Here are the resuts of running IOZone on an ext3 filesystem on an LVM volume residing on a SAN LUN (it was ran with 5 parallel processes).

"Throughput report Y-axis is type of test X-axis is number of processes" "Record size = 4 Kbytes " "Output is in Kbytes/sec"

" Initial write " 81628.55

" Rewrite " 83354.72

" Read " 115595.02

" Re-read " 119306.09

" Reverse Read " 47684.20

" Stride read " 10011.09

" Random read " 16751.27

" Mixed workload " 5659.77

" Random write " 1661.85

" Pwrite " 36030.83

Now this is all nice and dandy, but my question is: how do I know whether the values are as good as they could be or there is something to tweak (and if so, what?)

The actual usage I will have for that Logical Volume is to act as virtual disk for a VM.

  • I get better results from this off an amazon virtual machine instance using ceph with 16 clients. I think there is something wrong with your setup. – Matt Aug 27 '13 at 9:19

Ok, well you've not told us how many disks of what type on what SAN but I'm guessing this is a RAID 5 array of 4/5 x 7.2krpm disks via iSCSI over a 1Gbps NIC (?) - if that's the case then that seems fine really - it depends what you're using it for, again you don't mention.

If it's anything else then come back with more details of what you've got and what you want it for and I'll do what I can.

edit - is this an IBM N5000/6000? It feel like it might be.

  • doing what? sat there being a DHCP server or busy like a DB/Exchange server - VM's IO usages varies as much as physical ones do – Chopper3 Feb 4 '11 at 14:41
  • The usage, as I said, it's to act as virtual disk for a VM (it will run an application that uses LDAP and PGsql a lot - so basically database). The SAN itself is a 4Gb (I believe) FC SAN, with 2 paths to the storage, of which only one is used at any time (I think that's called active/standby). The configuration of the storage is 12 x 3.5" 3Gb SAS drives (10K RPM), RAID 5. – persson Feb 4 '11 at 14:41
  • 1
    In that case it feels a little slow yeah, especially on random writes. – Chopper3 Feb 4 '11 at 14:43
  • Alright, I suspected that. So what do I do to (try to) make it better? Thanks. – persson Feb 4 '11 at 14:55
  • it's an IBM DS3400 – persson Feb 4 '11 at 15:21

Test "Stride read", "Random read", "Mixed workload" and "Random write" with very poor results. Other tests at the level of one SATA disk.

OLTP(MySQL,PostgreSQL,...) database will work poorly.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.