Re: cache effectiveness testing

assuming you want to measure the benefit/savings of response caching you 
don't want to bypass dns and connection setups - while variable, they 
are a big part of the caching savings and the reason you want to cache!

you also need to measure/model it across a variety of RTTs - the benefit 
varies according to the network - and decide what ranges are important 
to you. Which makes sense becuase the value of caching a slow object is 
higher than caching a fast object. (and I personally think that should 
be part of the cache-replacement function.) Cache benchmarks taken over 
unshaped lans and localhost are just silly - its all overhead without 
any upside.

If you're just trying to measure overhead, that's another story.

On 1/20/2012 11:29 PM, RIUM+ wrote:
> Since we're just trying to isolate the impact of caching&  nothing
> else, it may be advisable to perform an initial run through the sites
> in a Private Browsing session, to prime the network connection. This
> would remove variability related to caching DNS requests&  other
> network issues. That can easily add a few hundred milliseconds of
> delay to whichever software package you test first (in this case
> Firefox run 1).
> _______________________________________________
> dev-tech-network mailing list
> [email protected]
> https://lists.mozilla.org/listinfo/dev-tech-network