Re: memory mapped store entries

Michael O'Reilly wrote:
>"Stephen R. van den Berg" <srb@cuci.nl> writes:
>> Michael O'Reilly wrote:
>> >Yes, but the squid executable is MUCH smaller, and more frequently
>> >accessed than the index data. Also the swap is on a seperate disk, not
>> >on the VERY busy cache disks.

>> The squid swaplog file we're mmapping *should* be on a separate disk
>> as well (it is here, it's on the log disk, it could be put on the swap
>> disk).

>Hmm. Isn't there one swaplog per cache disk?

Yes, but you can configure where they should be put. I put all
swapfiles together on a separate disk.

>> >Hmm. I'm doing over 8000 requests/min, on a Pentium-II 300, with
>> >512Meg of ram. I'm running at over 80% CPU. Using async-io with 32
>> >threads (16 won't keep up) with a 40Gig cache over a 6 disk array.

>> Interesting, what is your byte-hit-ratio?

>Depending on how you measure it, about 15%.

*That* is actually rather depressing. I get somewhere between 10% and 50%
depending on the time of day. The average being, I think, around 24%.
I would have guessed that with such a large cache, the byte-hit ratio
would rise.

>Ps. Don't get the idea that I think mmap shouldn't be added. I'd love
>to give it a go. :) These 5 mins squid restarts are a serious pain,
>and squid 1.2b23 at least has a lot of bugs that get triggered on busy
>caches during restart.

Yes, well, I think the rationale behind the pros and cons of mmapped store
entries have been discussed at length now. I think everyone understands
the potential benefits/risks now. So if someone would now comment on
the appearance of the patch, I could wrap up something and everyone can
give it a whirl.

-- 
Sincerely,                                                          srb@cuci.nl
           Stephen R. van den Berg (AKA BuGless).
Sign at a nursery school: Please pick up your children promptly after service.
                          No refunds after Wednesday.