Infinite I/O

Cache warming and VM migration

Posted by Alan Brandon on Aug 8, 2013

Cache warming is the process of preloading a cache with the appropriate, most frequently accessed data needed to speed up system performance. For server-side caches aimed at speeding up storage performance in virtualized environments, the caching software can take days to analyze the data and load the cache. This delays when the cache becomes most effective. When a VM is migrated to another host its cache is generally cleared, and the cache must be warmed all over again adding further delay.

In a blog post for Storage Switzerland, senior analyst Colm Keegan explores how to avoid repeated cache warming due to VM migration.

"The single biggest cause of unneeded cache rewarms are VM Migrations. Following a VM migration, cache warming can significantly degrade application QoS. The reason is most software caching technologies require that data be evicted from the cache on the original host before the VM can be migrated to another host. This means that the migrated VM will be downgraded to conventional disk performance speeds until its new cache resource can be re-warmed with all the correct data." 

Read Colm's column, and then check out the on-demand webinar, The Challenges with SSD Caching in VMware Environments, featuring George Crump from Storage Switzerland and our director of product marketing, Peter Smith.

When you view the webinar you will be able to download a copy of Storage Switzerland’s latest report, The Challenges with Server Side Caching in NFS Based VMware Environments.

challenges with server side caching white paper 1

Topics: Talking Tech