Multiple clusters for the same application, memory backend, and RPC layer

I am investigating how to design an application as a monolith code base, with roles instead of services (or microservices), and in particular replace memcached and REDIS with FDB using the memory backend.

Since createCluster is deprecated, it is not possible anymore to connect to two different FDB clusters from the same POSIX processus. One way to workaround that, is to have a separate processus, that connects to the other cluster, hence it requires coordination between the different process, which is more difficult, hence error prone.

Use case: I want my application to work as a single process in dev setup, and offer an easy path forward web-scale. So that there is really 1 + 1 + 0.5 infrastructure knowledge and effort required to develop and operate the whole application. Here is how I came up with the previous computation:

  • One unit of effort and knowledge invested into the monolith with roles code base
  • One unit of effort and knowledge invested into FDB using SSD backend
  • Half a unit of effort and knowledge invested into FDB using memory backend to take into account specifics about the memory backend (if any!)

I gave a some context to help decide whether it is interesting to be able to access different clusters from the same processus (I think it does). There may be other use-cases, I am not aware of (chime in!).

It seems the RPC Layer is the future.

  • Can it be confirmed that the RPC Layer will be another POSIX processus that should be deployed along side the FDB cluster?

  • What is the plan regarding developer setup? NB: I prefer to avoid the need for an FDB or RPC Layer emulator.

  • Will the future setup with the RPC layer support access from a single process multiple clusters?

You can use FDB.open() multiple times in the same process to connect to different clusters.