We are in the process of running jetstress for exchange 2010 using ESX hosts across FCoE links and the JetStress test is causing the FCoE links to saturate. Looking at the storage subsystem, the test is doing 16Kb writes and 256Kb reads. Is there any way to manipulate the I/O size so that we can achieve the number of IOPS without pushing the bandwidth as that much data is saturating the FCoE links? We cannot get the JetStress to complete successfully but we are moving 1.8GB/Sec. We are getting throughput of 4000 IOPS from the disk array, but JetStress is seeing network saturation as disk latency. All documentation is around IOPS, not bandwidth.

Hi You can try to reduce number of database and copy of “define database configuration” Tab. Maybe it is helpful to you to manipulate the I/O size. But I also can’t find clue of network bottleneck.