Realtime at Facebook

This talk will cover how Facebook collects data in as reliable a fashion as possible given machine failures and transient network outages. It will cover the data flow from "the outer rim" known as scribed, through "scribeh" (which is a legacy name--it speaks scribe-thrift, but is pure java), and the vital end component, ptail, a client app.

Speaker Sam Rash will also discuss a flagship application on top of this store-and-forward network with the twist of data that can be queried at any hop. That application is Puma, a system that aims to be the "Hive" of both streaming queries and moderate rollups of those results.