0 Replies Latest reply on Oct 29, 2018 6:57 PM by erives

    vSan stretched cluster Latency.  PFTT=1, SFTT=1, Raid5

    erives Novice

      I am running HCIbench on a newly deployed homogeneous stretched cluster made of 14 cisco UCS c240M5 each with 2 disk groups each ( 400GB cache 5x1.9TB capacity all flash) w/ 2 10Gb VICs 2 fault domains (7 nodes on each site). Running on esxi 6.7. The site to site latency is between 1-3ms.

      I am using the Easy run builtin test (56VMs total with 8vmdk-100ws-4k-70rdpct-100random-4threads) for vsan  and these are the results i am getting while testing a VM storage policy with PFTT=1, SFTT=1, Raid5. Deduplication and compression are enabled.

       

      Are these latency numbers expected for this config/test.

       

      Datastore:  vsanDatastore

      =============================

      Run Def:    RD=run1; I/O rate: Uncontrolled MAX; elapsed=3600 warmup=1800; For loops: None

      VMs         = 56

      IOPS        = 241033.10 IO/s

      THROUGHPUT  = 941.55 MB/s

      LATENCY     = 7.4583 ms

      R_LATENCY   = 5.3595 ms

      W_LATENCY   = 12.3659 ms

      95%tile_LAT = 16.0863 ms

      =============================

      Resource Usage:

      CPU USAGE  = 26.76%

      RAM USAGE  = 10.69%

      VSAN PCPU USAGE = 9.773%