Performance Design Principles - Business Layer Considerations
- J.D. Meier, Srinath Vasireddy, Ashish Babbar, Rico Mariani, and Alex Mackman
Instrument your code up front.
Instrument your application to gather custom health and performance data that helps you track whether your performance objectives are being met. Instrumentation can also provide additional information about the resource utilization associated with your application's critical and frequently performed operations.
Design your instrumentation so that it can be enabled and disabled through configuration file settings. By doing so, you can minimize overhead by enabling only the most relevant counters when you need to monitor them.
Prefer a stateless design.
By following a stateless design approach for your business logic, you help minimize resource utilization in your business layer and you ensure that your business objects do not hold onto shared resources across calls. This helps reduce resource contention and increase performance. Stateless objects also make it easier for you to ensure that you do not introduce server affinity, which restricts your scale-out options.
Ideally, with a stateless design, the lifetime of your business objects is tied to the lifetime of a single request. If you use singleton objects, you should store state outside of the object in a resource manager, such as a SQL Server database, and rehydrate the object with state before servicing each request.
Note that a stateless design may not be a requirement if you need to operate only on a single server. In this case, stateful components can actually help improve performance by removing the overhead of storing state outside of components or having the clients send the necessary state for servicing the request.
Partition your logic.
Avoid interspersing your business logic with your presentation logic or data access logic. Doing so significantly reduces the maintainability of your application and introduces versioning issues. Interspersed logic often results in a chatty, tightly coupled system that is difficult to optimize and tune in parts.
It is essential for scalability that you free limited and shared resources, such as database connections, as soon as you are finished with them. You must also ensure that this occurs even if exceptions are generated.
