superjared

"Massive" in this case is terribly subjective. I think the typical definition is concurrency between 10k and 50k. Whereas when we (at Urban Airship) wanted to hit 500k, the only reasonable way to do that without resorting to pure C was by using the JVM.

Finally, a use for Web Snippets.

The irony here is you'd lose serious performance since you're not using sendfile(2).

This is exactly why I created StaticGenerator for Django: https://github.com/luckythetourist/staticgenerator

"Ever get the feeling there's more to life than sitting around: sitting at work, sitting in the car, sitting in front of the TV? Do you want to get active but haven't time for the gym?"

Stop watching so much TV.

These additions make compact data storage and access much simpler from our perspective. Instead of loading and parsing an entire entry, we have access to just the data we want.

Nice work, Salvatore. Seriously.

Storm On Demand has a pretty cool offering with decent prices. They even have a "bare metal cloud".

http://www.stormondemand.com/cloud-hosting/

Is anyone here using Smalltalk for a production application? What are your thoughts on this VM?

I found the lyrics for 4.8 lacking, especially when compared with the lyrics from 4.3:

http://openbsd.org/lyrics.html#43

You forget your place, sir.

This guy must have boring coworkers. Seriously, because when I go to lunch with my colleagues we talk about work or related things and it's almost always productive. We either walk away with a more knowledge or a decision to a hard problem.

What's the status of BERT-RPC and its usage at GitHub? Have you noticed any drawbacks to it since you first created it?

That's kind of the point. Yesterday Spolsky asked why Digg had 500 servers and less traffic than StackOverflow, yet StackOverflow had 5 servers.

False dichotomies huzzah!

FYI Aimonetti's title wasn't the worst part about the presentation. Unfortunately the slides are now "private" so you can't see that.

This is especially true since you develop for WP7 on proven and well-known technologies (.Net-ish, C#, XNA). A great many developers are already intimately familiar with these and will have little or no barrier of entry with WP7.

There's also a Scala version: http://github.com/cscotta/Octobot-Scala

Which I find weird because you can write a task in any JVM language for either of the Octobots. Cool stuff.

You're probably right. I paid $800 in Nov '08.

That shows the perception between quality, though. So many people fight the fight with regard to price -- which I did, and failed. Now I have a PC which isn't really upgradeable because the CPU Socket (AM2+) and RAM-type (DDR2) are obsolete.

I'll agree with you here. I bought my wife a Mac Pro the same time I built myself a PC. I paid about 1/3 for the PC and it feels really dated. Her Mac Pro, meanwhile, is still incredibly fast. It plays SC2 on max settings without slogging while my PC slogs on Medium (and enough Marines running around).

Touché :)

Gentoo effectively copied FreeBSD's port system.

Does anyone here have feet in both camps? I'd like to read that person's opinion.

I work with the guy and know all about the project he refers to.

Hopefully I can clarify the situation. The server application he worked on was fairly new territory for him, but moreso to us. He had experience in the language (Java) that we didn't have so it was really his project. The application has had numerous subtle bugs that never caused huge problems, just weird behaviors here and there. As time pressed on, it seemed for a while that the bugs would never cease. Keep in mind that this is a server heavy with NIO internals and several worker threads, so maintaining state is a very delicate process. Race conditions that seldom occur are perhaps the worst kind of bug to tackle.

Additionally, he's the new guy. Being the new guy is hard when tasked with such an important task. Especially when the task itself seems to haunt you every day for several weeks. We encouraged him and never once blamed him for anything (as if we would have known better?). He simply took his responsibility personally. Perhaps too personally.

This sort of humanity in software is uncommon and humbling. He's a good kid who isn't out to prove anything to anyone, just to be the best programmer (person?) he can be. I think he can be forgiven for being too self-deprecating.

When using unfamiliar APIs under a deadline, this can happen. Especially when the documentation for those APIs is poor and you have to plod along in the dark.

Edit: Spelling

You are right. The part I didn't really make clear is that we only serve on the single external port. Were we to use multiple, then yes, we could have 64k * 64k per IP pair.

On an EC2 Large we hit 520K.

The C10k solutions are effectively the same as for C500k, those being epoll (Linux), kqueue (BSD), etc. Our Java NIO server utilizes epoll to handle C500k.

Regarding the `net.core` parameters. We do modify those, but my assumption (probably wrong) was that the `net.ipv4` changes would override the core configs. I'll take a look and update the post. Good point about `netdev_max_backlog`, I need to read up on that one too.

Thanks for the feedback!

Kelner is set to give a talk about it tomorrow[1]. Anyone know if they create videos for these?

[1]: http://www.csail.mit.edu/events/eventcalendar/calendar.php?s...

Octobot (like Celery, Resque, et al) is a worker, meaning that it takes messages from a queue, such as Rabbit or Redis, and processes that message based on a task that you write. Imagine Octobot being used to create thumbnails for Flickr--a job that should be done asynchronously.