CSE Roadmap

Core tier · about 4 hours · 8 parts

Part 4 of 8 · 1 min read

The big decisions

Monolith or distributed, where the data lives, sync or async.

After collecting inputs, FSA says the architect makes three determinations in order.

Decision 1: monolith or distributed. Use the quantum idea (page 4). Does one set of characteristics suit the whole system, or do parts need different ones? One set suggests a monolith, although other factors can still push toward distribution. Different sets mean a distributed style. Page 6 showed this with the auction: bid capture for thousands of bidders needs scale, while capture for the single auctioneer needs reliability.

Decision 2: where does the data live? A monolith typically has one relational database or a few. In a distributed style the architect decides which services keep data and how data flows to build workflows (page 5, page 8). FSA says to think of structure and behaviour together and not to fear iterating.

Decision 3: synchronous or asynchronous between services? FSA's rule is short: use synchronous by default, and asynchronous when necessary. Synchronous calls are simpler to design, build and debug, but can hurt scalability and reliability. Asynchronous calls bring benefits in performance and scale, but also data synchronisation, deadlocks, race conditions and hard debugging (page 7).

The output of the process, in FSA's words, is a topology, records of the hardest decisions (architecture decision records from page 1) and fitness functions, which are automatic checks that protect key principles and operational characteristics.

Read it as

Start simple at each step. One deployment unless different parts need different characteristics. One shared data store unless the services must be independent. Synchronous unless a specific need forces a queue.

Check yourself: 1. Which question comes first, data or deployment? Why? 2. What is FSA's default for communication? 3. Name the three outputs of the choosing process.