zfs very poor performance compared to ufs due to lack of cache?

[Jeremy Chadwick]

On Wed, Sep 15, 2010 at 10:42:24AM +0300, Andriy Gapon wrote:
> on 14/09/2010 20:30 Steven Hartland said the following:
> > Going to run the following tests:-
> > 1. run a live test with "sendfile off" in the nginx config
> > 2. run a live test with "sendfile on" in the nginx config.
> > 
> > During these tests I'm going to monitor the following every minute:-
> > time, kstat.zfs.misc.arcstats.size, vm.stats.vm.v_pdwakeups,
> > vm.stats.vm.v_cache_count, vm.stats.vm.v_inactive_count,
> > vm.stats.vm.v_active_count, vm.stats.vm.v_wire_count,
> > vm.stats.vm.v_free_count
> > 
> > Anything else that should be monitored?
> > 
> > Before each test the machine will be rebooted to try to ensure as direct a
> > comparison
> > as possible.
> > 
> > Anything else that I should add / change before running said tests or should
> > monitor?
> 
> This sounds sufficiently good.
> If you could arrange to draw the graphs of the data it would be terrific :)

Please be aware the OP is using RRDTool to store the sample data, which
means the values you see in the graphs are going to be averaged unless
he's taken the time to use MIN/MAX/LAST on both the CF and the DS (there
is a difference):

> > Now monitoring these each minute to an rrd and text file and updated
> > 8-STABLE ...

What I'm trying to say: averaged data may not show you what you're
looking for, depending on what that is.  :-)

-- 
| Jeremy Chadwick                                   jdc at parodius.com |
| Parodius Networking                       http://www.parodius.com/ |
| UNIX Systems Administrator                  Mountain View, CA, USA |
| Making life hard for others since 1977.              PGP: 4BD6C0CB |