"Cal Leeming [Simplicity Media Ltd]" <cal.leem...@simplicitymedialtd.co.uk> writes: > Came across the following blog entries today, for those interested in this > sort of thing: > > http://attentionshard.wordpress.com/2011/04/26/scaling-django-to-30000-requests-per-second/ > > http://attentionshard.wordpress.com/2011/04/03/scaling-to-30k-two-level-caches/ > Interesting approach. Although you'd have to be careful what context it was > used in (i.e. if your code is written under the assumption that the caching > server is atomic). > > http://attentionshard.wordpress.com/2011/04/26/scaling-to-30k-tsung/ > Never heard of tsung before, looks pretty nice. Will try it out for sure. > > http://attentionshard.wordpress.com/2011/05/03/scaling-to-30k-haproxy-on-ec2/ > Can't comment much on this, as we've never used haproxy. > > The stack they have used is quite interesting. Although they are using > Apache w/ mod_wsgi (which tends to be a lot slower than using nginx with > uwsgi), they still seem to have got some decent performance out of it. > > http://attentionshard.wordpress.com/2011/06/21/behind-the-scenes-using-cassandra-acunu-to-power-britains-got-talent/ > Explains a bit of their usage of using Cassandra. Would be interesting to > see some benchmarks though. > > Cal
Hi Cal, really thanks for sharing these posts! Malcolm Box <malcolm....@gmail.com> writes: > Thanks for the recommendations Cal. I hope this stuff is of use to some > people - let me know if anyone wants to know more. > > On 10 July 2011 21:06, Cal Leeming [Simplicity Media Ltd] < > cal.leem...@simplicitymedialtd.co.uk> wrote: > >> Came across the following blog entries today, for those interested in this >> sort of thing: >> >> >> http://attentionshard.wordpress.com/2011/04/26/scaling-django-to-30000-requests-per-second/ >> >> >> http://attentionshard.wordpress.com/2011/04/03/scaling-to-30k-two-level-caches/ >> Interesting approach. Although you'd have to be careful what context it was >> used in (i.e. if your code is written under the assumption that the caching >> server is atomic). >> > > Actually atomicity is preserved for updates, as all updates go to the L2 > cache servers (so are as atomic as memcache over multiple servers is). The > main gotcha is that you have to be happy that different servers can have > values that are up to N seconds out of date - but generally that's not too > bad, as it's pretty much the same situation as using a reverse proxy cache. > > >> >> http://attentionshard.wordpress.com/2011/04/26/scaling-to-30k-tsung/ >> Never heard of tsung before, looks pretty nice. Will try it out for sure. >> >> > Can't recommend Tsung highly enough. It Just Works for testing large numbers > of concurrent connections, which is more than I can say for other tools I've > tried (apachebench, JMeter, swarm of bees). There's several companies around > that offer to do high load testing but last time I checked their rates were > sky-high. +1. Tsung was maybe the only tool fit for this kind of testing, a while back, when I was doing some scaling tests. > > >> The stack they have used is quite interesting. Although they are using >> Apache w/ mod_wsgi (which tends to be a lot slower than using nginx with >> uwsgi), they still seem to have got some decent performance out of it. > > > Indeed, it's on the to-do list to try comparing the performance of > Apache/mod_wsgi with nginx/uwsgi, but my gut feel is that the webserver + > WSGI container makes only a marginal difference to the overall site > performance. Of course, I Could Be Wrong. > I had exactly the same thoughts although I'm just intrested (out of plain curiosity) in how nginx/uwsgi scales with and without the caching framework. It would be intresting though to see the difference between the two stacks. >> >> >> http://attentionshard.wordpress.com/2011/06/21/behind-the-scenes-using-cassandra-acunu-to-power-britains-got-talent/ >> Explains a bit of their usage of using Cassandra. Would be interesting to >> see some benchmarks though. >> > > What benchmarks would you like to see? We sustained 10K writes/second into a > 2 node m1.large cluster if that helps. > > Malcolm -- You received this message because you are subscribed to the Google Groups "Django users" group. To post to this group, send email to django-users@googlegroups.com. To unsubscribe from this group, send email to django-users+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/django-users?hl=en.
