Re: [patch 2/2] fix file counting

[Dipankar Sarma <dipankar@in.ibm.com>]

On Tue, Jan 31, 2006 at 11:33:49AM +0100, Eric Dumazet wrote:
> Dipankar Sarma a écrit :
> >>Putting an atomic op into the file_free path?
> >
> >Here are some numbers from a 32-way (HT) P4 xeon box with kernbench -
> >
> 
> Hi Dipankar, thank you very much for doing these tests.
> 
> To have an idea of the number of files that are allocated/freed during a 
> kernel build, I added 4 fields in struct files_stat.
> 
> # cat /proc/sys/fs/file-nr
> 153     0       206071  153     755767  755620  5119
> 
> So thats (755767-39169)/(4*60+56) = 2420 files opened per second.
> 
> Number of changes on central atomic_t was :
> (5119-1131)/(4*60+56) = 13 per second.
> 
> 13 atomic ops per second (instead of 2420*2 per second if I had one 
> atomic_t as in your patch), this is certainly not something we can notice 
> in a typical SMP machine... maybe on a big NUMA machine it could be 
> different ?

Depends on what you call big :) The one I ran on is a 2-node NUMA
16(32)-way with HT. I can't find anything bigger than this
in our server farm. On a 8-way ppc64 box too, there was no difference
at all.

Can you think of some non-microbenchish workload with more open/sec ?
I privately pointed out this thread to John Hawkes from SGI to see
if they care about the potential issues. In the mean while,
I will investigate some more.

Thanks
Dipankar