This is *awesome*! I have been thrilled to see the results and also impressed with the systematic approach towards attacking performance. Great job everyone!
On Friday, March 18, 2011, Robert Collins <[email protected]> wrote: > I meant to send this yesterday.. oops > > The hard time is now 11 seconds, and we saw 800 timeouts (about 0.02% > of requests) > > There are about 20 page ids we need to improve before we can drop to > 10 seconds : there are another 1000 page renders falling between 10 > and 11 seconds at the moment. That means we could be dropping the > timeout to 10 seconds as early as the start of April... as usual I'll > be reevaluating as we go. > > This is pretty exciting to me: we've come an incredibly long way since > we started systematically measuring and focusing on render times [as > the first plank in becoming fast overall] : thats a 9 second drop in 9 > months. I credit our achievement to the following things: > - you guys > - some small coding rule changes (like permitting cached property on > model objects, and query count tests) > - the restructure allowing folk the time to dig into a timeout > problem over several days without being pressured by feature work (and > vice verca, allowing features to be made fast without interrupts for > day-to-day issues) > - a willingness from our user base for us to make engineering > tradeoffs (such as moving the safety net of the hard timeout every > time we've got room to improve without causing a mass increase in the > timeouts, removing features that are underperforming or of little > benefit but high cst) > > I've been casually polling folk as we've gone through the process so > far, and in the last couple of weeks a very interesting change has > occured: rather than hearing 'gee LP is slow, its great you are > improving it, I've notice some improvements already', I'm now hearing > 'you know, I haven't been pissed off by LP being slow recently.' That > might come across as faint praise, but its really significant: > *** We've moved out of the zomg slow bucket and into the > just-another-website bucket *** > > Now, we have 2 more seconds to drop our hard timeout by, to reach > Francis' challenge of 9 seconds for the next epic : I'm sure we'll > manage that without any difficulty. > > -woo- > > -Rob > > _______________________________________________ > Mailing list: https://launchpad.net/~launchpad-dev > Post to : [email protected] > Unsubscribe : https://launchpad.net/~launchpad-dev > More help : https://help.launchpad.net/ListHelp > -- Elliot Murphy | https://launchpad.net/~statik/ _______________________________________________ Mailing list: https://launchpad.net/~launchpad-dev Post to : [email protected] Unsubscribe : https://launchpad.net/~launchpad-dev More help : https://help.launchpad.net/ListHelp
