Buildy: 1k HN users placed 250k objects in our HTML5 sandbox(blog.playbuildy.com)
blog.playbuildy.com
Buildy: 1k HN users placed 250k objects in our HTML5 sandbox
http://blog.playbuildy.com/2012/11/14/hn-postmortem/
3 コメント
Thank you for maintaining sockjs-tornado! Performance has been shockingly good.
Setup: sockjs-tornado behind haproxy (on the same ec2 m1.large machine)
Over the 6-hour period the post was at the top of HN, there were about 480 concurrent users. At that time, the python process was using a grand 33 megabytes of memory and the overall system load was at 6%. Go sockjs-tornado :)
In the past couple days we've pushed 12+ gigabytes of real-time data through the server without restarting it. Memory usage has grown to a measly 60mb.
2. We've been using sockjs-tornado to push around ephemeral world update events, and Firebase for more persistent state-sync things like chat and presence. That's a pretty vague answer, I'll write it up better in a blog post!
Please feel free to reach out if you have any more questions -- my address is in my profile
Setup: sockjs-tornado behind haproxy (on the same ec2 m1.large machine)
Over the 6-hour period the post was at the top of HN, there were about 480 concurrent users. At that time, the python process was using a grand 33 megabytes of memory and the overall system load was at 6%. Go sockjs-tornado :)
In the past couple days we've pushed 12+ gigabytes of real-time data through the server without restarting it. Memory usage has grown to a measly 60mb.
2. We've been using sockjs-tornado to push around ephemeral world update events, and Firebase for more persistent state-sync things like chat and presence. That's a pretty vague answer, I'll write it up better in a blog post!
Please feel free to reach out if you have any more questions -- my address is in my profile
great follow up. I was on the site and remember being pleasantly surprised by the good performance, so As for the ideas for technical posts mentioned at the end, I'd be most interested in the realtime event publishing with firebase and tornado and rendering and optimization engine stuff :)
Very interested in the MongoDB and MySql issue, and trade-offs you've perceived with each.
Two questions:
1. As sockjs-tornado maintainer I'm very interested in project performance: what is average server load, how memory consumption looks like, etc;
2. Why do you use both Tornado and Firebase and not just one of them?