I am curious how others are doing this especially in a VMware environment with SAN storage.
For example. Say you are running Exchange 2010 on Windows Server 2008 R2. The OS is installed on a datastore where the LUN is using 4k blocks. I believe recommendations are still 32k for its Exchange mailbox database and 16k for logs. So would you create a 16k and 32k LUN, mount them as datastores and create separate VMDKs for each, or create the LUNs and map via iSCSI in the guest OS?
Also, at what point do these block sizes really start to have a performance impact? Meaning if you are running Exchange for only 20 users, do you even bother with the different block sizes or just run everything off a 4k LUN?
And what do you do if you are using NFS for mounting datastores?
For example. Say you are running Exchange 2010 on Windows Server 2008 R2. The OS is installed on a datastore where the LUN is using 4k blocks. I believe recommendations are still 32k for its Exchange mailbox database and 16k for logs. So would you create a 16k and 32k LUN, mount them as datastores and create separate VMDKs for each, or create the LUNs and map via iSCSI in the guest OS?
Also, at what point do these block sizes really start to have a performance impact? Meaning if you are running Exchange for only 20 users, do you even bother with the different block sizes or just run everything off a 4k LUN?
And what do you do if you are using NFS for mounting datastores?