Hello,
Im writing you guys because I have recently started a new project that aims to integrate a DB cluster solution into my company's grid computing platform, and the Sequoia middleware looks like the best option so far; however, despite the fact that both softwares are written in Java and share many features and architectural structures, were still in doubt of which would be the best way to put the two pieces of code together.
Well For now I would like to make the following question: which are the restrictions that prevent two or more Sequoia Controllers to share a backend? Im asking this because, as far as we can see now, the best way to adapt the Sequoia to our platform would be deploying multiple controllers connected to the same set of backends. This happens because, in a parallel execution, there will be multiple clients connected to the DB cluster via the controllers, and we would like to avoid the duplication of requests on the network. Let me explain this better:
- if we have a single controller, then the requests generated by the app instances would travel from the driver to the controller and from the controller to the backends, duplicating the usage of the network,
- on the other hand, if each app instance has its own controller running locally, and if all controllers share the same set of backends, then the requests would be sent to the local controller and forwarded to the backends, making the network busy only once.
Does it sound reasonable?
Another thing that we couldnt understand so far is: why does the Sequoia use group communication among the controllers if they dont manage the same set of backends? Whats to be synchronized?
F. Lucchese
_______________________________________________ Sequoia mailing list [email protected] https://forge.continuent.org/mailman/listinfo/sequoia
