We recently replaced our estate of trusty PowerEdge R620 Dell servers, with shiny new Cisco UCS C220 M5SX Servers.
On each server, VMware ESXi was installed. We used LAG with 4 10 GB NIC’s, each server had 10 7200rpm drives setup in a RAID 6 (no SSDs, don’t ask why).
All looked great, but when moving the VM’s to the Cisco servers with vMotion, the process was super slow.
After a lot of digging, we discovered the issue was with the way the Wite Policy was setup on the RAID, we’d used Write Through, and we should have used Write Back Bad BBU:
– Write Through— Data is written through the cache and to the physical drives. Performance is improved, because subsequent reads of that data can be satisfied from the cache.
– Write Back— Data is stored in the cache, and is only written to the physical drives when space in the cache is needed. Virtual drives requesting this policy fall back to Write Through caching when the BBU cannot guarantee the safety of the cache in the event of a power failure.
– Write Back Bad BBU—With this policy, write caching remains Write Back even if the battery backup unit is defective or discharged.
Once we changed it, everything worked superfast.
