Hi,
Hardware:
FAS2554@cdot 8.3.2 with 24*4TB-SATA-Disks, 2*10G Network connectivity
This FAS exports a NFS-Volume to some ESXi-Hosts. This FAS/volume is dedicated to receiving large sequential writes, f.e. I use GhettoVCB to backup some VMs towards this FAS. The VMs reside on SSDs directly on those ESXi Hosts, which are also connected with 10G. There is no other load on this FAS and MTU size is standard.
During such an backup, write speed on this FAS is about 160-200 MByte/s. Disk utilization (sysstat -x) is about 30%. I wondered about a possible bottleneck and did "node run -node (nodename) sysstat -M 1":
ANY1+ ANY2+ ANY3+ ANY4+ AVG CPU0 CPU1 CPU2 CPU3 Nwk_Excl Nwk_Lg Nwk_Exmpt Protocol Storage Raid Raid_Ex Target Kahuna WAFL_Ex(Kahu) WAFL_XClean SM_Exempt Cifs Exempt SSAN_Ex Intr Host Ops/s CP
100% 76% 42% 18% 59% 46% 59% 47% 86% 1% 0% 35% 0% 2% 7% 0% 0% 1% 159%( 98%) 18% 0% 0% 8% 0% 2% 4% 3229 41%
100% 76% 41% 14% 58% 46% 56% 41% 89% 1% 0% 36% 0% 4% 10% 0% 0% 1% 154%( 98%) 8% 0% 0% 12% 0% 2% 2% 3190 80%
100% 79% 49% 26% 64% 51% 61% 53% 90% 1% 0% 34% 0% 5% 14% 0% 0% 1% 151%( 99%) 27% 0% 0% 11% 0% 2% 8% 3002 66%
100% 69% 33% 9% 53% 40% 51% 34% 87% 1% 0% 34% 0% 2% 4% 0% 0% 1% 156%( 98%) 0% 0% 0% 7% 0% 2% 3% 3173 69%
100% 87% 61% 31% 70% 63% 69% 60% 87% 1% 0% 35% 0% 6% 17% 0% 0% 1% 162%( 98%) 27% 0% 0% 19% 0% 2% 9% 2918 100%
Is the FAS CPU a bottleneck here?
About 3000IOPS@200MB/s means 64K per IO. Is that common/normal?
