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
Jul 29 2019

DCPMM – Fujitsu presents new class of memory

In today’s data-centric world, all data must be stored, processed and analyzed, leading to an enormous and ever-increasing need for greater computing performance within the IT infrastructure. With memory bandwidth becoming the most common bottleneck, a new class of memory has arrived to solve your business performance issues — Data Center Persistent Memory. As always, Fujitsu is here to help your company, and is currently the only provider in the market able to deliver this solution.

Aiming to accelerate the performance of critical enterprise data workloads, at the beginning of April Fujitsu refreshed and turbo-charged its entire x86 server portfolio, fully exploiting the capabilities of the new Intel® Optane™ DC Persistent Memory. Adding breakthrough non-volatile random-access memory (NVRAM) technology to its new-generation PRIMERGY and PRIMEQUEST servers based on the second generation of Intel® Xeon® Scalable family processors, Fujitsu is removing performance bottlenecks that hold back today’s applications.

The technology that dominates traditional main memory, DRAM, is quickly accessible but small, expensive and volatile. Storage is large, cheap, durable, but only slowly accessible. Data Center Persistent Memory is disrupting the traditional storage hierarchy by creating a new layer to close the storage gap and provide more performance, efficiency and affordability.

 Image

Intel® Optane™ DC Persistent Memory supports the 2nd generation Intel® Xeon® Scalable family CPUs, and is available in capacities of up to 512GB/module. Persistent memory modules are DDR4 socket compatible and co-exist with conventional DDR4 DRAM DIMMS on the same platform.

The DCPMM modules are able to operate in three different modes:

  • Memory mode – the persistent memory acts as volatile system memory and DRAM is used only for caching. This mode works transparently with any application;
  • App Direct mode – the persistent memory is accessed directly by specifically coded applications. Total system memory is the combination of DCPMM and DRAM capacity;
  • Mixed mode – some percentage of DCPMM capacity is directly accessible to specific applications (App direct), while the rest serves as system memory. The App direct part of DCPMM is displayed as persistent memory, while the rest of DCPMM capacity is displayed as system memory. DRAM DIMMs act as cache.

Addressing several customer pain points, DCPMM offers benefits that companies have been looking for for a long time:

 Image

With the emerging need for applications to handle much larger datasets, with higher throughput and lower latencies, DCPMM enables entirely new use cases and applications:

  • In-memory databases: DCPMM makes fast access to very large volumes of data both practical and affordable;
  • Big Data and real-time analytics: Enterprises and service providers need to exploit the information as quickly as possible, using it to make immediate decisions;
  • HPC: High-performance computing demands fast processors and fast access to very large volumes of data;
  • Very large VMs and dense populations of containers: As virtualized platforms expand, so does their requirement for ever-larger memory spaces. DCPMM will make it possible to run more VMs per system. The same need will apply as container usage expands;
  • Very fast storage: DCPMM allows for both fast and high-capacity cache in front of traditional storage, and on its own;
  • Machine Learning and AI: Machine Learning is at its most effective when large sets of data are used for training models.

 

Fujitsu & NEXTGenIO

The new PRIMERGY and PRIMEQUEST servers benefit from Fujitsu’s pioneering work in the NEXTGenIO project, part of EU Horizon 2020¹, the European Union’s biggest ever research and innovation program. Using a requirements-driven co-design process involving all the stakeholders² of the NEXTGenIO project, Fujitsu and Intel developed a prototype system using DCPMM memory. As the key output of the project, a 34-node cluster was delivered in June 2019 by Fujitsu and installed at the University of Edinburgh’s Advanced Computing Facility, where it is now in the final testing and integration phase before going into full production in the summer.

The success of this EU project made it possible for Fujitsu to introduce these unprecedented I/O capabilities to industry standard x86 PRIMERGY and PRIMEQUEST servers.

In order to learn more about Fujitsu, DCPMM, and to review newly-published results of independent benchmark measurements please visit this link.



¹ The NEXTGenIO project has received funding from the European Union’s Horizon 2020 Research and Innovation program under Grant Agreement no. 671951.c

² The consortium partners are EPCC, Intel, Fujitsu, Technische Universität Dresden, Barcelona Supercomputing Center, the European Centre for Medium-Range Weather Forecasts, Arm (formerly Allinea) and Arctur.

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.