Re: [squid-users] Re: Unbalaned Cpu cores with squid 3.4.3 with centos 6.4 64 bit

On 16/02/2014 12:54 a.m., Dr.x wrote:
> hi alex ,
> [root_at_squid ~]# egrep 'cpu cores|physical id' /proc/cpuinfo | sort -u
> cpu cores : 6
> physical id : 0
> physical id : 1
>
> what i was meaning is that low byte ratio !!!!!
> low bandwidth saving ??!!!!
>
> when using noraml squid 1 instaned with 3 hardsisk each with 90 giga ,
> i was saving about 30-40Mbps
>
> now
> with same load and same users and same BW
> im only saving about 10-12Mbps !!!!!!

Total? or per-worker? where is that measurement coming from?

>
> that wt i was meaning .
>
> the question is frequently being asked .
>
> how i implement sauqid so that it handle more requests (SMP) and make more
> bandwith saving??
> i think the biggest problem of smp is that shared issue which limitation in
> memeory and rock store to 32 KB

Many others have agreed with you on that one. Some even paid for the
work to be done. Squid currently meets their needs more than others,
just because they could pay for the development work in specific areas.

There is still a fairly long way to go overall. So all help whether it
be finding, fixing and/or testing these things is welcome.

>
> why all the new efforts of squid go to increase the shared memory and shared
> objects to about 1M ???
>

Can you explain what you mean by that a bit more?

> this is a big truoble in my think and need to be fixed .
>
> agian , we all appreciate ur efforts and tired in updating squid and fixing
> bugs
> but i thuink this issue need a priority in work with squid developers
>
>
> agian , gret thankfull for you (Alex) and for squid team developers
>

Thank you, on everyones behalf.

Amos
Received on Sat Feb 15 2014 - 14:10:29 MST