If you consider the heavy use of Lustre in the world of HPC, it is clearly the best choice for extension into a cloud-based HPC service. Lustre is the main file system for 3 of the 5 Top supercomputers in the world, and over half of all supercomputers use Lustre. To appreciate just how powerful and fast Lustre is, consider that Oak Ridge National Laboratory’s supercomputer runs Lustre and has more than 25,000 clients, over 10PB of storage, and aggregate IO throughput of 250 GB/sec.
With demand for the combination of AI, machine learning, and big data analytics just starting its surge, offering Lustre-based HPC means winning and retaining customers in a potentially vast and broad market that needs cost-effective, high-speed, random read access to large amounts of data. Further evidence of the market betting on Lustre can be seen from Amazon’s recent announcement of Amazon FSx for Lustre which clearly underscores the company’s view that Lustre is the best HPC path, rather than NFS-based file services.
How Kmesh SaaS Accelerates Your Planned HPC Offering
Kmesh is software (SaaS) that transforms your centralized data (file system data, unstructured data, NoSQL data) into distributed data, operating over multiple clouds, countries, and edges as a single global namespace. In addition, Kmesh invested in Lustre as our own core technology, which is why we work closely with the Open SFS community as a Participant that can shape features of Lustre and influence other players in the community (see Kmesh next to Intel and below Cray).
Participants | OpenSFS: The Lustre File System Community
Kmesh SaaS incorporates all the functionality needed to real-time manage HPC app data across cloud, hybrid cloud, and multi-cloud deployments. The Kmesh SaaS solution distributes file system data over as many of your cloud (and edge) locations as you choose and provides a single global namespace for each customer’s data. Customers configure their business rules / data orchestration policies through the Kmesh portal and easily deploy their data over any cloud location(s).
The main three elements of the Kmesh Saas solution for CSPs to develop HPC offerings include the following:
Lustre-as-a-Service: CSPs can use Kmesh SaaS to exploit Lustre’s performance and scalability to support latency and throughput of HPC apps. Additionally, there is nothing to manage to make the solution work. It is completely turnkey.
Multi-Cloud Data Orchestration: CSPs can sync data from on-premises HPC data into less expensive storage and cloud-based HPC clusters. HPC results data can be efficiently synched with data for other apps, both on-premises and cloud.
Cloud Caching for Remote Data Access: Sensitive on-premise data can be accessed from public cloud without experiencing security issues, while HPC apps can be brought up faster on cloud – without waiting for data transfer.
Since Kmesh SaaS operates within any existing cloud infrastructure and security architecture, you do not have to make changes to any of your data storage infrastructure to leverage Kmesh SaaS quickly.
When it comes to HPC offerings in the cloud, Kmesh gives you what you need to get started today:
Move to HPC in the Cloud Fast, Remain Flexible
Because Kmesh SaaS does not require installation and management (due to auto-instances), you can trial and run the software fast. Given its features of multi-tenancy, configurability, and policy-driven operation, you can run new HPC offerings in ways that flex to your business model and architecture. With auto-scaling, simple APIs and an intuitive GUI, your HCP services will flexibly grow as you desire.
These are exciting times for CSPs as enterprises large and small turn to the cloud for HPC use cases. So, contact Kmesh today to learn more about streamlining your path to HCP offerings or email firstname.lastname@example.org.