[OpenAFS] Reporting on some recent benchmark results

[Andy Cobaugh]

On 2011-04-06 at 16:06, Simon Wilkinson ( sxw@inf.ed.ac.uk ) said:
>
> On 4 Apr 2011, at 22:18, Garrett Wollman wrote:
>
>> Over the past few days I have performed several benchmarks comparing
>> the performance of various OpenAFS server and client configurations.
>
> Thanks for this - it makes for really interesting reading.
>
> The statistic I'm really interested in at present, unfortunately, isn't 
> one that you cover. With the imminent release of 1.6.0, what would be 
> really interesting to know is a direct comparison between 1.4.14 and 
> 1.6.0 on the same hardware, for the same workload. I know of workloads 
> in which I can clearly show that 1.6.0 is faster, what would be really 
> useful is to see, and to understand, is workloads for which it is 
> slower.

All of my iozone tests are here:

http://www.bx.psu.edu/~phalenor/afs_performance_results/

For each run, you get the raw iozone output, and an 'info' file that 
collects information about the client: version, memory, afsd options, and 
location of the test volume.

I'm only interested in single client, single thread performance - when 
your users are dealing with files 10's and 100's of GB in size, that's all 
you really care about.

The recent tests on the 'c2' machine are my attempt to decide whether to 
deploy 1.6.0pre4 on all of our clients in place of 1.4.14.

1.6.0pre4 with memcache is looking very promising so far, easily capable 
of saturating a gigabit connection under the right conditions..

Of course, none of our tests are with encryption turned on. In our 
experience, it's far too easy for just a few clients to bring down even 
some of our fastest fileservers when they're all on gigabit.

One observation regarding 1.6.0pre4: In stop'ing and start'ing the client 
via the init script, AFS shows up as being mounted several times in the 
output of 'mount' - is that to be expected?

--andy