Re: Squid memory footprint (was: MemPools rewrite)

Chemolli Francesco (USI) wrote:
>> I wonder how can you eliminate StoreEntry? IMHO it contains crucial
>> information that allows squid to skip disk accesses. Moving parts
>> of this data into squidfs doesn't seem to change much in ram usage.
>> Moving this crucial information onto disks implies enormous
>> performance
>> penalty, doesn't it?

>I don't know if and how it could be done, but what about moving
>all storeentry structures in a BIG chunk mmapped on the
>swap.state file?

Been there, done that.

> This way the OS would be in charge of doing
>pagein and outs for us.
>Besides, we could save up tons of startup time for swap.state
>parsing and checks: just mark the file clean, compute an MD5
>checksum and store it somewhere This could allow for
>simultaneous access by two processes (synchronous double-checking).
>No heap fragmentation problems either, reallocs are easy,
>disk space is cheap.

Yep, works like a dream. Been running squid like this for more than
two years like this.

>It all seems too perfect. Only disadvantage I can see, we'd have
>to implement memory-management on the thing. But if we KNOW
>that all objects have the same size (which is not entirely true,
>but we could maybe do some storage organization tricks), it should
>be relatively easy.

It is. The patches I've created have been published on squid-dev
in June 1999.

-- 
Sincerely,                                                          srb@cuci.nl
           Stephen R. van den Berg (AKA BuGless).
Gravity is running out!  Conserve gravity: walk with a light step, use tape,
magnets or glue instead of paperweights, avoid showers... take baths instead.