Organizations of all sizes can put the storage technology behind the world’s fastest supercomputers to work for them. See how the HPE storage portfolio does it... and you can, too.
The storage technology supporting the world’s largest supercomputers isn’t exclusive to them. It can help organizations of all sizes cope with the impact of relentless data explosion driven by higher fidelity simulations, larger AI models and larger data sets in high performance data analytics.
All of the world’s most powerful supercomputers are using parallel file systems—either the community-owned Lustre file system or IBM Spectrum Scale (formerly known as GPFS).
The convergence of classic modeling & simulation (MOD/SIM), artificial intelligence (AI/ML) and high-performance data analytics (HPDA) has driven massive improvements on the software layer for both parallel file systems.
- File size: Small, large and mixed
- Access type: Read, write and mixed
- Access pattern: Sequential and random access
When leadership computing sites look at storage systems for supercomputers, they typically select storage systems that provide three core values:
- Performance
- Cost effectiveness
- Operational excellence
We are humbled and proud that four out of the 10 largest supercomputers on the planet have selected parallel storage systems from HPE for the reasons mentioned above.
We are providing a full HPC storage portfolio as shown in the graphic below to help organizations of all sizes and file system preferences to address the top three challenges with legacy HPC storage systems:
As the #1 HPC computer vendor we know about the top HPC storage challenges of our customers who see their legacy HPC storage systems break either architecturally, economically and/or operationally in the era of converged MOD/SIM, AI/ML, and HPDA workflows that drive unprecedented storage requirements—both in terms of storage performance and storage capacity.
1. I/O bottlenecks leading to job pipeline congestion
Job pipeline congestion often causes escalations to due missed deadlines and regrettable attrition of top talent like data scientists as they cannot work productively.
Unlike others, we engineer our systems to deliver more I/O performance than you will ever need.
2. Capacity-based storage licensing leading to exploding storage costs
Customers who used to have hundreds of terabytes of storage capacity now have single digit petabytes and are planning for double digit petabytes. HPC users who used to have single digits of petabytes storage capacity, now have double digit petabytes and plan for having triple digit petabytes in the future. When storage capacity requirements grow by an order of magnitude in just a few years, then capacity-based storage licensing quickly can become cost-prohibitive.
Unlike others, we do not charge licensing per terabyte or per storage drive in our storage systems.
3. Vendor “finger pointing” leading to unnecessary delays in problem resolution
In case of business- or mission-critical system issues that are spanning both the compute and the storage layer it can be very frustrating for the customer to have different vendors accuse each other of being the root cause of the issue—leading to unnecessary delays in problem resolution.
Unlike others, we offer operational support for the full “system stack” with HPE Pointnext Tech Care.
Leadership class storage for all
If you are struggling with one or more of the typical HPC storage challenges mentioned above for your MOD/SIM, AI/ML and HPDA workloads running on clusters of HPE ProLiant DL servers, dense HPE Apollo systems, “fat memory” HPE Superdome Flex systems or on HPE Cray EX supercomputers than we can help you!
And you do not need to be a Top 10, Top 100 or Top 500 HPC user. We can help you even if you are Top 5000 HPC user as our parallel storage systems start as low as 1/10th of an enterprise rack for a purchase price of less than $ 50,000!
This is what Aleph Alpha, a German AI startup, has done when they adopted the HPE Machine Learning Development System to train their multimodal AI, which includes Natural Language Processing (NLP) and computer vision. They started with just four rack units of HPE Parallel File System Storage for their “alpha ONE” AI supercomputer that currently ranks on place 72 of the Top500.org list of the largest supercomputers.
We also provide you with unprecedented options on how to consume our offerings:
- Classic purchase of our infrastructure
- Financing of our infrastructure with HPE Financial Services
- Subscription-based consumption of our infrastructure with flexible capacity
- A fully managed private HPC cloud with HPE GreenLake for HPC either in your premises or in a colocation facility
If you are struggling with one or more of the typical HPC storage challenges mentioned above for your MOD/SIM, AI/ML and HPDA workloads reach out to your account manager today to get a dedicated briefing on how we can help you to cope with the impact of the relentless data explosion.
To dig deeper right away:
Download the paper: Spend less on HPC/AI storage and more on CPU/GPU compute
Find more details online:
Uli Plechschmidt
Hewlett Packard Enterprise
twitter.com/hpe_hpc
linkedin.com/showcase/hpe-ai/
hpe.com/us/en/solutions/hpc
UliPlechschmidt
Uli leads the product marketing function for high performance computing (HPC) storage. He joined HPE in January 2020 as part of the Cray acquisition. Prior to Cray, Uli held leadership roles in marketing, sales enablement, and sales at Seagate, Brocade Communications, and IBM.
