"We know that ESX can handle a large number of IOPS, but we set out to see just how far we could push it. We went to EMC and got as many storage arrays as we could lay our hands on. We picked a realistic workload of 100% random with a 50/50 read/write mix and 8KB block size and then fired it up. We kept pushing more and more IOPS and managed to get up to 100,000 IOPS before running out of storage array bandwidth. And we still had plenty of headroom to spare with ESX!"

Read the full article here

Posted by Gabriel Maciel