Re: cache effectiveness testing

On 1/17/2012 12:55 PM, Taras Glek wrote:
> Hi,
> As part of the Snappy effort, we asked QA to compare cache effectiveness
> of Firefox against Opera (which some people claim has a superior cache).
> We have some preliminary data now.
>
>
> The testplan is described at
> https://wiki.mozilla.org/Performance/Snappy/Testing:BFCache_Sprint#Reference_Test_Case
>
>
> I put up a snapshot of the results at
> http://people.mozilla.com/~tglek/snappy/cache1.html
>
> I was suspecting that our cache hit-rates may be worse, but the data
> proved that theory wrong.
>
> Data indicates that either writing to our cache is expensive or we are
> not http-pipelining/etc properly. My theory is that disk cache is to
> blame, so I asked QA to rerun this comparing to Firefox to Firefox with
> disk cache disabled.

No-disk-cache followup appears to confirm that writing cache absolutely 
annihilates pageload perf.

http://people.mozilla.com/~tglek/snappy/cache2.html

I wonder if this is a symptom of 
https://bugzilla.mozilla.org/show_bug.cgi?id=695399

Taras