Ignore the Mem: line, it's not important. It's the +/- buffers line you
need to read.

The difference is the filesystem buffers, maintained autoatically by the
OS to speed up disk I/O when there is free memory. Reclaimed
automatically if applications needs memory.

ons 2009-02-04 klockan 12:00 +0900 skrev Ja-Ryeong Koo:
>  
> Hello,
>  
>  
>  
> I am writing this email to ask something regarding "squid -z".
>  
> In advance, cheers and thanks for your time.
>  
>  
>  
> My system environment is like the follows.
>  
> -----------
>  
> Squid.2.6.STABLE22
>  
> Intel Xeon 3.0*8
>  
> 8G mem
>  
> -----------
>  
>  
>  
> Squid has been configured with COSS like the follows.
>  
> ------------
>  
> cache_dir coss /data/var/cache0/coss 4096 max-size=4718592
> block-size=512 membufs=50 maxfullbufs=24
> cache_dir coss /data/var/cache1/coss 4096 max-size=4718592
> block-size=512 membufs=50 maxfullbufs=24
> cache_dir coss /data/var/cache2/coss 4096 max-size=4718592
> block-size=512 membufs=50 maxfullbufs=24
> -----------
>  
> Cache_mem is 1024 MB.
>  
>  
>  
> The problem is,
>  
>  
>  
> before doing "squid -z", free output indicates that about 90% of total
> memory is "free memory".
>  
> But, unfortunately, AS SOON AS I do "squid -z", "free memory" on free
> output becomes very low. I am very sad.... :(
>  
> This phenomenon happens just before starting polygraph test.
>  
>  
>  
> I have searched relevant issues but I could not find clear answers for
> this.
>  
>  
>  
> The following is the result with "free -m" which is captured during
> Polytest (fill state).
>  
> -------------
>  
>                   total       used       free     shared    buffers
> cached
> Mem:          8104       7948        155          0          0
> 6370
> -/+ buffers/cache:       1577       6527
> Swap:            0          0          0
> -------------
>  
>  
>  
> I have done "swapoff xxx" before testing in order to shut down swap
> funtionality.
>  
> But, kswapd daemon consumes my CPU usage partially.
>  
>  
>  
> Why does "squid -z on COSS storage scheme" comsume so many memory
> before Polygraph testing?
>  
>  
>  
> I woud greatly appreciate any kinds of comments for this matter.
>  
>  
>  
> Thank you...
>  
>  
>  
> With the best regards,
>  
> Ja-Ryeong Koo
>  
> -- 
> Ja-Ryeong Koo,
> Department of Computer Science,
> Texas A&M University-College Station,
> TX, 77843-3112, USA,
> Phone: +1-979-204-8021
> 

Reply via email to