Neil Harris wrote:
Oh, one last thing:
Looking at the recent Alexa traffic graphs reminded me of the recent
E-mail to Wikitech-l, telling the developers that the Wikipedia
infrastructure was (and I quote)
"U N S C A L A B L E"
I think that should now read "unscalable, apart from the ever-improving
performance and continued exponential growth in load", or something?
I wasn't reading wikitech-l at that time, I had to look up the archives
to find out what you were talking about. I for one was originally
worried about the lack of scalability inherent in MySQL, but those fears
seem to have been unfounded. Any of our database servers can keep up
with Wikipedia's write load several times over, and we've been able to
distribute the read load effectively to maintain good performance.
Growing at this rate, Wikipedia will see market saturation within the
next year or two, and I'm confident we won't hit a performance brick
wall before then.
By the way, what _has_ caused the recent traffic spike? Added hardware?
Software improvements? Better load-balancing? Something else?
According to [[Wikipedia:Announcements]] for April 2:
"Traffic continues to skyrocket, with Wikipedia's daily Alexa ranking
peaking at a daily rank of 89 and a weekly average rank of 97. Several
key placements on search engines have aided in the encyclopedia's
popularity, such as #2 for online poker, #3 for Terri Schiavo, and #9
for Pope John Paul II."
Since then we've been given special placement in Google's definition and
Q&A services, and better placement in Yahoo. The timing of this increase
in demand was fortuitous, since we solved some performance problems at
about the same time -- we took load off the image server (albert) by
moving backups to benet and implementing a metadata cache, and we added
another perlbal server, which seemed to be a bottleneck. Domas deserves
special credit, for analysing our performance problems, fixing some of
them and providing excellent advice on how to fix others.
-- Tim Starling