Bogofilter simulator.

[Matthias Andree]

David Relson <relson at osagesoftware.com> writes:

> Matthias,
>
> Here're my results:
>
>
> mount - /dev/sda1 on / type ext3 (rw)
>
> Adaptec AIC7XXX EISA/VLB/PCI SCSI HBA DRIVER, Rev 6.2.8
> <Adaptec aic7890/91 Ultra2 SCSI adapter>
> aic7890/91: Ultra2 Wide Channel A, SCSI Id=7, 32/253 SCBs
>
> SEAGATE   Model: ST34520W          Rev: 1281
> Direct-Access                      ANSI SCSI revision: 02
>
> [relson at osage MatthiasAndree]$ time ./simbf > /tmp/junk
> 0.39user 13.55system 2:25.03elapsed 9%CPU (0avgtext+0avgdata 0maxresident)k
> 0inputs+0outputs (76major+16minor)pagefaults 0swaps

Andrew Morton essentially says 1. the many overwrites (I estimate we
have around 99% in Greg's setup) are evil, and 2. the interval between
writes (5 s in ext3, 30 s in ext2) distorts our results and may make for
exponential increase of write times, because overwrites can be locked
out until the writes have completed, and the longer it takes, the more
flush operations the kernel will conduct; other file systems might also
show these problems as the size of our DB increases or the commit
intervals decrease. All that sounds pretty reasonable IMO.

So we should see if we can reduce the number of overwrites somehow and
improve the locality. I'll think about that later.

If your kernel supports it, try mounting with "commit=30" and see if
that helps.

-- 
Matthias Andree