Excelero showcases Elastic NVMe for AI and HPC workloads at SC19
As the Excelero team is on its way to Denver for SuperComputing 2019 (SC19), I thought it would be a good idea to provide an overview of all the exciting stuff we’ll have going on at the event. Once again, we chose to leverage our strong HPC partner ecosystem. As a result, you will find Excelero all over SC19, in over half a dozen of booths, promoting our HPC and AI use cases. There will be live demos featuring technologies from Mellanox, AMD, SuperMicro, Kioxia and Penguin Computing, and presentations with Mellanox, Micron and ThinkParQ. Cherry on the cake will be the announcement of STFC as our newest AI customer, a joint success with Boston Ltd., Mellanox and NVIDIA. Here is a selection of the exciting stuff we’ll have going on this week:
After successful collaborations at SC17 and SC18, Excelero is once again hosted in the Mellanox boot (#601). As for previous editions, the Excelero team will be offering live demos of NVMesh. To demonstrate the flexibility of our software-based elastic NVMesh technology, we set up multiple configurations:
High Availability with Mellanox BlueField and NVMe-of
This will be a joint technology demonstration highlighting the value of Mellanox BlueField NVMf initiator offload with Mellanox SNAP technology and ultra-high performance, scalable Elastic NVMe provided by Exclero’s NVMesh. The demo consists of both Microsoft Windows and Linux application servers accessing NVMesh logical volumes at very low latency and high bandwidth, with various levels of protection against both drive and server failures. Mellanox SNAP technology will allow for NVMesh logical volumes to appear to the application servers as local, NVMe drives without the need for special block drivers and without consuming precious server CPU. The logical volumes themselves will be provided by Excelero NVMesh over industry standard NVMf protocol, with volumes being highly available and performant, without the need for any special configuration or software on the application servers.
Best price/performance leveraging AMD EPYC and Mellanox ConnectX – in partnership with Penguin Computing
Penguin computing and Excelero are teaming up with AMD and Mellanox at SC19 to give a live demo of the leading price/performance NVMe storage servers on the market. In the demo, Excelero will show how it leverages the record-breaking number of PCIe lanes that come with AMD EPYC™ processors. Excelero’s elastic NVMe software was designed to not consume CPU on the targets, which makes it the only scale-out NVMe storage solution that can fully utilize the EPYC PCIe performance without making the CPU a bottleneck. Additionally, when paired with Mellanox ConnectX® PCIe Gen4-capable network adapters, those PCIe slots double the available data throughout. This makes AMD’s EPYC platform and Mellanox networking solutions combined with Excelero’s NVMesh Elastic NVMe an extremely efficient and scalable system to provide shared NVMe access at local speed. We will demonstrate AMD’s the highly efficient AMD’s EPYC efficient CPU building block for scale-out NVMe.
ThinkParQ booth (#1373), Tuesday 2pm – Supercharging AI and HPC workloads with BeeGFS on Elastic NVMe
The new generation of AI & HPC applications significantly raises the bar for shared file storage. This comes with the demand for a new solution that not only provides very high streaming throughput, but also efficiently handles random access and lots of small files to unlock the full performance potential of modern of GPUs and CPUs. This presentation will show how to easily eliminate the traditional storage bottleneck by combining Excelero’s NVMesh and the BeeGFS parallel file system to take full advantage of NVMe. Excelero’s patented Remote Direct Drive Access (RDDA) technology and the MeshProtect flexible data protection options allow remote NVMe access at the same speed that a local drive would provide, enabling a very fast resilience layer and easy management of NVMe drives. The BeeGFS technology works on top of this fast, resilient block layer to provide a file system view to the storage at unmatched performance. Thanks to the intelligent software stack, the solution runs on commodity servers to provide full flexibility and enables converged systems for optimal cost efficiency, eliminating the need to procure separate servers for NVMe drives and file services.
Micron booth (#335), daily at 2pm – Excelero’s NVMesh – Software defined elastic storage for the NVMe era
The new generation of AI & HPC applications significantly raises the bar for shared file storage. This comes with the demand for a new solution that not only provides very high streaming throughput, but also efficiently handles random access and lots of small files to unlock the full performance potential of modern of GPUs and CPUs. This presentation will show how to easily eliminate the traditional storage bottleneck by combining the latest generation of Micron’s highly efficient NVMe drives with Excelero’s NVMesh and the BeeGFS parallel file system. Excelero’s patented Remote Direct Drive Access (RDDA) technology and the MeshProtect flexible data protection options allow remote NVMe access at the same speed that a local drive would provide, enabling a very fast resilience layer and easy management of NVMe drives. The BeeGFS technology works on top of this fast, resilient block layer to provide a file system view to the storage at unmatched performance. Thanks to the intelligent software stack, the solution runs on commidity servers to provide full flexibility and enables converged systems for optimal cost efficiency, eliminating the need to procure separate servers for NVMe drives and file services.
Mellanox booth (#601), Tuesday at 5pm – Elastic NVMe for AI and HPC
The biggest advantage of modern GPU computing is also creating its biggest challenge: GPUs have an amazing appetite for data. Excelero’s NVMesh eliminates any compromise between performance and practicality, and allows GPU optimized servers to access scalable, high performance NVMe flash storage pools as if they were local flash. This technique ensures efficient use of both the GPUs themselves and the associated NVMe flash. The end result is higher ROI, easier workflow management and faster time to results.
Excelero and Boston Ltd. are proud to announce that the Science and Technology Facilities Council (STFC) has deployed a new high performance computing (HPC) architecture to support computationally intensive analysis including machine learning and AI‐based workloads using the NVMesh™ elastic NVMe block storage solution. The deployment, done in partnership with Boston Limited, a provider of high performance, mission‐critical server and storage solutions, is enabling researchers from STFC and the Alan Turing Institute to complete machine learning training tasks that formerly took three to four days, in just one hour – and other foundational scientific computations that researchers formerly could not perform.
Boston Ltd. (booth #1849) will be promoting the Flash-IO Talyn: Utilizing Excelero’s NVMesh architecture, the Boston Flash-IO Talyn is able to extend the promise of SDS to low-latency workloads by leveraging server-side NVMe-based flash storage to deliver a scalable converged infrastructure for next level performance.
Supermicro (booth #1211) will be promoting the NVMesh BigTwin Reference Architecture, which implements the next-generation of intelligent infrastructure platforms designed to provide applications with all of the raw performance you can expect from local storage — and all of the flexibility, manageability and scalability typically experienced with traditional storage area network-based solutions. Leveraging the latest NVMe and PCIe standards, this reference architecture uses a low-latency, high-bandwidth networking to connect compute and storage together in a flexible way that will fit almost any application’s architecture requirements
Lenovo (booth #763) Lenovo and Excelero have joined forces to deliver a high-performance, low-latency elastic NVMe solutions leveraging Lenovo servers with NVMe flash, high-performance networking and Excelero NVMesh. Marrying high-performance, reliable, standard servers from Lenovo with innovative and revolutionary software from Excelero enables what was previously unattainable: the cost savings of standard servers, the performance of local flash with the convenience and protection of centrally managed storage.
QCT (booth #1671) QCT and Excelero have partnered to develop a High-Performance Storage solution that delivers the high IOPs performance capabilities of NVMe Flash at scale without the traditional storage bottlenecks. Targeted at HPC and AI users, QxStor Excelero NVMesh combines the QCT next-generation NVMe storage servers with Excelero’s NVMesh® software-defined block storage for shared NVMe at local performance.