
Sycomp Intelligent Data Storage Platform, deployed on Google Cloud Platform (GCP) with A3 Ultra virtual machines, meets the performance criteria mandated by this industry-leading research and development (R&D) company for artificial intelligence (AI) and machine learning (ML) workloads.
The MLPerf Storage v1.0 benchmark was successfully executed, demonstrating a ‘train_au_meet_expectation: success’ result on a single A3 Ultra compute node utilizing a five-node Sycomp Intelligent Data Storage Platform cluster. The system achieved an average mean throughput of approximately 23 GiB/s and demonstrated the capability to scale to keep pace with future demands.
Using standard FIO workloads, Sycomp Intelligent Data Storage Platform delivers high-performance storage capabilities, achieving 157,000 IOPS for 4KiB random reads and 28.5 GiB/s for 64KiB sequential reads.
These tests showed Sycomp Intelligent Data Storage Platform is capable of supporting the most demanding AI workloads on Google Cloud.
Other Resources

Achieving 97% GPU Utilization on Google Cloud with Sycomp Intelligent Data Storage Platform

Preparing for the Retirement of Azure HPC Cache: What You Need to Know

Sycomp Intelligent Data Storage Platform Unlocks Scalable, High-Performance Data Access for Global Financial Market Maker
