CSx55: Distributed Systems

• Thread safety and concurrent programming
• Algorithms underpinning peer-to-peer systems and distributed hash tables including for systems such as Chord, Pastry, Tapestry, Napster, Gnutella, and BitTorrent.
• Foundational issues (including Brewer's CAP theorem) in the design of cloud scale storage systems such as the Google File System and Amazon Dynamo.
• Design considerations in systems for scalable analytics such as MapReduce and Spark.
  

Course Objectives
By the end of the course, students should be able to

1. Build scalable servers using thread pools that leverage lightweight concurrency primitives. This is
2. Design and build distributed, fault-tolerant file systems from the ground-up
3. Write MapReduce programs that execute concurrently on multiple machines
4. Design efficient data representation formats for communications between distributed
   components
5. Understand the role of overlays in content disseminations
6. Scale as the number of entities in the system increase
7. Can handle and process large data volumes using Hadoop/MapReduce and Spark
8. Design systems that are loosely coupled, transactional and eventually stable. These systems must be able to sustain failures and recover from them.
9. Understand issues related to the role of order and distributed mutual exclusion