To optimize database connectivity, an application should store and retrieve data from the database as quickly as it can. It should optimize the use of memory, CPU cycles, and network bandwidth. And with all optimizations, there will be tradeoffs. Good software will be tunable, so the scarcest resources in the system can be preserved while still providing acceptable performance.

Database connectivity may seem like a small issue in the context of a large enterprise solution. But in a very real sense, the database is nothing to the client but the database connectivity software used to connect to it. In the .NET world of managed code, connectivity software that is 100% managed can be safer, easier to maintain, and more performant than software that uses a mixed .NET/COM model. That software should efficiently use the memory, CPU cycles, and network bandwidth. It should be tuned to the workload that it is presented, and provide facility for monitoring and troubleshooting its own performance. It should also provide scalability features to allow it to work in the load balanced, always available world of mission-critical database applications.