On 3 Apr 2002, at 22:53, Joe Cooper <joe@swelltech.com> wrote:
> Adrian Chadd wrote:
> > On Wed, Apr 03, 2002, Joe Cooper wrote:
> >
> >>I couldn't crash it with a polymix-4. Both commloops and
> >>chunked_mempools seems to have a harder time during fills than the older
> >>Squids, but otherwise performance is very much equivelent.
> >
> > harder time during fills?
>
> Yep...This isn't a conclusive statement of fact, but both of the 2.6
> branches I benchmarked had a slow but steady climb in latency during the
> fill--though they recovered completely for the top phases on the runs
> and performed nearly precisely equally to 2.4 Squids. This can
chunked_mempools adds distinct overhead during process size growth -
creation of chunk is heavy op. This can be the case at fill phase,
and also during sudden change from idle squid to overloaded one.
Also, when chunked_pools is unable to keep idle mem below configured
limit, it tries hard and cleans up every pool and chunks to find if
some chunk can be released. This is done every 1 second by event.
With combination of heavy pool activity, fragmentation and too low
idle limit, this can cause some overhead every 1sec.
When squid load stabilises, so also should fragmentation get lower
over time and mentioned overhead is avoided. But still, notable
increase of latency is surprising. Under normal loads, chunked
pools should be actually notably more efficient than old pools.
------------------------------------
Andres Kroonmaa <andre@online.ee>
CTO, Microlink Online
Tel: 6501 731, Fax: 6501 725
Pärnu mnt. 158, Tallinn,
11317 Estonia
Received on Thu Apr 04 2002 - 00:20:24 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:14:57 MST