Distributed Design Patterns

~2 mins read

https://learn.microsoft.com/en-us/azure/architecture/patterns/category/data-management

There are so many patterns but you don’t need to memorize them all because they are based on a smaller list of key ideas. When you encounter a problem, you can arrive at the same pattern using the fundamentals.

• caching

• separating interfaces

• decoupling with queues or message brokers

• reducing coordination necessity by assigning each worker a subset of work

• batching work

• reusing

• async work

• buffers

• doing work upfront to simplify later stages, like indexing

• dividing your data into more manageable pieces, sharding

• using references to avoid moving large pieces of data

• selecting an appropriate data store for the use case, like an object store for static assets, etc.

• using a gateway to aggregating multiple requests to one

• using separate subsystems for different concerns like auth, config, etc.

Many of them include a spectrum, for example

• central control vs autonomy

• divide vs aggregate

• couple vs decouple

• copies vs references

• sync vs async

Again, the way is not to memorize and apply patterns but to reason about the problem at hand