Hi,
Looking through the latest Linux magazines, I came across the
interesting point that some Unix's use a "wake all" approach when a
request comes in for Apache, while others use a "wake one only" model.
Let me explain.
Say you have a busy Apache web server, and you start up a process
that uses a lot of RAM. Half of the Apache processes are swapped out
in whole or in part. Then a request comes in.
Under some cases, ALL processes have to wake up, with one process
grabbing the request, and the others not getting it. Then the other
processes go back to sleeping. In this scenario, there would be a
lot of swapping as each swapped out Apache process has to come back
into RAM.
However, some Unix's can wake only ONE process up, and not wake the others.
Which model does OpenBSD use? If it uses the "wake all" model, is
there a way to configure the "wake only one" by recompiling the
server?
Cordially
Patrick Giagnocavo
|