SPEC releases major new version of its storage solution benchmark

New features include AI and genomics workloads, expanded customization, massively better scaling, and statistical visualization

GAINESVILLE, Va., December 15, 2020 — The Standard Performance Evaluation Corp. (SPEC) has released a major new version of its storage solution benchmark.

The SPECstorage Solution 2020 benchmark measures performance of an entire storage configuration as it interacts with application-based workloads. The latest version includes new workloads for artificial intelligence (AI) and genomics, expanded custom workload capabilities, massively better scaling, and a statistical visualization mechanism for displaying benchmark results.

"The new SPEC storage benchmark is a valuable tool to help IT managers make better purchasing, configuration and optimization decisions based on actual applications and real-world scenarios," says Don Capps, chair of the SPEC Storage subcommittee.

New workloads and customization

The new AI workload in the SPECstorage Solution 2020 benchmark represents AI Tensorflow image processing environments, with traces collected from systems running COCO, Resnet50 and CityScape datasets. The new genomics workload comes from commercial and research facilities that perform genetic analysis. Other workloads in the benchmark address video data acquisition (VDA), electronic design automation (EDA), and software build environments.

New capabilities in the benchmark enable users to customize existing workloads or create new ones and test them to gain insight into storage performance issues.

Massive scaling with fewer constraints

New internal IPC mechanisms in SPECstorage Solution 2020 significantly reduce the number of TCP ports and DNS lookups, improving scalability, reliability, and start-up and run-time performance. Scaling has been increased from 60,000 load-generating processes in previous versions to more than 4 million processes that can be distributed throughout the world. The new efficiency is especially significant for testing cloud-based storage solutions.

"Among cloud providers, TCP ports are a scarce resource, with most virtual machines able to accommodate only a few thousand at best," says Capps. "The new benchmark never requires more than 1,000 TCP ports. This enables users to scale up to 4 million processes without running into cloud constraints."

Visualization for greater insights

The new statistical visualization mechanism within SPECstorage Solution 2020 allows users to extract runtime counters and connect them to a database for visualization. This provides greater insight into the behavior and operation details of the system under test.

Available now for downloading

The SPECstorage Solution 2020 benchmark is available for immediate download on the SPEC website for $2,000, with discounts for qualified research and academic institutions. Results from the new benchmark are not comparable to those from past SPEC storage benchmarks.

SPEC corporate members active in the development of the benchmark include Dell, IBM, Intel, iX Systems, NetApp, Pure Storage, and WekaIO. Supporting individual contributors include Udayan Bapat, Sorin Faibish, and Brian Pawlowski.

About SPEC

SPEC is a non-profit organization that establishes, maintains and endorses standardized benchmarks and tools to evaluate performance and energy consumption for the newest generation of computing systems. Its membership comprises more than 120 leading computer hardware and software vendors, educational institutions, research organizations, and government agencies worldwide.

###

Media contacts:: Bob Cramblitt, Cramblitt & Company, 919-481-4599, info@cramco.com
Brigit Valencia, Compel PR, 360-597-4516, brigit@compel-pr.com
SPECstorage™ and SPEC® are trademarks of the Standard Performance Evaluation Corporation. All other product and company names herein may be trademarks of their registered owners.

Graph of SPECstorage 2020 SWBUILD Op_rates

The new statistical visualization mechanism within SPECstorage Solution 2020 allows users to extract runtime counters and connect them to a database for visualization.