Formal specifications for microservices

This is a lightning talk presented at DDD Europe 2020, introducing TLA+ and a Petri Net based language to translate DDD concepts into an executable graph.

Transcript

1. Formal specifications for Microservices Lightning talk © GuanxiLabs 2019
2. Hello 2 François Royer CEO & Principal Consultant francois.royer@guanxilabs.com +33 (0) 6 72 33 75 02
3. What are formal methods and why should you consider them? « Engineers naturally focus on designing the ‘happy case’ for a system, i.e. the processing path in which no errors occur. […] That code path must solve the customer’s problem, perform well, make efficient use of resources, and scale with the business; these are all significant challenges in their own right. Once the design for the happy case is done, the engineer then tries to think of “what might go wrong?”, based on personal experience and that of colleagues and reviewers. The engineer then adds mitigations for these classes of scenarios, prioritized by intuition and perhaps some statistics on the probability of occurrence. » Use of formal methods at Amazon Web Services (2014) Chris Newcombe, Tim Rath, Fan Zhang, Bogdan Munteanu, Marc Brooker, Michael Deardeuff
4. Meet TLA+, by Leslie Lamport The problem definition Invariants and liveness properties (The trains can’t be stuck)
5. Formal methods such as TLA+ can help design complex distributed systems Use of formal methods at Amazon Web Services (2014) Chris Newcombe, Tim Rath, Fan Zhang, Bogdan Munteanu, Marc Brooker, Michael Deardeuff
6. Microservices are hard and are naturally a target for DDD practitioners Building systems using microservices requires us to think more deeply about failure isolation and testing. TLA+ as a formal specification language can be useful in both these scenarios. For failure isolation, TLA+ can be used to identify invariants in your system that can be monitored directly. An invariant can be the ratio of number of requests to one service to the number of requests to a second service, for example. Any change in this ratio would lead to an alert. Source: Thoughtworks Radar
7. The problem … TLA+ can be a significant investment upfront for most projects • Everything is a state machine, which is fine if you’re a physicist or academic CS… • The domain needs to be completely mastered • The syntax takes some time to be accustomed to…
8. DDD methods are great at defining microservice boundaries and events…
9. Software developers must still understand the system behaviour and produce formal specs SHARE AS DOC EXPORT AN API SPEC FOR THE DEV TEAM DEPLOY ON ANY CLOUD FOR COSTING ANALYSIS
10. The DDD ontology appears to define its own regular language, i.e. a state machine… © Alberto Brandolini Entity Repository Factory Value Object Hierarchy And other concepts from the blue book…
11. Let’s build a set of primitives, starting with Commands and Events …
12. … and add time dependency to satisfy liveness …
13. … and finally Entities to add persistence to the workflow
14. Views, bounded contexts follow the same conventions (drag, overlap, link…)
15. francois.royer@guanxilabs.com Interested? www.txime.com @TximeApp