zeeg

27 ms·

241.
by zeeg 13y ago
I have absolutely no idea how HN replies work, but this is in response to the PyPy memory usage. I'm optimizing for my application. If you want to use PyPy, use it. I couldn't care less. If I was having a hard time already max
242.
by zeeg 13y ago
No I wanted reasonable tradeoffs. e.g. 20% faster execution for 100% more memory In my case it was 15% faster execution (give or take), with a 400% memory increase. This was on a simple view which did nothing, so it's hard to tell how
243.
by zeeg 13y ago
What kind of headaches? I'm not aware of anyone actually using Mongrel 2 in a large scale Python application, but I'd be curious to know how well it works out.
244.
by zeeg 13y ago
I actually lied here, this is the 90th percentile, and average is a couple ms higher. That'll teach me to burn through a post quickly
245.
by zeeg 13y ago
I saw 5x memory usage, which was already something I was trying to optimize. There's a ticket I opened on the PyPy tracker, but this is sort of expected if I understand JITs correctly. I didn't exhaustively benchmark it, but the i
246.
by zeeg 13y ago
gevent is a very simple webserver that wont handle many things that the "real" servers do. One of the main things I'm talking about in this post is load balancing, which gevent (and uwsgi, and gunicorn, etc) don't implem
247.
by zeeg 13y ago
There was absolutely no tuning other than adjusting buffer sizes (which only matters because it accepts large POST packets). The options that a standard application will use are almost identical between the two. UWSGI just provides numerous
248.
by zeeg 13y ago
Also it does nothing with the disk, but does quite a bit over the network (queueing data, talking to caches). I'm content with 20ms. Hell, I'm content with 100ms, as long as its consistent and you can argue the value.
249.
by zeeg 13y ago
In many cases, it uncompresses 1mb of text. Bet you didn't see that one coming
250.
by zeeg 13y ago
PyPy uses too much memory and isnt that much faster for many things. (Sorry PyPy guys!) I dont have numbers/charts to pull up for this right now, but I was testing it out to see if I could push more cpu on a server, and the memory cost
251.
Using Nginx and UWSGI(justcramer.com)
113 points by zeeg 13y ago|49 comments
252.
by zeeg 13y ago
What is "close to the metal interaction" when you're talking about set of frameworks that run in an embedded interpreter that runs in a browser that runs on a GUI that runs on an OS that runs on.. oh wait.
253.
by zeeg 14y ago
If there are 1/10 developers, there are, in fact 10x developers. Guys, math and stuff. In all seriousness, there's a lot of things being done in the productivity and quality spaces to try and mine data to measure things. I recently joined t
254.
by zeeg 14y ago
You guys should feel bad about generating such awful benchmarks. How the hell can you compare accessing a MySQL database to accessing a MongoDB database? Its like comparing apples to piles of poop. Also when you're testing things like Djang
255.
by zeeg 14y ago
My point is that the metrics used are mostly meaningless
256.
by zeeg 14y ago
These stats hardly rank some companies like Disqus where they should be. In all honesty, I'm surprised (with the metrics used) that it ranks Disqus as well as it does at all. It also can't gauge the most important metric of them all: revenu
257.
by zeeg 14y ago
First time I've seen this use-case for Sentry :) (I didn't actually see Sentry within the source code, was I missing something?)
258.
by zeeg 14y ago
You won't fit that much data into Mongo anyways, so does it matter?
259.
by zeeg 14y ago
Does this quote have an origin?
260.
by zeeg 14y ago
Here's a very simple gevent hello world app. This is run from inside AWS on an m1.large: https://gist.github.com/dcramer/4950101 For the 50 dyno test, this was the second run, making the assumption that the dynos had to warm up before the
261.
by zeeg 14y ago
If this is such a problem for you, why are you still on Heroku? It's not a be-all end-all solution. I got started on Heroku for a project, and I also ran into limitations of the platform. I think it can work for some types of projects, but
262.
by zeeg 14y ago
Looks very similar to http://getsentry.com . What would you say makes your product better?
263.
by zeeg 14y ago
Agree completely. Hard to editorialize something very well when it's on TechCrunch :) We'd also rename service to platform, as an example
264.
by zeeg 14y ago
There's not, but you can check out this public event from our JavaScript client: https://app.getsentry.com/sentry/sentry-js-test/group/323091... (This was recorded before we annotated sourcemap data, but the original error here was actual
265.
by zeeg 14y ago
It's very similar to Crashlytics with the primary difference being Sentry's very strong focus on web applications (initially) and our slowly expanding support for desktop and mobile applications
266.
by zeeg 14y ago
It's not ticket tracking like those, instead we automatically collect errors within your application and aggregate them together.
267.
by zeeg 14y ago
It quickly becomes expensive if you're routinely sending a measurable amount data to it. Disqus has two physical servers dedicated to it. Both of them are larger than any of the web/worker machines getsentry.com runs. On top of those costs,
268.
by zeeg 14y ago
I'd be curious to hear if any HN readers are using Sentry and how they like the service. We started as primarily Python based, so it's especially interesting if you're using it for another language/platform.
269.
Dependency Graph Nightmares in Python(justcramer.com)
11 points by zeeg 14y ago|3 comments
270.
by zeeg 14y ago
If you call out bad code, you call out bad people? Half of you included, look at the ridiculous shit you say, and how much more offensive you are than anything anyone originally said. Also this is great for a laugh: http://dl.dropbox.c

More ›