Adrian Chadd disse na ultima mensagem:
> On Mon, Jun 18, 2007, Michel Santos wrote:
>
>> yes I tried it. Even if the aufs threads are running on all CPUs and it
>> gives a better performance in comparism to ufs, still, diskd runs
>> faster,
>> specially under load and lots of connections. Response time of squid
>> with
>> diskd under load is much better here.
>
> Then someone's going to have to do some digging into fixing the issue.
> To be honest, someone should do some digging into the last ten years of
> HTTP cache filesystem work and just write something to replace COSS
> that isn't so temperamental and difficult to configure; I want to do it
> but I just haven't any time anymore.
>
> (and then contribute it back to Squid, unlike 100% of the papers I've
> read on this subject which used Squid in their development and testing
> phases.)
so if I had enough money I would even spend on this project to help the
"time issue" out :) ... so if there is a developer who does not pretend to
charge as ibm or cisco consulting can contact me to fiddle something out
anyway, I have something for you Adrian ... the NASA developed in order to
sincronize earth-marth teams a clock for the MARS Rover expedidition
where the day is 39 minutes shorter (on Mars) so you take one of this
clocks running Mars-Time and you get almost an hour extra each day! And in
exchange for this particular benefit you make COSS starting faster for me
ok?
Michel
...
****************************************************
Datacenter Matik http://datacenter.matik.com.br
E-Mail e Data Hosting Service para Profissionais.
****************************************************
Received on Mon Jun 18 2007 - 10:59:52 MDT
This archive was generated by hypermail pre-2.1.9 : Sun Jul 01 2007 - 12:00:04 MDT