Disk Performance on Vmware

[EL_Beginner]

Sorry I am in the wrong subgroup here. Please let me know if I am.

I'm hoping you can point me in the right direction on if I should disable journaling or not on disposable servers?

Background -
I'm on 6th working day of supporting CentOs 5.x for a team of 300+ developers, dev environments. Roughly 8000 VMs. And I am kinda new to this. I'm told money is tight.
Our Stack: NetApp >> vSphere 5.1 >> CentOS setup >> JBOSS

Problem -
We have discovered our IO is pretty high. Tipped our NetApps over ( Kjournald alone is generated millions of IO. We are using Ext3 for no other reason than that is what it install by default. Since these servers are DISPOSABLE and there is no data stored on these boxes it would seem running a journaling file system offers no value, but high cost. NetApp has come in and we're looking at min $600k in labor, hardware, licensing etc to get our IO right.

Proposed Solution -
If I can save us a a few hundred thousand dollars or extend the use of our hardware I would like to. Ideally as environments are reinstalled I wanted to move the to ext4 and disable journaling. Again servers are disposable and we wouldn't use this kick start for any machine purposed for real work. So I am not concerned at all about that. We can reinstall and redeploy code in minutes. So I am not even concerned about the down time.

Test -
I did this in a local virtual machine and did indeed reduce IO substantial once I generated some IO.

Question -
Is there anything I am missing? Some feature other than fast recovery and protection of data during a crash I would loose? Other input?

[EL_Beginner]

Thanks for allowing my post. 80+ views so far and no replies so I might be at the wrong place. Any other recommended forums I might be able to get input on this from? I guess this is less a CentOS question and more a Linux file system optimization issue. Any favorite forums that might help me out there?

Barring anyone telling me not to, I am going to provision a 100 servers and give it a shot tomorrow. I'll post my results.

[aks]

The journal is good for recovery of lost or partial writes. That is what you would loose.

[drk]

You can reduce I/O further by mounting your filesystem(s) with noatime and nodiratime options.

You might also set the elevator to noop and let VMware handle it.

[EL_Beginner]

Just wanted to give a little follow up here. Seeing major improvements in IO so far no issues having to rebuild machines. We did have to FSCK some machines after an ESXi host died. So far so good.

Thanks for everyone's input.

[EL_Beginner]

We have had this running for some time and have noticed staggering reduction is in IO against our virtual infrastructure. So far no issues to speak of. We're still only using this for disposable servers. But it's worked out great.