is it possible to graph disk-related latency?

[sneakyimp]

I have a site I've been writing in php and it's really slow. I'm hoping to locate the bottlenecks and was wondering if cacti can collect data on slowness related to hard disk access.

my cpu load is never very high during some intensive queries and my suspicion is that the disk is the weak link.

[gandalf]

If you have the script/SNMP OID, I have the howto's. See very first link of my signature for more help
Reinhard

[Cova]

Some information about what OS you're trying to get that information from would be helpful.

If it's windows, the important disk counter to watch is Avg. Disk Queue Length - it should typically be less than 2 per physical spindle (so, if it's just a drive, 2 or less is good - for a 4-disk RAID array, 8 or less).

Actual disk latency is watched by the Avg. Disk sec/Transfer counter. I don't really watch that one though, and don't know what the recommended values would be.

I can't help with other OS's - and Gandalf's howto's should cover getting the counter into cacti.