DragonFly users List (threaded) for 2008-01
Re: rsync vs. cvsup benchmarks
Justin C. Sherrill wrote:
The only minor thing I'd bring up is that I recall one reason for cvsup is
that rsync placed a relatively higher load per client on the server.
That needs to be established. We already heard that cvsup - contrary to
claims - is not competitive with rsync, on the client side. So I can very
well believe that this is also true for the server side. I for myself
always notice that when syncing from chlamydia, the server basically
traverses all 60k files *instantly*, while it takes quite some time on my
desktop. So the load doesn't seem to be a problem once the directory
structure is in the buffer cache.
Of course, that may complaint may date from when people only had 400Mhz
CPUs and older versions of rsync, so I doubt it's a strong reason to stay
with cvsup any more.
Quick test on chlamydia: rsync of already synced repo:
% time rsync --delete -aH chlamydia.fs.ei.tum.de::dragonfly-cvs .
rsync --delete -aH chlamydia.fs.ei.tum.de::dragonfly-cvs . 0.72s user
1.43s system 36% cpu 5.865 total
considering that rsync spends half of the time on the local side, that's <
3s of load on the server:
53331 nobody 161 0 4536K 3888K select 0:00 355.68% 33.89% rsync
nobody cares about that. it might take some more cycles when transfering,
but so what. seriously. I don't care, this is peanuts.
Serve - BSD +++ RENT this banner advert +++ ASCII Ribbon /"\
Work - Mac +++ space for low €€€ NOW!1 +++ Campaign \ /
Party Enjoy Relax | http://dragonflybsd.org Against HTML \
Dude 2c 2 the max ! http://golden-apple.biz Mail + News / \