Hi,
I have a report from a VMware Capacity Planner exercise which shows my average Disk I/O (Trans/Sec) as 257 and also the Disk I/O (MB/sec) as 5.
Each of my ESX hosts has a pair of 4Gb HBA's connected to an HP array with disk groups containing 32 x 15k FC disks.
I know that my 4GB HBA's can shift 400MB per minute so that's not a bottle neck but I was wondering how I can measure the Disk I/O trans/sec against my storage configuration to see if they are likely to be a bottle neck.
Does 257 Trans/sec sound low, average or high?
Thanks
247 is pretty low, honestly. Thats about 2 drives worth of IO.
There are really 2 ways to gauge what your array can do:
1) Talk to your vendor, and have them calculate what your array should be able to do in that config.
2) Benchmark it using IOmeter, vdbench or similar and see what it can do. Note - this can put a lot of stress on your array, so be careful about doing it on production arrays.
That being said, 32 15K drives would be expected to handle *about* 3000 (RAID5) IOPs or about 4500 (RAID10) IOPs, in total.
247 is pretty low, honestly. Thats about 2 drives worth of IO.
There are really 2 ways to gauge what your array can do:
1) Talk to your vendor, and have them calculate what your array should be able to do in that config.
2) Benchmark it using IOmeter, vdbench or similar and see what it can do. Note - this can put a lot of stress on your array, so be careful about doing it on production arrays.
That being said, 32 15K drives would be expected to handle *about* 3000 (RAID5) IOPs or about 4500 (RAID10) IOPs, in total.
Thanks for your quick response.
Can you confirm that when people are talking about I/Ops generally are they speaking about transactions per second or something else?
Cheers
Confirm
(Reads + Writes) / sec = Transactions /sec = IOPs
Thanks, this has been very helpful.