POOLING NVME WITHIN GPFS NSDS ENABLES EFFICIENT BURST BUFFER
POOLING NVME WITHIN GPFS NSDS ENABLES EFFICIENT BURST BUFFER
4 PagesCOPYRIGHT © 2017 | INFO@EXCELERO.COM | WWW.EXCELERO.COM CASE STUDY POOLING NVME WITHIN GPFS NSDS ENABLES EFFICIENT BURST BUFFER USE CASE Large-scale modeling, simulation, analysis and visualization CHALLENGE Complete checkpoints within 15 minutes to meet availability SLA's SOLUTION NVMesh enables a petabyte-scale unified pool of distributed high-performance NVMe flash as burst buffer for checkpointing RESULTS • 80 pooled NVMe devices • 148 GB/s of write burst (device limited) • 230GB/s read throughput (network limited) • Well over 20M random 4k IOPS BENEFITS • Meets the 15 minutes checkpoint window • Extremely cost effective • Unheard of burst buffer bandwidth SciNet is Canada’s largest supercomputer center, providing Canadian researchers with comput
