Can anyone speak as to why Netflix needs so much computing power? As far as I can tell, content delivery is handled separately, so all these talks are about the non-content side.<p>Sure we need to see various personalised lists when first connecting, but no activity while actually watching things. The personalisation seems very light - eg their lists of comedies look like they show the same to everyone. And it doesn't seem intelligent at all, like for example hiding items you have already seen. <a href="https://news.ycombinator.com/item?id=9083196" rel="nofollow">https://news.ycombinator.com/item?id=9083196</a>
The SCALE 13x videos are going up here: <a href="https://www.youtube.com/channel/UCN2nbMPLJWv3Y__4JuF_hMQ" rel="nofollow">https://www.youtube.com/channel/UCN2nbMPLJWv3Y__4JuF_hMQ</a> Does anyone know in what room and day Brendan spoke?
For those who missed today's post <a href="https://news.ycombinator.com/item?id=9121015" rel="nofollow">https://news.ycombinator.com/item?id=9121015</a> because it got downvoted (wrongly) for being a duplicate: accompanying blog post and video now available: <a href="http://www.brendangregg.com/blog/2015-02-27/linux-profiling-at-netflix.html" rel="nofollow">http://www.brendangregg.com/blog/2015-02-27/linux-profiling-...</a>
I really like the workflow chart on slide 22. I can't help but think this is a useful way to visualize common workflow sequences of commands (think git add, git commit, git push).<p>Honestly I wouldn't mind if man pages came with ascii art to visualize this type of documentation.
Those flame graphs look hella useful, hopefully the JDK patches make their way in so we don't have to futz around with customizing too much. Thanks for this!
Omitting the frame pointer is a default on x86_64 its specified in the abi so it's going to be mighty tough to change that. I don't understand why you recommend not using the debug info when you already have to use it for symbol name resolution.