In this video from the 2015 OFS Developer’s Workshop, Bob Pearson from Cray presents: NVM Burst Buffers.
The Cray XC series DataWarp applications I/O accelerator technology delivers a balanced and cohesive system architecture from compute to storage. It allocates storage dynamically in either private (dedicated) or shared modes. Storage performance quality of service can be provided to individual applications, based on the user’s policies. While leveraging Cray’s proven domain expertise in storage, the DataWarp accelerator can be used as a global storage cache for parallel file systems (PFS) such as Lustre, General Parallel File System (GPFS) and PanFS. In these scenarios, the applications I/O accelerator capabilities drive up the overall utilization of the parallel file system by buffering performance across this new tier. This enables new levels of pure performance boosts by delivering the right storage performance to applications. Thus, the accelerator improves overall application performance by decoupling application I/O from the corresponding PFS I/O, while also improving the performance and resiliency of conventional disk-based solutions.
Download the Slides (PDF) * See more talks from the Open Fabrics Workshop
