How Facebook Squeezes More From Its Machines

[hshah]

http://gigaom.com/2009/12/16/facebook-dyno/

[jsteele]

Oh no, have they started to embed scribd on web-pages. I absolutely hate scribd. Worst thing to hit the internet since that guy crying about Brittney Spears on youtube.

[s3graham → jsteele]

I love this Chrome extension: https://chrome.google.com/extensions/detail/nnbmlagghjjcbdhg...

Solves PDF to ~99%.

[imurray → s3graham]

Here’s a user script that does a similar in thing for several browsers: http://www.cs.toronto.edu/~murray/code/#googleviewer

[rs → jsteele]

The embedded paper is on Facebook's engineering notes as well:

http://www.facebook.com/note.php?note_id=203367363919

[houseabsolute]

Only 30,000? I would have guessed much more.

[koko775 → houseabsolute]

And you'd be right. That figure is way old, and even small for then.

[Andys → houseabsolute]

Wonder how many of them are just almost idle, running something like memcached?

[lsb → Andys]

Facebook's actually encountered scaling problems from memcached taking up too much CPU time; that's why there's the binary protocol, and they're talking about the multiget hole at http://highscalability.com/blog/2009/10/26/facebooks-memcach...

[grinich → houseabsolute]

They're all in the US, too.

[mtw]

any link with more info about Dyno? the article assumes readers know about this testing tool.