monotone-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: misc stuff for mainline (was Re: [Monotone-devel] updates to net.ven


From: Eric Anderson
Subject: Re: misc stuff for mainline (was Re: [Monotone-devel] updates to net.venge.monotone.experiment.performance)
Date: Thu, 10 Aug 2006 01:09:12 -0700

Nathaniel Smith writes:
 > On Wed, Aug 09, 2006 at 10:42:37PM -0700, Eric Anderson wrote:
 > > 22457095fd36ea02a44652d45b5e7a6788cdea06: whitespace trimming
 > 
 > Looks fine.

Will separate out.

 > > -----------
 > > 4d389c13b3bb1235c720b8392f9574f1ddb72d13: inline verify
 > > 
 > > Move the verify function to be inline so that it disappears from callgrind
 > > output to make it easier to find real problems.  A non-statistically 
 > > significant reduction in user time on client and server.  
 > 
 > Umm... sure, whatever.

Will separate out.

 > > -----------
 > > 27c06ef5b20ade167011e489bc2e5333eed00faf: configurable vcache size 
 > > 
 > > Allow for size of the vcache to be set by a lua hook so that people
 > > can choose their tradeoff between memory and cpu usage.  Going from
 > > the default size to 32MiB was worth a 1.52x cpu reduction on the
 > > server, 32MiB to 128MiB was worth another 1.37x cpu reduction for a
 > > cumulative 2.08x improvement from the original setting to 128MiB.
 > 
 > Umm... sure, I guess.  In general this is probably a silly thing to
 > have configurable, but as an undocumented knob for testing, it
 > probably doesn't matter that much either way... especially since all
 > that database caching stuff is already a bit of a hackish mess and
 > cruising for a cleanup.

Why is it silly to have configurable?  If I have a server with lots of
spare memory, I set the cache size to large and trade memory for CPU.
If I have a desktop with a small amount of memory, I leave it at a
small default and trade CPU for memory?  A 2x CPU speedup is huge, but
some people may not be able to tolerate the mtn process going to
hundreds of megs (guarenteed if cache size is 128MiB).

It is worth cleaning up if possible, I modified the vcache stuff
during testing to record how many times it inserted every single key.
The hit rate on the LRU cache was very bad; I don't know if the order
things are sent out could be changed to make more things hit in the
LRU cache.
        -Eric





reply via email to

[Prev in Thread] Current Thread [Next in Thread]