Introduction to Distributed Systems

Last updated Nov 2, 2022

• Coursework
• 6.824

The core of a distributed system is a set of cooperating computers that are communicating with each other over a network to get some coherent task done. Examples are storage, big data computations, and peer-to-peer file sharing. A lot of critical infrastructure out there is built by distributed systems. If you’re designing a system, if you can possibly solve it with a single computer, do that. It’s always easier that way. Try everything else before a distributed system, because they’re not simpler.

1. Performance: How do I get 1000 computers to give me 1000x the compute?
   • The reason people use lots of cooperating computers is because they need high-performance, and to accomplish that they need parallelism.
2. Fault Tolerance
   • Have two computers do the exact same thing, if one fails give the task to the other one.
3. Physical
   • Some problems are naturally spread out into space and become inherently physically distributed systems.
   • Two Banks across the country dealing with the transfer of money.
4. Security
   • Improve security by splitting things between multiple computers, resulting in isolation.

Because distributed systems compute concurrently, you get all the problems that come up with concurrent programming. Complex interactions and weird timing dependant stuff make the world a scary place.

Distributed is hard because with all the different computers in addition to a network, you will have unexpected failure patterns. If you were working with a single computer, it’s usually the case that the computer is working, or it’s not. A distributed system may have multiple computers not working, or a single computer not working. A part of the network might be down. These result in partial failures.

Though we build computers to obtain high-performance, there’s a lot of roadblocks in the way. It takes careful design to get the performance you feel you deserve.