Not logged in » Login
Mar 19 2015

Maximum Productivity in a 1U Housing: FUJITSU Server PRIMERGY RX2530 M1

/data/www/ctec-live/application/public/media/images/blogimages/34631_PRIMERGY_RX2530_M1_3.5-inch_open_2_scr.jpg

Performance and energy efficiency have long been widely accepted as key criteria for evaluating server systems. Lately, however, a third category has emerged: with more and more organizations being hard pressed for floor space, density – i.e. the ability to pack an unprecedented amount of high-end components into the smallest possible enclosure – is about to become an equally important factor.

In today's hyper-connected economy, virtually all data centers, regardless of their age or actual size, have one thing in common: they're too small. Or more precisely, they cannot continue to grow at the current pace, even if the amount of servers, storage and networking gear supposed to contain the ever-growing data flood does. The problem not only affects transnational corporations or omnipresent state agencies, whose headquarters and branch offices often tend to outgrow the communities they're located in, up to the point of requiring their own power and water supplies and waste management systems. In fact, it might be even more obvious for start-ups or small and medium enterprises that don't own large properties. Consequently, users and ICT vendors alike began to explore how they could pack more processing power into smaller structures and, ultimately, server units. The term density was coined to describe this approach – and has since turned into a value proposition that sparks the interest of pretty much every CIO or head of data center.

The Quest for Density
Historically, most of us first may have heard about density some five or six years ago, when people started asking how Facebook's or Google's servers could be so responsive that it almost seemed they knew search queries before they were even typed. The answer was that both companies' data centers strongly relied on custom-made server architectures that were based on standard components, but optimized for speed and provided higher core counts, more RAM, faster memory interfaces, and flash storage within tightly packed enclosures. Before that, similar arguments were heard when the first blade servers appeared in the early to mid-2000s. More generally speaking though, the concept of density has been around since at least the early 1970s and is basically a derivative of the ongoing miniaturization of processor layouts and manufacturing processes. Fujitsu's PRIMERGY RX2530 M1, which has been globally available since February, now re-introduces the idea to the classic rack server market in order to deliver maximum productivity in a 1U housing. Customers searching for configurations with even higher density may want to take a look at our scale-out system FUJITSU Server PRIMERGY CX400 M1.

Specifications
Technically, the PRIMERGY RX2530 M1 is a dual-socket server designed to run in various particularly demanding usage scenarios, ranging from fulfilling the hypervisor role in virtualized environments to HPC applications such as data analysis or simulation. To achieve this, our engineers bundled latest-generation Intel® Xeon® processor technology with numerous features not usually available with dual-socket systems. Here's an excerpt of a much longer list that can be found in the data sheet:

  • All systems are equipped with up to two processors from Intel's Xeon® E5-2600 v3 product family that offer a maximum of 18 cores each and can handle up to 36 threads thanks to HT; the total core/thread count for the maximum configuration is 36/72.
  • Each RX2530 M1 uses DDR4 modules as main memory; capacities currently range from 8 to 768 GB. The key advantage of DDR4 modules over their predecessors is that they offer much higher data transfer speeds of 2133 MT/s at just 1.2 volts – a 100% increase in bandwidth paired with a 20% decrease in voltage compared to standard DDR3 modules. Initially, a drop of just 0.3 volts doesn't sound like a lot; however, experts like Anandtech's Ian Cutress have pointed out that this translates into between 1 and 2 watts of power savings per module per system. In data center settings with dozens of servers, this allows for reasonable cuts of the power bill. Moreover, DDR4 technology can support up to 1.5 TB of main memory – a feature that will be added as soon as 64 GB memory modules appear on the market.
  • Internal storage capacities vary according to individual configuration. For customers who prefer standard 3.5 inch SAS or SATA HDDs, we offer systems with four drive bays and a maximum of 16 GB disk space. For more challenging scenarios, customers can pick from a host of 2.5 inch SAS or SATA SSDs and HDDs, PCIe SSDs, and SATA DOMs. These small form factor configurations can either hold four disks (upgradable to eight) or a maximum of ten drives, four of which may be PCIe SSDs. Maximum capacities range from 14.4 TB in a system with eight 2.5 inch SAS HDDs to 17.6 TB in a server equipped with four 2 TB PCIe and six 1.6 TB SAS SSDs. SATA DOMs are available with 32, 64 or 128 GB of storage space.
  • Major improvements also include the integration of two major Fujitsu technologies, DynamicLoM and Cool-safe® Advanced Thermal Design (Cool-safe® ATD). – DynamicLoM (cf. Fig. 1) is Fujitsu's version of the LAN-on-motherboard architecture that allows for the selection of an interface card that best meets the needs of a specific server setup. For this purpose, the RX2530 M1 uses Emulex XE CNA chips that connect to 2 x 1, 4 x 1, and 2 x 10 Gbit/s interface cards upon release; we expect to add 4 x 10 and 1 x 40 Gbit/s cards in the future. The advantages of DynamicLoM are manifold: on the one hand, it enables flexible scaling of data transfer/connection speeds to external systems; on the other, support staff no longer has to disassemble the entire server just to replace a LoM chip. What's more, DynamicLoM supports a variety of advanced data center technologies such as iSCSI, FCoE, network virtualization, Single Root I/O Virtualization, and RDMA over Converged Ethernet for live migration. – Available optionally at no extra charge, Cool-safe® ATD is an extension of our original Cool-safe® concept (involving heat pipes and straight-through cooling) that aims to improve the performance, efficiency and reliability of PRIMERGY servers even under today's adverse data center conditions. The main goal is to enable selected servers to operate within a temperature range between 5° C/ 41° F and 40° C/105° F – as opposed to 10 °C/50 °F and 35 °C /95 °F, which was the 'normal' range for older Cool-safe® servers. As a result, customers can put the servers into hotter-than-usual rooms without having to invest in additional cooling and cut the existing cooling system's power consumption by up to 20% for each 5° C/9° F temperature increase. A welcome side effect is that Cool-safe® ATD also aligns well with more modern cooling concepts such as fresh air cooling. With selected Fujitsu storage solutions and certified switches, the entire data center can thus be run at a higher temperature.

 

Fujitsu's DynamicLOM Architecture

 

Fig. 1: Fujitsu's DynamicLOM Architecture

 

  • As one of two dual-socket systems, the RX2530 M1 is available in a long lifecycle configuration. Selected models hereby become available for an extended time period, making it ideal for long roll-outs, for companies which require extensive certification work or which need codified system characteristics.
  • Like all of Fujitsu's x86-based servers, the PRIMERGY RX2530 M1 supports a broad selection of leading server operating systems (Windows Server 2008 R2/2012/2012 R2, several Linux flavors, VMware vSphere™, and Citrix® XenServer®) and comes equipped with a full-fledged management suite and comprehensive security features.

Conclusion
Fujitsu has long focused on building highly reliable x86 standard servers that are at the same time powerful, energy-efficient and adaptable to a wide variety of environments and usage scenarios. The FUJITSU Server PRIMERGY RX2530 M1 not only adheres to this concept, but adds in a number of substantial enhancements and effectively closes the gap between entry- and enterprise-level systems. The success of these efforts is underlined by a new SPECpower_ssj2008 energy efficiency benchmark result achieved in the class of 1U rack servers, which set a new industry-best level of efficiency with a score of 9811 overall ssj_ops/watt (operations per watt).1

[1] Competitive benchmark results stated above reflect results published on www.spec.org as of February 4, 2015.

Score of 9,811 overall ssj_ops/watt published by Fujitsu on the PRIMERGY RX2530 M1, see: http://spec.org/power_ssj2008/results/res2015q1/power_ssj2008-20150114-00683.html.

For the latest benchmark results, visit www.spec.org.

 

Marcel Schuster

 

About the Author:

Marcel Schuster

Senior Specialist Marketing Manager – Data Center Systems / Server

SHARE

Comments on this article

No comments yet.

Please Login to leave a comment.

X

Please login

Please log in with your Fujitsu Partner Account.

Login


» Forgot password

Register now

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

» Register now