I am just guessing, but you seemed to have answered your own question. if it wasn't 
dma then the cpu
would have spiked but because it is, that means the dma controller is doing all the 
work leaving the
cpu free to do some calculation.
but, its not the whole deal. the unresponsiveness could be due to the fact that 
because the dma has
control for the particular process you are working on, the cpu cannot pull data from 
memory or from
swap or from the actual data on that disk. now normally it shouldn't happen since i 
guess the
schedueler should take care of a process hogging all the io, dma, etc calls. however, 
check the nice
value maybe something is wrong around that area. run top see if something is really 
strange, check
the logs, check the "dmesg" maybe some hardware is throwing faults, check memory size, 
thrashing is
not out of question.

again, its all a guess.

> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED] Behalf Of Alex Shnitman
> Sent: Wednesday, September 10, 2003 9:46 AM
> To: [EMAIL PROTECTED]
> Subject: CPU load & load average
>
>
> Hi,
>
> I have a program that continuously writes multiple streams into files on
> the disk. I'm seeing a strange phenomenon -- the CPU load is in the
> single digits (both "user" and "system"), however the load average is
> very high, and the system is very unresponsive. The disk is running in
> DMA mode so that shouldn't be a problem. I thought maybe it's the high
> number of system calls per second that causes high context switching
> overhead, but the same program can also read those streams from the
> files instead of writing, and in that case the load average stays near
> zero, with the same number of system calls.
>
> I'm using 2.4.20 with the ext3 filesystem, tried both in data=ordered
> and data=writeback modes. Do you think it may be a filesystem problem?
>
> And in any case, I'd really like to understand how it's possible to have
> very light CPU load and a high load average. If I understand correctly,
> the latter is the average of the number of processes ready to run at any
> given moment. In that case, why is the CPU load so low -- why aren't
> they running?
>
>
> --
> Alex Shnitman <[EMAIL PROTECTED]>
> http://www.hectic.net/   UIN 188956
> PGP 0xEC5D619D / E1 F2 7B 6C A0 31 80 28  63 B8 02 BA 65 C7 8B BA
>
> =================================================================
> To unsubscribe, send mail to [EMAIL PROTECTED] with
> the word "unsubscribe" in the message body, e.g., run the command
> echo unsubscribe | mail [EMAIL PROTECTED]
>
>
>



=================================================================
To unsubscribe, send mail to [EMAIL PROTECTED] with
the word "unsubscribe" in the message body, e.g., run the command
echo unsubscribe | mail [EMAIL PROTECTED]

Reply via email to