Re: [Ext2-devel] Re: [RFC][PATCH 0/2]Extend ext3 filesystem limit from 8TB to 16TB

[Ravikiran G Thirumalai <kiran@scalex86.org>]

On Tue, Apr 18, 2006 at 12:57:00PM +0200, Laurent Vivier wrote:
> 
> I made tests on same system (x440) with postmark-1.51 :
> 
> pm> set numbers 100000
> pm> set transactions 250000
> pm> run
> 
> With atomic_t:
> 
> Time:
>         3761 seconds total
>         2414 seconds of transactions (103 per second)
> 
> Files:
>         225064 created (59 per second)
>                 Creation alone: 100000 files (87 per second)
>                 Mixed with transactions: 125064 files (51 per second)
>         124961 read (51 per second)
>         124895 appended (51 per second)
>         225064 deleted (59 per second)
>                 Deletion alone: 100128 files (503 per second)
>                 Mixed with transactions: 124936 files (51 per second)
> 
> Data:
>         731.14 megabytes read (199.07 kilobytes per second)
>         1359.02 megabytes written (370.02 kilobytes per second)
> 
> With percpu_counter:
> 
> Time:
>         3787 seconds total
>         2422 seconds of transactions (103 per second)
> 
> Files:
>         225064 created (59 per second)
>                 Creation alone: 100000 files (85 per second)
>                 Mixed with transactions: 125064 files (51 per second)
>         124961 read (51 per second)
>         124895 appended (51 per second)
>         225064 deleted (59 per second)
>                 Deletion alone: 100128 files (503 per second)
>                 Mixed with transactions: 124936 files (51 per second)
> 
> Data:
>         731.14 megabytes read (197.70 kilobytes per second)
>         1359.02 megabytes written (367.48 kilobytes per second)

Can we get oprofile output for these tests please?  It will give us a clue as
to how much of hot spots the ext3 atomic counters are with this benchmark.
Also, it will be nice to have results for 3-5 iterations of the test
to make sure we are looking at statistically significant numbers.

Thanks,
Kiran