LSWS 4.2 + suEXEC = 50% performance decrease

[Deltabee]

Hi,

We just upgraded to 4.2 and followed the steps to enable suEXEC.

On an AB benchmark we're seeing almost 50% less requests per second than if we set Run on start-up to no (or yes).

Any ideas?

[NiteWave]

with opcode cache enabled ?

in daemon mode, can install xcache or apc or eAccelerator.

if non-daemon mode, only eAccelerator disk-cache is available(opcode cache not in shared memory)

[Deltabee]

Originally Posted by NiteWave

with opcode cache enabled ?

in daemon mode, can install xcache or apc or eAccelerator.

if non-daemon mode, only eAccelerator disk-cache is available(opcode cache not in shared memory)

Yes, eaccelertor is installed. Selecting yes or no, gives around 200 requests a second on a wordpress installation, then selecting suEXEC brings it down to 100 req a second.

However, just running it on a phpinfo() page, gives approximately the same result.

So it could be the wordpress cache isn't using the opcode properly. But does this mean on sites not using opcode, we'll see 50% less performance?

[eva2000]

Originally Posted by Deltabee

Yes, eaccelertor is installed. Selecting yes or no, gives around 200 requests a second on a wordpress installation, then selecting suEXEC brings it down to 100 req a second.

However, just running it on a phpinfo() page, gives approximately the same result.

So it could be the wordpress cache isn't using the opcode properly. But does this mean on sites not using opcode, we'll see 50% less performance?

haven't tried 4.2 yet myself, but what do you have set in litespeed server configuration for PHP suEXEC Max Conn ? default 10 ? tried raising that value ?