Re: some linux tuning (fs benchmarks)

Alex Rousskov writes:
> On Mon, 28 Jun 1999, Doug Renner wrote:
> > FS speed, specifically I/O bandwidth, is uberimportant to caches.
>
> I/O bandwidth is not a bottleneck for most Squid setups. File system
> overheads, blocking processes, no application level seek optimization,
> etc. are.
>
> > Unless we already have a standard benchmark, I propose that we create
> > one. I would like to see published results in some standard, moderately
> > trustable, format. Maybe there is already a project that has FS specific
> > benchmarks. I would like to see them with these different hardware
> > combinations.
>
> Testing raw disk performance or raw file system performance seems
> counter-productive to me. You can use Web Polygraph to test disk
> subsystem of a proxy with Web-like workload:
> http://polygraph.ircache.net/

  I think that Doug's suggestion was to measure file system performance
independently, to be able to understand it as a limiting factor in
Squid or other proxy performance. Raw disk performance is irrelevant,
I agree. However, if file system throughput is an asymptotic limit for
performance tuning of a proxy, then that's an argument for focusing on
that area as one approach to abstracting "intrinsic" Squid performance.
I thought one user's result that Squid delivers over 45Mb/s when
running purely from memory was very interesting.

  However, I think there already are some standard cross-UNIX
benchmarks for file system throughput. Trying to devise one here is
reinventing the wheel - to avoid unnecessarily reinventing the
industry-standard cliche. Some quick web searches have already turned
up references to the SPEC SFS97 and LADDIS (SFS93) benchmark for NFS,
an Andrew benchmark, and a Chen benchmark. These would probably be
good places to start, if people want to start evaluating performance of
their systems. Simply reporting sequential file write times is usually
not useful.

  -- Clifton

-- 
 Clifton Royston  --  LavaNet Systems Architect --  cliftonr@lava.net
        "An absolute monarch would be absolutely wise and good.  
           But no man is strong enough to have no interest.  
             Therefore the best king would be Pure Chance.  
              It is Pure Chance that rules the Universe; 
          therefore, and only therefore, life is good." - AC