optimal opcode setup for large shared hosting plesk server

[strebel]

This is no good on a large server with hundreds of vhosts.

/dev/shm/ea fills up in about a minute. Interrupts, Fork Rate, Load, and Disk I/O all go up and the node becomes unstable. 8cores and 11gb of ram on this machine.

Any other ideas on how to run a opcode with suExec affectively?

[NiteWave]

then set cache_dir to /tmp/ea
you can clean up /tmp/ea daily

[strebel]

Disk iowait is the problem.

Wait time shoots up as it is trying to read/write from the SAN.. not fast enough.

[NiteWave]

yeah, network disk not recommended. We experienced very slow disk access when /tmp is NFS in a few customer's servers.
using local disk, maybe a dedicated partition to store /tmp or for cache_dir only.