[OpenAFS-devel] Default params for afsd?

[Jeffrey Hutzelman]

On Friday, May 13, 2005 09:43:05 AM +0200 Niklas Edmundsson 
<Niklas.Edmundsson@hpc2n.umu.se> wrote:

> I have a faint memory of the memcache only being able to store
> cachesize/chunksize number of files even if the files are much smaller
> than chunksize, ie that the memcache is an array of chunksize blocks. If
> this is correct, then you should probably reduce the chunksize when doing
> small-sized-memcache to be able to fit enough files.

You bring up an important point which I failed to mention.  The proposals I 
made regarding chunkSize, cacheFiles, and dCacheSize are all relevant only 
when using disk cache.  The memory cache architecture works somewhat 
differently; memcache chunks are fixed-size and take up the same amount of 
storage regardless of how "full" they are.  So, the cache can store only 
cachesize/chunksize chunks.

As you suggested, this argues for much smaller chunks than would be used 
for a disk cache, and in fact the default chunkSize for memcache is only 13 
(8K).  I'm not suggesting changing this default, though of course folks who 
think they can spare the memory could choose to raise it (and cacheBlocks) 
an order of magnitude or two.

Furthermore, separate configuration of cacheFiles and dCacheSize is 
meaningless.  The value of cacheFiles is ignored entirely; the number of 
chunks is exactly the number that will fit.  The value of dCacheSize must 
be the same as the number of chunks, because memcache has no CacheItems 
file, and the dcache is the authoritative index rather than merely a cache. 
In practice what this means is that you cannot simultaneously specify both 
cacheBlocks and dCacheSize; whichever is left out will be computed from the 
other.

> Regarding the memcache, is there a known maxsize? Again, my faint memory
> reminds me of someone mumbling that noone has tested memcache at sizes a
> modern machine can handle (ie gigs of memcache).

I do not believe there is a known maximum; I've certainly not used a 
memcache on that order.  We used to use memcache many years ago during 
Linux installs, because it was easier to configure in the limited 
environment available in the installer at that time.  These days we use a 
disk cache on an in-memory ext2 filesystem.



> A clear step up from the current situation. I have to thank Jeffrey for
> his thorough explanation of the different tuning options and I sincerely
> hope that it finds it way into documentation and/or the afsd sourcecode.
> It's clearly better than what's in there, especially the comments like
> "another guess, perhaps Honeyman will have a grad student write a paper"

Well, you know, if Peter had ever found someone to write that paper....

-- Jeffrey T. Hutzelman (N3NHS) <jhutz+@cmu.edu>
   Sr. Research Systems Programmer
   School of Computer Science - Research Computing Facility
   Carnegie Mellon University - Pittsburgh, PA