4 ms·
As these slides are very old, here's some updates: * We use Flask and nginx in various areas now (the main app is still Django). Our realtime app, for example,
by zeeg 14y ago
As these slides are very old, here's some updates:
* We use Flask and nginx in various areas now (the main app is still Django). Our realtime app, for example, is powered off of uwsgi and Flask.
* There are nearly 1b monthly uniques across the network serviced by the platform.
* ~300 servers
* Still Postgres (with Slony, multiple clusters), Redis, Memcache, and some Cassandra for newer things (not comments).
Also mostly confident we're still the "largest Django app" in terms of traffic.
- recuter 14y agoInteresting. Care to update on the other stats like requests at peak? Also, I wonder - if you're at 1b uniques is there anywhere to grow from here other then more engagement? Oh, also: Would you say the 1% contribute, 9% curate, 90% consume heuristic holds true in your experience at this scale or is it even more skewed?
- reinhardt 14y agoVirtual servers or bare metal? What do you use for herding them?
- lucian303 14y agoI second that. Would be interesting to know a bit more about the systems setup. Also why the decision to use an ORM? To manage the partitions?
- zeeg 14y agoEventually you need abstraction. Python is fast enough, as is Django, that the abstraction costs us less than the value it provides us. That said, we do manage our partitions (currently) using an ORM layer (at the application level). We do want to break this out into a proxy middleware at some point though.
- lucian303 14y agoRight, just interested in the choose of such an anti pattern. What proxies are you considering?