved energy efficiency, better management effectiveness, and overall lower investments. The new generation builds on these virtues and expands them, mainly by incorporating fresher, more powerful hardware. Plus, it finally breathes life into a feature that was part of the product line all along, but so far had less impact than originally intended: expandability. Let's take a look at the details.

Like its predecessor, the Fujitsu PRIMERGY CX400 S2 consists of a chassis that either has room for four 1U, half-wide dual socket server nodes or for two 2U units of the same type. These nodes – the PRIMERGY CX250 S2 and CX270 S2 – serve up the latest Xeon® E5-2600 v2 processor technology with up to 12 cores.

Do more with less
But a powerful CPU alone might not make the CX400 a suitable system for deployment in Big Data, Cloud, and HPC environments. That's why these CX400 nodes come with a number of other features. First and foremost, the systems support a maximum amount of RAM per CPU and node of 512/1,024 GB. As a result, the 4-node configuration comes with a maximum of 4 TB of main memory.

Further features include security enhancements that lay the foundation for a safer computing environment. Most of these functions, namely Intel® Secure Key and Intel® OS Guard, are baked into the chips and serve to isolate VMs from each other, ensure a system hasn't been tampered with, accelerate encryption, and protect the server against attacks based on OS vulnerabilities. With these characteristics plus 10 GbE or 56Gb InfiniBand interconnects, the four-server configuration of PRIMERGY CX400 S2 with CX250 nodes is perfectly positioned to carry out distributed parallel processing in Big Data settings, e.g. as slave nodes in Apache Hadoop installations.

Designed for HPC
Various HPC scenarios, however, require a different setup, and that's where the two-server configurations with CX270 nodes and additional GPGPUs or coprocessors come into play. Customers who buy the latest CX400 scale-out servers can choose between Tesla K20/K20X cards that offer 3.95 and 3.52 single precision and 1.31/1.17 teraflops double precision floating point peak performance. Or they can opt for Intel's Xeon® Phi 5110P or 7120P coprocessors instead with 1.98/2.4 gigaflops peak performance for single precision and 0.99/1.2 gigaflops peak performance for double precision floating point calculations. The idea here was to expand the power and capabilities of a Xeon® CPU by adding a coprocessor that understands code written for the original platform, so that highly parallel applications can be carried out using a regular x86 architecture. As a result, the new CX270 nodes can be used for applications like weather forecasts, finite element analysis, financial services, and ray tracing. To see the quality of the results, check out the design study at the top of this page, created by our long-term customer Realtime Technology (RTT).

To learn more, see the PRIMERGY Scale-out Servers microsite and the product page.

The PRIMERGY scale-out servers are suitable for various usage scenarios. Did you already consider offering a PRIMERGY CX400 to your customer? If yes, for which projects?


Florian Frimmel


About the Author:

Florian Frimmel

Global Product Marketing Manager, responsible for PRIMERGY tower and scale-out servers, at Fujitsu 


Comments on this article

No comments yet.

Please Login to leave a comment.


Please login

Please log in with your Fujitsu Partner Account.


» Forgot password

Register now

If you do not have a Fujitsu Partner Account, please register for a new account.

» Register now