ClusterStor L300F Lustre-Based Scalable Flash Storage System From Cray

Cray Inc. announced the ClusterStor L300F Lustre-based scalable flash storage solution, designed for applications that need high-performance scratch storage to store and retrieve intermediate results.

The L300F system enable users such as engineers, researchers and scientists to reduce the run time of their applications, allowing jobs to reliably complete within their required schedules, supporting more iterations and faster time to insight.

Supplementing the company’s ClusterStor portfolio with an all-flash storage option, the L300F will integrate with and complement the existing L300 and L300N models to provide a comprehensive storage architecture. It will allow customers to address performance bottlenecks without needlessly over provisioning HDD storage capacity, creating a cost-competitive solution for improved application run time.

“Flash is poised to become an essential technology in every HPC storage solution,” said Jeff Janukowicz, research VP, SSDs and enabling technologies, IDC. “It has the unique role of satisfying the high-performance appetite of artificial intelligence applications even while helping customers optimize their storage budget for big data. With the ClusterStor L300F, Cray has positioned itself to be at the leading edge of the next generation of HPC storage solutions.“

The L300F will simplify storage management for storage administrators, allowing them to stand up a performance flash pool within their existing Lustre file systems using existing tools and skills. This eliminates the need for product-unique training or to administer a separate file system. Using ClusterStor Manager, administrators can reduce the learning curve and accelerate time to proficiency, thereby improving ROI.

When coupled with the firm’s monitoring application, Cray View for ClusterStor, administrators get an end-to-end view of Lustre jobs, network status and storage system performance. View for ClusterStor provides visibility into job run time variability, event correlation and trend analysis, and offers custom alerts based on any selected metric.

The flash L300F system will eliminate the I/O latency of HDDs and allows system architects to dynamically increase the performance SLAs to users for IO/s while lowering the effective cost ($/IOP) of providing the service. It will offer balanced performance, ensuring that IO/s performance is consistently delivered throughout the system. Like all ClusterStor options, the L300F model will contain both object storage servers and object storage targets, and it will scale linearly and with no performance loss. Moreover, users can be assured their data is well protected, as ClusterStor L300F storage will deliver redundancy throughout the storage system and features HA.

“At Cray we are committed to continue making ClusterStor the most capable HPC storage platform in the industry,” stated John Howarth, VP, storage, Cray. “The L300F is another step in addressing critical customer requirements in the demanding HPC storage market.“

The ClusterStor L300F model joins the ClusterStor L300 and ClusterStor L300N storage systems in the company’s portfolio. It’s a flexible storage system that will offer heterogeneous building blocks and support for Lustre 2.11 functionality. Packing 24 SSDs into a 2U enclosure is for customers running large volumes of small I/O workloads, and the L300F solution takes the firm’s storage offerings for HPC environments to a new level.

The existing L300 model is an all-HDD Lustre solution, suited for environments using applications with large, sequential I/O workloads.

The L300N model, by contrast, is a hybrid SSD/HDD solution with flash-accelerated NXD software that redirects I/O to the appropriate storage medium, delivering cost-effective, consistent performance on mixed I/O workloads while shielding the application, file system and users from complexity through transparent flash acceleration.

ClusterStor L300N
The ClusterStor L300N system with NXD flash accelerator software optimizes multiple mixed workloads on the same storage platform, eliminating data access bottlenecks through flash-based small file acceleration. Organizations will be able to automatically support multiple applications that require a range of storage access patterns, such as small block and large block, without making any changes to applications or processes. ClusterStor NXD technology scales performance linearly with the size of the system.

Click to enlarge

View for ClusterStor
View for ClusterStor software tracks job performance on Lustre file systems, and it enables the correlation of system events that cause job performance issues using streaming telemetry with customizable visualization, enabling real-time and historical analysis. It provides administrators with the tools to resolve problems faster.

The company will be at ISC 2018 on June 24-28 in Frankfurt, Germany.