General feedback to class - 03/02/2006
- Make sure that your hard coding of server names/addresses does not become a limiting factor when you go into the scalability phase.
- Do not adjust a client-side timeout at this stage as yet -- it will become onerous. Use a timeout in the real-time phase of the project. Use exceptions instead in this phase of the project to indicate server failure.
- Try to use a dual-mode fault injector (kudos to team 4 for doing this already) so that it helps you in your testing and automated fault-injection.
- Make sure that your system notifies the end-user when you are all out of replicas. You can implement a number of retries to keep looking for alive replicas, but after a while, the system should give up and notify the users.