The service that is likely the most useful in this is RevertRisk-Multilingual: it's a service that we want to continue to support, it would benefit from faster response times, and its req/response semantics are well suited to caching.
The steps needed to implement this are:
- Set up user(s) and permissions on Cassandra
- Add Istio config to make Cassandra reachable from LW
- Copy RRML repo to new RRMLca repo for experimentation
- Setup RRMLca LW service config (Deployment service etc), including Cassandra config
- Decide schema for caching table(s)
- Add code to RRMLca repo to read from and write to cache
- Benchmark speed difference between cach hits and misses
- Decide how cache invalidation should be handled and implement it.
- Benchmark again, also check what a good cache size might be.
- Distill insights into documentation and/or dummy example service that illustrates how to add caching to a LW service.