HPC Leadership Computing Trusts DDN

By Rob Farber

June 13, 2016

As we approach an Exascale future, the focus is on how to provision and use that computational capability. In order to realize the full societal impact of Exascale computing, storage systems to support Exascale supercomputers are equally important else those valuable (and expensive) compute cycles will be wasted during IO operations. Thought leadership in the HPC community agrees that increasing core count is clearly the direction for computation (although there are strong differences of opinion on how those increased core counts are to be implemented). However, the storage picture is more complicated.

Unlike computation or in memory systems, data retained in storage is persistent over a period of years and even decades. Further, storage systems cannot ever risk losing or delivering bad data at any point in the data lifecycle.  This requires deep technical capabilities and experience in large scale computing and data management.  As we look to the future, past performance truly is a predictor of future success in the storage market, which is why more than 2/3 of the world’s fastest computers rely on DDN (Data Direct Networks) for their storage needs. From pre-Petascale supercomputers to the current generation of double-digit PF/s (petaflop per second) machines, DDN has preferentially been selected to partner with end users and technology integrators to expand the limits of HPC computing. Looking to an Exascale world, DDN is investing 10’s of millions of dollars and opening new research and development facilities to create the end-to-end storage technologies that will meet the data requirements of current users and future Exascale supercomputers. DDN storage simply works, is fast, expandable, power efficient, and cost effective, which is why DDN is the storage vendor of choice for HPC professionals and those tasked with advancing the state-of-the-art in leadership class supercomputing.

The recent announcement of the Japanese Oakforest-PACS 25 PF/s supercomputer is the latest double-digit Petascale machine that will utilize a combination of DDN burst buffer, application acceleration, SSD and file system technologies together to achieve results faster than conceived possible even just 2 years ago.  The Oakforest storage system is comprised of 25 DDN IME14KX caching appliances to provide 1.4 TB/s of low-latency flash-based cache. These cache devices will work in conjunction with DDN supplied storage to deliver 400 GB/s of peak Lustre bandwidth to meet the storage bandwidth needs of this latest generation multi-PF/s supercomputer. As can be seen in the figure below, Lustre is just one option as tiered DDN storage works with any parallel file-system.

Figure 1: DDN devices work with any parallel file-system
Figure 1: DDN devices work with any parallel file-system

Infinite Memory Engine

DDN’s IME (Infinite Memory Engine) represents a new IO tier for HPC that treats small IO in precisely the same manner as large sequential IO. This is a revolutionary change from existing parallel filesystems results in near wire-speed performance regardless of random IO patterns, IO size, and shared file access. DDN’s IME product line also has the ability to work with future storage media such as 3D XPoint and others.

Figure 2: Rack performance IME (Image courtesy Cray Users Group)
Figure 2: Rack performance IME (Image courtesy Cray Users Group)

IME “burst-buffers”

The DDN IME intelligently decouples storage performance from the traditional view of ‘storage’ to greatly accelerate HPC workloads – especially for frequently performed checkpoint/restart operations.

As can be seen in the figure below, Burst Bandwidth has traditionally required overprovisioning of storage to meet peak bandwidth needs. Checkpoint/restart operations are an example of a common IO operation that requires storage overprovisioning to quickly move the data and prevent wasting valuable compute cycles. The DDN IME caches can be configured to act as burst buffers that can quickly handle bursts of extremely high IO activity. This is the reason why the Oakforest-PACS supercomputer has been provisioned with 1.4 TB/s of DDN IME bandwidth.

Figure 3: Bursty IO patterns require overprovising
Figure 3: Bursty IO patterns require overprovising

IME positions HPC for the Exascale

Looking ahead to the Exascale, DDN IME caches can save significant capital and operational dollars by reducing the number of devices required to achieve Exascale-capable levels of storage performance. To put this in perspective, Gary Grider famously pointed out in his 2009 presentation, Preparing Applications for Next Generation IO/Storage that plotting Exascale storage costs of millions of dollars in log scale means you have hit the big time!

Figure 4: 2009 projected costs of storage for an Exascale system (image courtesy HPC User Forum)
Figure 4: 2009 projected costs of storage for an Exascale system (image courtesy HPC User Forum)

In contrast, the Oakforest-PACS procurement only required 25 DDN IME14KX caching appliances. As the industry leader, DDN has dramatically redefined the storage landscape and costs associated with Exascale storage systems since 2009 as shown in the graphic below.

Figure 5: DDN has redefined the storage landscape since 2009
Figure 5: DDN has redefined the storage landscape since 2009

For HPC, DDN IME devices makes high-performance clusters, multi-PF/s systems, and Exascale computation both possible and affordable.

Figure 6: A DDN IME14k (click to see more)
Figure 6: A DDN IME14k (click to see more)

The many uses of IME

Of course, IME storage works great for databases, out-of-core solvers, and a variety of other scientific and commercial HPC workloads.

Figure 7: Additional uses of a DDN IME product
Figure 7: Additional uses of a DDN IME product
  1. A Write Accelerating Burst Buffer absorbing the bulk application data into the IME14K NVMe solid state cache significantly faster than the file system can absorb it.
  2. A File System Accelerator and Application Optimizer as IME reorders application I/O to optimize flushing the cache to long term storage (enabling purchasing as little expensive cache possible).
Figure 8: Dataflow in the client
Figure 8: Dataflow in the client
  1. A Read-optimized Application-I/O Accelerator that enables out-of-band API configuration of the IME appliance to optimize both reads and writes, allowing more simultaneous job runs, shortening the job queue and enabling significantly faster application run time to the user. The API integrates IME with the job schedulers and pre-stages / warms the cache for new jobs, accelerating first read.

Standard script operations make utilization of DDN IME appliance capabilities straight-forward. The following shows how to use the DDN IME as an application accelerator.

Figure 9: IME acts as an application IO accelerator
Figure 9: IME acts as an application IO accelerator

Robustness and Scalability are key!

Cost and power savings are for naught if the storage solution is not robust and scalable as well.

DDN gives the customer the option of using a technique called erasure coding to protect against storage failures. Erasure codes are primarily used in scale-out object storage systems where erasure encoded data blocks are distributed across multiple storage nodes to provide protection against both media and node failures. Erasure encoding can literally save racks of storage nodes when compared to the alternative, three- or four-way mirroring/replication [For more information click here].

Option 1: Data protection is optional. The IME server and associated storage media are considered “just cache” where the data can be recreated if lost.

Option 2: Erasure coding is calculated at the client:

  • Exhibits excellent scaling and can run with high client counts.
  • Servers don’t get clogged up.
  • There is a tradeoff as erasure coding does reduce usable client bandwidth and IME capacity according to IME count by roughly 11% (in an 8+1 configuration) to 25% (in a 3+1 configuration).
Figure 10: Erasure encoding distributed across multiple IMEs
Figure 10: Erasure encoding distributed across multiple IMEs

Managing the full spectrum of end-to-end data lifecycle management

Robust, scalable, and performant storage are but part of the HPC storage picture as data archive must also be considered as well as full life cycle data management and distributed cloud based storage. Similarly, questions are being raised about the efficacy of POSIX based file-systems in future HPC systems. For this reason, object storage systems are undergoing rapid development.

To address current and future end-user storage needs – even at the Exascale – DDN has created a complete portfolio of end-to-end storage products that work together as an extremely flexible data lifecycle management toolset. DDN claims these tools that can be applied anywhere and at any scale.

Figure 11: DDN end-to-end big data lifecycle management
Figure 11: DDN end-to-end big data lifecycle management

Briefly, the DDN storage portfolio covers:

  • Fast data and compute: Addressed through the DDN family of IME products.
  • File-system appliances: DDN products include the GRIDScaler® and EXAScaler®.
  • Persistent data: Persistent data for a variety of commercial and big data workloads are addressed via the SFA14k™ storage array products.
  • Object and cloud storage: The WOS® Object storage for private and hybrid clouds take DDN customers beyond traditional file-systems. WOS is described in the DDN white paper, WOS® 360° full spectrum object storage.
Figure 12: WOS object storage
Figure 12: WOS object storage

For more information

For more information, visit http://www.ddn.com.


Rob Farber is a global technology consultant and author with an extensive background in HPC and storage technologies that he applies at national labs and commercial organizations. He can be reached at [email protected].

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industry updates delivered to you every week!

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion XL — were added to the benchmark suite as MLPerf continues Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing power it brings to artificial intelligence.  Nvidia's DGX Read more…

Call for Participation in Workshop on Potential NSF CISE Quantum Initiative

March 26, 2024

Editor’s Note: Next month there will be a workshop to discuss what a quantum initiative led by NSF’s Computer, Information Science and Engineering (CISE) directorate could entail. The details are posted below in a Ca Read more…

Waseda U. Researchers Reports New Quantum Algorithm for Speeding Optimization

March 25, 2024

Optimization problems cover a wide range of applications and are often cited as good candidates for quantum computing. However, the execution time for constrained combinatorial optimization applications on quantum device Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at the network layer threatens to make bigger and brawnier pro Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HBM3E memory as well as the the ability to train 1 trillion pa Read more…

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing po Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HB Read more…

Nvidia Looks to Accelerate GenAI Adoption with NIM

March 19, 2024

Today at the GPU Technology Conference, Nvidia launched a new offering aimed at helping customers quickly deploy their generative AI applications in a secure, s Read more…

The Generative AI Future Is Now, Nvidia’s Huang Says

March 19, 2024

We are in the early days of a transformative shift in how business gets done thanks to the advent of generative AI, according to Nvidia CEO and cofounder Jensen Read more…

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, codenamed Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. Wh Read more…

Alibaba Shuts Down its Quantum Computing Effort

November 30, 2023

In case you missed it, China’s e-commerce giant Alibaba has shut down its quantum computing research effort. It’s not entirely clear what drove the change. Read more…

Nvidia H100: Are 550,000 GPUs Enough for This Year?

August 17, 2023

The GPU Squeeze continues to place a premium on Nvidia H100 GPUs. In a recent Financial Times article, Nvidia reports that it expects to ship 550,000 of its lat Read more…

Shutterstock 1285747942

AMD’s Horsepower-packed MI300X GPU Beats Nvidia’s Upcoming H200

December 7, 2023

AMD and Nvidia are locked in an AI performance battle – much like the gaming GPU performance clash the companies have waged for decades. AMD has claimed it Read more…

DoD Takes a Long View of Quantum Computing

December 19, 2023

Given the large sums tied to expensive weapon systems – think $100-million-plus per F-35 fighter – it’s easy to forget the U.S. Department of Defense is a Read more…

Synopsys Eats Ansys: Does HPC Get Indigestion?

February 8, 2024

Recently, it was announced that Synopsys is buying HPC tool developer Ansys. Started in Pittsburgh, Pa., in 1970 as Swanson Analysis Systems, Inc. (SASI) by John Swanson (and eventually renamed), Ansys serves the CAE (Computer Aided Engineering)/multiphysics engineering simulation market. Read more…

Choosing the Right GPU for LLM Inference and Training

December 11, 2023

Accelerating the training and inference processes of deep learning models is crucial for unleashing their true potential and NVIDIA GPUs have emerged as a game- Read more…

Intel’s Server and PC Chip Development Will Blur After 2025

January 15, 2024

Intel's dealing with much more than chip rivals breathing down its neck; it is simultaneously integrating a bevy of new technologies such as chiplets, artificia Read more…

Baidu Exits Quantum, Closely Following Alibaba’s Earlier Move

January 5, 2024

Reuters reported this week that Baidu, China’s giant e-commerce and services provider, is exiting the quantum computing development arena. Reuters reported � Read more…

Leading Solution Providers

Contributors

Comparing NVIDIA A100 and NVIDIA L40S: Which GPU is Ideal for AI and Graphics-Intensive Workloads?

October 30, 2023

With long lead times for the NVIDIA H100 and A100 GPUs, many organizations are looking at the new NVIDIA L40S GPU, which it’s a new GPU optimized for AI and g Read more…

Shutterstock 1179408610

Google Addresses the Mysteries of Its Hypercomputer 

December 28, 2023

When Google launched its Hypercomputer earlier this month (December 2023), the first reaction was, "Say what?" It turns out that the Hypercomputer is Google's t Read more…

AMD MI3000A

How AMD May Get Across the CUDA Moat

October 5, 2023

When discussing GenAI, the term "GPU" almost always enters the conversation and the topic often moves toward performance and access. Interestingly, the word "GPU" is assumed to mean "Nvidia" products. (As an aside, the popular Nvidia hardware used in GenAI are not technically... Read more…

Shutterstock 1606064203

Meta’s Zuckerberg Puts Its AI Future in the Hands of 600,000 GPUs

January 25, 2024

In under two minutes, Meta's CEO, Mark Zuckerberg, laid out the company's AI plans, which included a plan to build an artificial intelligence system with the eq Read more…

Google Introduces ‘Hypercomputer’ to Its AI Infrastructure

December 11, 2023

Google ran out of monikers to describe its new AI system released on December 7. Supercomputer perhaps wasn't an apt description, so it settled on Hypercomputer Read more…

China Is All In on a RISC-V Future

January 8, 2024

The state of RISC-V in China was discussed in a recent report released by the Jamestown Foundation, a Washington, D.C.-based think tank. The report, entitled "E Read more…

Intel Won’t Have a Xeon Max Chip with New Emerald Rapids CPU

December 14, 2023

As expected, Intel officially announced its 5th generation Xeon server chips codenamed Emerald Rapids at an event in New York City, where the focus was really o Read more…

IBM Quantum Summit: Two New QPUs, Upgraded Qiskit, 10-year Roadmap and More

December 4, 2023

IBM kicks off its annual Quantum Summit today and will announce a broad range of advances including its much-anticipated 1121-qubit Condor QPU, a smaller 133-qu Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire