Fujitsu
Not logged in » Login
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
Apr 16 2018

From Blades to Modular Eco-Systems: FUJITSU Server PRIMERGY CX400 M4 Paves the Way

/data/www/ctec-live/application/public/media/images/blogimages/image-cx400blog2_TC_april2018.jpg

From the early 2000s into the first half of the 2010s, blade systems were the platform of choice for many companies that had to shoulder large computing and storage requirements while at the same time coping with budget and space restrictions. Meanwhile, however, the former cost-cutters are facing serious technical challenges that in turn are likely to drive up infrastructure spending. The best way to avoid this predicament is to progressively retire blade systems and replace them with a set of modular servers that provide identical or better functionality while also allowing for efficient data center operation and management.

For roughly a decade, businesses from all industries would sooner or later hit upon blade systems when searching for proper hardware that would help them master extensive virtualization and consolidation projects or implement high availability as well as the use of data-rich HPC applications. That trend was particularly strong among large and midsize businesses, whose efforts to rebuild their IT infrastructures in a way that met the increasing requirements often ran into a brick wall of technical and economic challenges. Typical issues at hand included high levels of complexity, huge design efforts, long project duration, botched and/or delayed implementation, and high cost deviation. Against this backdrop, blade systems with their combination of server and storage blades and networking components, all contained within a single chassis, had to look like the perfect solution: Instead of undergoing the complicated and error-prone process of finding and configuring adequate kit, IT departments could simply use the new class of systems as building blocks, that way saving time, administrative efforts, and money. So, once they had these systems up and running, all was well again. Or was it?

Key Challenges for Blade Systems
As has been pointed out above, the rise and subsequent reign of blade systems largely coincided with the welcoming of virtualization into the mainstream of data center architecture and management concepts. By the mid-2000s, however, innovative as well as more traditional hard- and software vendors were looking into even more effective means to streamline IT operations and services. Once again, the idea was to help companies build platforms that enable employees to seamlessly access data, applications and systems hosted at one or more data centers. But this time around these efforts would not be hampered by technical boundaries – instead, all resources would be available for virtually all employees, regardless of their whereabouts. Consequently, infrastructure providers began to build modular ecosystems consisting of compute units that can be optimized to cover a wide variety of workloads and use cases, from everyday tasks through VDI and cloud scenarios to HPC and big data analytics.

Apart from such conceptual issues, blade systems are increasingly subject to several technological trends that will shape the marketplace over the next few years. For instance, ever-growing speed demands have created a need for more advanced I/O technologies that can handle larger amounts of data in less time, such as 32 Gbit/s Fibre Channel, 100 Gbit/s Ethernet or 100 Gbit/s InfiniBand. These technologies would need to be supported by blade midplanes, but to get there could cost the industry tons of money because it's very complicated to achieve reliable signal integrity. Moreover, the power consumption of regular dual-socket servers based on Intel® Xeon® processors is still pointing upward; today they typically eat up 165 watts in TDP. Assuming this trend keeps up, future blade system and blade server generations will sooner or later draw too much power and run a substantial risk of overheating, which in turn would either lead to configuration restrictions or to much lower density. Thus, a main advantage of blade systems could simply evaporate.

The FUJITSU Server PRIMERGY CX400 M4 Eco-System
So which alternative to a stagnating technology do Fujitsu and its channel partners have to offer? Cues can be taken from the early adopters of cloud computing, who usually relied on a combination of industry standard servers, capacious storage systems and fast networking gear that was then further tweaked to accommodate typical cloud-based services. However, such solutions are extremely difficult to build from scratch, and customers will be much better off with an 'integrated package' that combines the density and efficiency of blade-like servers with the simplicity and cost efficiency of rack-based systems. This is precisely what we're offering in the form of the PRIMERGY CX400 M4 Eco-system.

As has been pointed out before, "CX400" serves as an umbrella term that denotes an entire product line and derives from the 2U chassis for 19-inch server racks that may contain up to four server nodes and 24 drive bays for 2.5-inch HDDs or SSDs. In addition, the chassis holds all other relevant components, such as up to two redundant, hot-pluggable power supplies (1,600 or 2,400 watts), up to eight hot-plug and redundant fan modules and the possibility to use the infrastructure with standard air cooling or optionally advanced liquid cooling units (CX2550 and CX2570 only).

Image

Fig. 1: FUJITSU Server PRIMERGY CX400 M4 with four 1U, half-width server blades

By implementing this server type, data center managers can improve efficiency, server density and consolidation by a common infrastructure. The comprehensive ecosystem include a TOR Ethernet switch as well as Infrastructure Manager Software (ISM) that further expands the advantages of these systems and provides tailor-made computing performance for a variety of workloads.

To populate the enclosure, customers can choose between the following server nodes that may either perform identical tasks in parallel or execute different workloads independently:

  • The FUJITSU Server PRIMERGY CX2550 M4 is a 1U, half-width, dual-socket server node that utilizes the compute power and flexibility of Intel's Xeon® Processor Scalable Family combined with up to 2 TB of DDR4 memory and up to two storage drives. More specifically, it's equipped with Platinum- or Gold-class CPUs that feature up to 28 cores and run at maximum clock speeds of 3.7 GHz. On top of that, they support a number of advanced technologies, such as Ultra-Path Interconnect (UPI, 3 links), to facilitate increased data exchange rates between CPUs in multi-node systems or the AVX-512 instruction set, which boosts the performance of protection mechanisms built into the processor hardware. As a result, the PRIMERGY CX2550 M 4 is the optimal system for cloud and HPC scenarios.
  • The FUJITSU Server PRIMERGY CX2560 M4 was developed for IT departments that struggle with space restrictions, but still wish to implement the new levels of efficiency and density modern server nodes can provide. The top-of-the-line processor for this half-width 1U model is an Intel® Xeon® Platinum 8164 processor with 26 cores. Moreover, each CX2560 M4 comes with up to six storage drives per node and a choice of connectivity options including OCP onboard LAN and DynamicLOM (4 x 1 Gbit/s or 2 x 10 Gbit/s Ethernet with RJ45 connectors, alternatively 2 x 10 Gbit/s and 4 x 10 Gbit/s Ethernet with SFP+ interface). Thanks to this broad range of configuration options, this node server will easily cover tasks like web serving, dedicated hosting, infrastructure virtualization and analytics or run mainstream enterprise applications, especially when paired with a full 2 TB of RAM. For less demanding scenarios, it may also be equipped with Silver- and Bronze-class processors.
  • Finally, the FUJITSU Server PRIMERGY CX2570 M4 is a half-width, 2U server node that relies on the same processor and RAM technology as its 1U cousins. Unlike both other server nodes, it also may be fitted with up to four NVIDIA® Tesla® P100 or V100 graphics cards. Thanks to this setup, the CX2570 M4 is a perfect match for graphics-heavy and/or compute-intensive workloads such as VDI, CAD, PLM and machine learning/deep learning.

Conclusion
The FUJITSU Server PRIMERGY CX400 M4 Eco-system, consisting of powerful hardware and advanced management capabilities available with Infrastructure Management software, was designed to help companies roll out their most ambitious virtualization projects, serve as a foundation for hyper-converged infrastructures within our PRIMEFLEX product family of integrated systems and can subsequently be used as a building block for a potential transition toward a software-defined data center.

Timo Lampe

 

About the Author:

Timo Lampe

Senior Specialist Marketing Manager – Data Center Systems / Server

SHARE

Comments on this article

No comments yet.

Please Login to leave a comment.