Optimizing Windows for Scale HC3
-
From personal experience - Windows has a lot of default settings that can be optimized to make them run better in a VM, stay smaller over time, reduce I/O load, reduce snapshot and replication size, etc.
One easy example is reducing the size of the windows OS page file... especially if you are taking frequent snapshots or replicating (there is also a flag available that support can set that will exclude a virtual disk from being snapshotted / replicated at all). Adding a little bit more ram to the VM is much better than having the VM swap to a pagefile.
But there are many other optimizations such as turning off 8.3 file name creation on NTFS, disabling background defragmentation tasks, etc.
One thing customers sometimes want to try is pre-allocating (full formatting) NTFS volumes ... we don't see any benefit to doing that and strongly recommend against that for many reasons (creating extra work for HC3 deduplication is one ... it's just going to find all those blocks with the "same" nothing written to them later anyway)
Another don't - don't defragment your NTFS file systems (and be very careful and purposeful about application level defrag / reindexing / compacting, as well) ... a lot of things that made sense with spinning disks and local, non mirrored RAID systems don't help at all with a system like HC3 designed to aggregate the I/O of storage devices spanning multiple nodes.
Most of the tricks that are suggested for other virtualization platforms provide the same benefits when applied to VM's running on HC3. I even have used VMware's free OS Optimization tool on HC3 VM's and seen it reduce snapshot and replication size substantially. But it's just a list of windows settings changes / registry changes that you could apply other ways (and obviously isn't VMware dependent).
https://labs.vmware.com/flings/vmware-os-optimization-tool#summary
Some of the optimizations likely only matter for virtual desktop VM's but others are more general.
As always YMMV, backup, snapshot and test for yourself!
