Is it possible to make my file server go faster?
I have a windows2003 32 bit fileserver running on an IBM DS4300 with a Raid5 VMFS partition with 900 GB of space.
The server has the following drives:
One C:\ drive with 10 GB space (including swap file) Yes i will increase this drive!
one D:\ drive with 100 GB space
one E:\ drive with 40 GB space
one G:\ drive with 332 space
one H:\ drive with 50 GB space
I included the config file for the file server.
What's the biggest recommended size of the drives?
Pro/Cons with different file sizes?
What about the SCSI reservations?
Something else to think about?
I can't change anything on my SAN!
Tuning your hardware's performance is a bit outside of the scope of this forum. But I can give a couple pointers that will help with ESX.
Some might recommend replacing your VMFS partition with an RDM to the LUN. This may squeeze out 5% performance but at severe cost to your system's flexibility. You can read more about this in our VMFS performance paper . As I say, some recommend it but I'd stay away from it.
Create your VMFS partitions with VirtualCenter and they will automatically be aligned. If VC is not available, manually align the partitions. See the partition alignment performance paper.
SCSI locking will not impact VMFS performance during normal usage. The only time that SCSI locks are used is during administrative operations (disk creation, deletion, etc.) With fixed-size VMDKs on your VMFS partition SCSI locks are not used and VMFS scales perfectly.