What is distributed system?

What makes a software distributed one?

• Commodity computers talk to other computers over a network to accomplish a certain task

Distributed System = Partial Failures + Unbounded Latency

Why make software distributed?

• Data too big to fit on one machine
• You want things to be faster using more computers
• Scalable ( support more data and compute)
• Isolate failures to avoid the entire system going down

What are the main features of distributed software?

• Characterized by partial-failures
• Partial failure means some parts of the system are NOT working; some are working.
• It runs on several nodes ( node is just a computer connected by the network)
• Designing around failures: Continues to operate in the presence of failures.

In comparison with High-performance computing, it treats partial failure as a total failure.

Fundamentals of the Distributed system

• What can go wrong when two nodes talk to each other over the network?
• If you send a request to another node and don’t receive a response, it is impossible to know why.
• Byzantine Faults: Responder is lying, refusing to answer
• The sender’s message could be lost, is slow on a congested network, or the destination crashed while processing
• The responding node is slow , its response might be lost during transit, or the responder crashed

References:

• Martin Kleppmann, Lindsey Kuper