Hacker News new | ask | show | jobs
by mishmash 5669 days ago
So the results were worse, even though everyone involved thought they would be better on the updated version.

Any theories?

2 comments

I have no idea, I am completely baffled. I really want to try without the node_redis library and do it all in memory, keeping the Array of elements in the returned list at less than 1000.

UPDATE: I should also mention, if you look at the charts, there's a point in time which the CPU hits 99%-100% and it's always reproducible. Once the CPU hits 100% I have just sat back and watched the memory get chewed up.

I'm sure someone w/ more experience than me can probably debug what's going on. The code is there, I really hope someone else can take a look at it. As my friend mentioned, it's reproducible, it can be fixed.

Don't know if you'll see this, but as someone playing with both node and erlang, I'd love to hear more about your erlang setup (and your blog was down earlier when I looked).
tumblr :/

The way that SyncPad was prototyped was using PHP/MySQL. I ported that to using Mochiweb. I manage all of the queuing of messages through a relatively straight forward OTP gen_server and Redis.

theorising by armchair experts arent always accurate?

That sounds more confrontational than I meant it to, point if you just guess at improvements without testing them, it is still a guess.

I'm not sure Ryan Dahl should be considered an armchair expert with respect to Node.js.
as I edited that sounded more confrontational than I meant :) and I wasnt aiming it at ryan dahl (there was a lot of commentators in that thread).

I just meant that without the numbers its an guess, educated are otherwise, and we all know how many assumptions we make that end up being wrong when they are implemented.