Apache ZooKeeper and The Art of Building Distributed Systems

Implementing distributed systems can be hard. Servers crash, become slow, and get partitioned away. These are all events that can happen in a real setting and your distributed system needs to be ready to deal with them.Apache ZooKeeper has been developed to deal with such problems. It is a replicated in-memory system that stores small files called znodes in a hierarchical manner. The ordering guarantees operations over znodes and the notification scheme it provides enables the implementation of a number of commonly used recipes, like master election, group membership, locks, barriers, etc. Without a component like ZooKeeper, implementing such recipes from scratch can be a significant burden because they require sophisticated algorithms with many corner cases that are easy to overlook. ZooKeeper, however, does not make the promise of completely hiding all the problems of a distributed system, but instead to simplify the task.This webcast covers:Some basic concepts of ZooKeeperDesign choices, and caveatsExamples of recipes one can implement with the ZooKeeper APIElaborate on the potential problems with naive recipesDiscuss ways around such problems