[Bioclusters] cache limit?

Joe Landman bioclusters@bioinformatics.org
04 Mar 2003 16:32:57 -0500

Hi Justin:

  The 2.4 cache behavior is substantially different than the 2.2.  After
the BLAST job ends, I believe the mmap'ed area is returned to the heap. 
The mmap'ed db is likely to be fully cached though (if it is small
enough), which means that mmap can create a small area, and simply
stream it in from cache.  This can be a good thing.

  Also, have a good close look at your limits.  You might need to adjust
them before the run.  The kernel parameters are tunable, but 2.2.16 is
somewhat on the old side (so is 2.4.2 for that matter), so you might not
find relevant information readily available.

On Tue, 2003-03-04 at 10:16, Justin Powell wrote:
> I've got a small blast cluster which includes a Quad Processor Dell
> (kernel 2.2.16-22enterprise #1 SMP) and a bunch of Dual Athlons (kernel
> 2.4.2-2smp #1).  Repeat BLASTS of the same DB on the Athlons run about 7
> times faster than the first BLAST, I assume this is due to the memory
> mapped dbs being retained in the cache, but this is not the case with the
> Dell.  When I use 'free' to look at memory the cache on the Dell never
> goes above 850MB whereas that on the Athlons happily goes up to 3.5GB
> after identical BLAST jobs.
> All systems have 4GB ram.  Are there some obvious parameters which can be
> adjusted to alter the behaviour of the cache under Linux (I can't find
> obvious references to such on Google) or is there some difference in the
> memory management of different versions of the kernel which makes some of
> them less useful for this sort of thing?
> Justin Powell
> jacp1@mole.bio.cam.ac.uk
> _______________________________________________
> Bioclusters maillist  -  Bioclusters@bioinformatics.org
> https://bioinformatics.org/mailman/listinfo/bioclusters
Joseph Landman, Ph.D
Scalable Informatics LLC,
email: landman@scalableinformatics.com
web  : http://scalableinformatics.com
phone: +1 734 612 4615