[Asterisk-Users] dimensioning: Where is the CPU vs Asterisk load table

Tzafrir Cohen tzafrir at cohens.org.il
Fri Jan 13 03:33:41 MST 2006


On Thu, Jan 12, 2006 at 03:05:15PM -0500, Erick Perez wrote:
> Hi, is there any good calculator/table/reference about proper dimensioning?
> I read the wiki and they basically say "xx users run fine in yy hardware"
> http://www.voip-info.org/tiki-index.php?page=Asterisk+dimensioning.
> 
> SO far I read that:
> -Run up to 4 E1s per CPU (which one? an i386 or a dual core?

This is just too general to be true. A dual-CPU PIII monster of a number
of years ago is pathertic cmpared to a single 3GHz P4.

>  -it is very CPU intensive to do transcoding. Try to minimize it.
>  -you can help the CPU by using DSP-based boards or optimized boards.

Up to a point. Normally it will be cheaper to just throw in more CPU
power.

>  -Lots of RAM is good too. (like 512MB or 8 GB?).

Depending on the task. Lots of RAM helps if you have much IO.

>  -A Front Side Bus of 800mhz is also good.

This follows the "get a fast machine" rule. 

>  -disable HT on Intel CPUs.
> -Use Ram disk to avoid some I/O bottlenecks specially on voicemail
> (hence, deploy more RAM).
> -two single core CPUs better than one dual core CPUs??.
> -And the most important I read was: Keep load under 5 in single CPUs
> and 10 in dual CPUs (didn't mention dual cores in the article).
> 
> Im not sure If Im asking properly, basically in this
> asterisk-heavy-load-learning stage, I want to know how to calculate
> computer needs based on customer needs.
> (i've only done 5 to 50 extens, and up to 2 E1)

That can probably be easily handled by the typical computer you buy on 
on any shop. Speed is not your main issue. Make sure the system is relieble.

-- 
Tzafrir Cohen         | tzafrir at jbr.cohens.org.il | VIM is
http://tzafrir.org.il |                           | a Mutt's  
tzafrir at cohens.org.il |                           |  best
ICQ# 16849755         |                           | friend




More information about the asterisk-users mailing list