Slashdot Mirror
How Facebook Stores Billions of Photos
David Gobaud writes "Jason Sobel, the manager of infrastructure engineering at Facebook, gave an interesting presentation titled Needle in a Haystack:
Efficient Storage of Billions of Photos at Stanford for the Stanford ACM. Jason explains how Facebook efficiently stores ~6.5 billion images, in 4 or 5 sizes each, totaling ~30 billion files, and a total of 540 TB and serving 475,000 images per second at peak. The presentation is now online here in the form of a Flowgram."
I put some short video clips on Facebook's video application (just stuff of my daughter for my friends and family to see). These are AVI files generated by my digital camera, about 20-30MB in size, lasting about 1-1.5 minutes each.
They uploaded pretty quickly, but then they were put in a queue to be encoded for their flash player. It took over 3 days for them to be online in my profile! It seems they don't need to just have large capacity for storage, but a bunch more CPU for processing.
Look at the tomato! Isn't it sad? He can't dance! Poor tomato!
The issue isn't the number of bytes per second, it's the number of distinct requests. The data is _way_ bigger than will fit in memory, and hard disks can only do 100-150 seeks per second so you need a lot of them to serve from disk. A naive implementation will go to disk many times for a single file, because filesystems aren't designed for this many small files. So this is really an issue of getting exactly the right stuff in memory so you can serve hot content from memory, and if you go to disk you seek exactly once instead of several times.