Gilles Dubuc, 09/01/2015 10:46:
Right now 17% of image requests in Media Viewer are Varnish misses, and
99.5% of those are pulling an existing thumbnail from Swift. Varnish
misses are twice as slow as hits on average.
Very interesting.
I plan to disable pre-rendering next week in order to confirm these
findings and determine for certain what percentage of image requests
pre-rendering is useful for on the set of sizes Media Viewer currently uses.
Isn't there another way to differentiate the images affected
pre-rendering? Maybe images uploaded before a certain time, or similar.
Gergo Tisza, 09/01/2015 20:54:
> IMO the next step should be differentiating between Varnish delay and
> network delay.
Or ensuring that we're not comparing apples and oranges. Considering
thumbnails of similar size and format would be an improvement. Some
formats, like GIF and very detailed PNGs, are both rarer and
significantly heavier (possibly also slower to parse).
>
> suggests that most of the time is network delay between Varnish and the
> browser so it is more useful to think about CDNs than about caching
> strategies.
Hm. If this were true, would it mean that varnish misses correlate to
geography?
> (Also I wonder what the missing 20% is. Seems a bit high for DNS / TCP
> handshakes.)
Those are a substantial portion of load time on many of our pages which
load resources from multiple domains (typically, current + bits + upload
+ login + meta for CentralNotice + another wiki for CSS/JS).
The connection view of
http://www.webpagetest.org/result/140605_N7_CSX/1/details/ seems a
typical example.
Nemo
_______________________________________________
Analytics mailing list
Analytics@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/analytics
