Distributed systems follow certain patterns that we can follow to achieve specific purposes. This lesson will introduce those patterns and some related concepts that drive or inform them.

By understanding these patterns and what differentiates them, we can start to classify distributed systems and to understand the trade-offs between different patterns. While some problems lend themselves to specific architectures (e.g. parallel processing almost always makes use of pipelined or tree-shaped process interconnection), other choices are not as clear cut.

This lesson begins by discussing coupling, the primary driver of pain in building and operating distributed system and thus one of the key properties to control and plan for in architecting them.

Next, we discuss different choices we can make when designing a distributed system that impact the three primary parts of any computation: processing, communication, and data storage.

Distributed may not mean what you think it means. We can classify distributed systems along three different axes related to the design choices we make.

Finally, we’ll look at some specific interconnection patterns and discuss some of their properties.

Lesson Objectives

After completing this lesson, you should be able to

Describe different types of coupling and classify simple coupling scenarios according to type.
Define heterarchy and contrast it with hierarchy.
Distinguish decentralized from distributed and both of those from heterarchical.
Describe common topologies of distributed architectures and their uses.

Required Reading/Viewing

Distributed System Patterns (Lesson Slidedoc) (PDF)
Coupling and Cohesion by Sam Newman (Chapter 1 of From Monolith to Microservices)
- You should have access to this through BYU's subscription to O'Reilly library.
Heterarchy
Tiered Distribution
Re-imagining Decentralized and Distributed

Additional Resources

The surprising math of cities and corporations
Enterprise SOA: Service-Oriented Architecture Best Practices, Section 3.5
Uncoupling
Balanced Asymmetry of Networks or How to avoid Hierarchies
Event-Based Programming by Ted Faison, Chapter 1 - Coupling
- This is long. You can safely skip the mathematical properties of coupling. Focus on concepts, converting from one form of coupling to another, and how coupling can be mitigated.