Originally Posted by erickotz
Disagree - stick with 4K clusters. While your logic is sound, cluster sizes other than 4K aren't well-tested. While Windows itself will handle it just fine (for the most part, I think even a few things in Windows didn't like it) any sort of 3rd-party utility that deals with files (AV, recovery, backup, cloning, encryption, etc) likely did not test with anything other than 4K clusters.
Up till about 18 months ago, I worked at MS, and among other things, supported storage for enterprise customers. We would generally tell them that non-4K clusters were a bad idea unless they had very specific reasons (I think SQL DB server was one of the few reasons), and even then, with caveats.
I'd really like to hear some specific examples on how using larger than 4k cluster sizes have caused issues with 3rd party software. Nothing I've used over the last 15 years I've been in IT has had an issue with it.
Currently I manage the storage, ESX, and backup environment for my company. We have EMC, Nimble, and custom built storage spread across our datacenters.
Let's use backups as an example. We do disk to disk backups for all of our data, as stated, our backup volumes are for the most part too large to use 4k sizes. No issues on any of them due to the cluster size being set at 64k, even on our replication servers that hold several million files with sizes ranging from a few k to close to a TB in size. Thankfully I can use larger than 4k sizes without worry.
We use three separate backup software for our backup and replication tasks. ALL of them recommend using the largest cluster size available for the volume in order to increase performance and reduce fragmentation. Testing has borne out their recommendations. Backup Exec for example really likes a 64k cluster size.
Use the cluster size that fits the application or data, not some mythical "anything over 4k is bad".