Re: Growing RAM usage by SQUID

We found that vanilla Squid was more likely to fall into swap (which would
result in stalled requests for clients). With NOVM Squid and good SCSI2
hardware (the bulk of the reads/writes being between 2-4k according to the
aic7xxx driver stats), Squid fits neatly into RAM and requests are handled
with good speed during peak time. A weekly shutdown of squid at 4am also
cleans up the swaplog reducing memory.

tom@interact.net.au

> Running on an unmodified 1.1.17. Not the NOVM version. The NOVM kept the memory
> usage much lower, but everyone complained about the response times. (sometimes a
> minute or more per request, before data started to happen).
>
> D
>
>
> Stephen R. van den Berg wrote:
>
> > Dancer wrote:
> > >I'm running an 8GB cache on 64MB of RAM (+64MB swap). Periodic shutdowns (every
> > >couple of days) are necessary. It isn't a leak. It's just bloat from in-transit
> > >allocations.
> >
> > 8GB with 64MB of RAM? What version of Squid are you running?
> > I think the minimum amount of RAM needed to run an unmodified 1.NOVM.18
> > with 8GB of swap is 112MB.
> > --
> > Sincerely, srb@cuci.nl
> > Stephen R. van den Berg (AKA BuGless).
> >
> > This signature third word omitted, yet is comprehensible.
>
>
>
> --
> Between jobs and looking for work in Australia or overseas.
> Got something? Let me know. http://www.brisnet.org.au/~dancer/resume.html
>
>
Received on Sat Dec 06 1997 - 04:55:19 MST