On 25 Dec 2002 at 8:52, Suresh Ramasubramanian wrote:

> That has a ton of 2.5.x patches (including most of Ingo's stuff)
> backported into it.

While I do acknowledge RedHat's contribution and patching, this is just to 
point out that 2.4.18-x onwards kernel has O(1) patch already backported. Only 
patches that are seriously missing and not integrated in main 2.4.x line are 
pre-memptive and low latency patches, I believe.

I do not think disk I/O has anything to do with CPU load. It is SCSI remember? 
If disk I/O or file sysetm are not upto it, it would affect throughput. For no 
reason, it should cap out CPU.

I believe culprit is squid. AFAIK, it has provision for using POSIX async I/O 
which is not really good in linux IMO. If squid options are tuned, it should be 
possible to bring down CPU usage. I agree that does not guarantee thourghput 
though.

I have done heavy testing on MDK9.0 with 40GB and 60GB databases on compaq 
machine. OK it was quad xeon and 4GB RAM machine, but CPU was no way noticable 
factor in any of the cases. It is the disk I/O and SCSI do not load CPU that 
way.

I agree with the sugestion of file system. I suggest reiser as it is very fast 
with small files and that is what large part of them in squid cache, isn't it?

Of course, you try and decide what fits best for your problem. but I don't 
think it has to do anything with kernel provided you have 2.4.18-x onwards..

HTH
Bye
 Shridhar

--
Larkinson's Law:        All laws are basically false.



-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
linux-india-help mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/linux-india-help

Reply via email to