Business-related metrics can provide helpful insights into the enterprise application. These metrics are a part of the business use case and therefore should be treated as such. Business metrics are ultimately impacted by other, technical metrics. It's therefore advisable to monitor these metrics as well.

The theory of constraints describes that there will be one ore more limiting constraints that prevent the system from infinitely increasing its throughput. In order to improve the application's performance the limiting constraint therefore needs to be eradicated. jPDM helps identifying the limiting constraints by finding the dominating consumer of the CPU first and using appropriate tooling to further investigate performance issues. It's advisable to investigate potential bottlenecks by following this process, which takes the overall situation into account, rather than to blindly peek and poke.

Rather than using high-frequency monitoring, engineers are advised to sample technical metrics with low frequency and to query, calculate, and investigate out-of-band. This has tremendously less impact on the application's performance. Distributed applications will need to meet SLAs. The backpressure approach as well as the bulkhead pattern can help achieve highly responsive, resilient enterprise systems.

Traditional logging should be avoided for a number of reasons, especially the negative performance impact. Enterprise applications are advised to only output log events in case of fatal, unexpected errors, which are written to standard output in a preferably straightforward way. For all other motivations, such as debugging, tracing, journaling, or monitoring, there are more suitable solutions.

Performance and stress tests running in simulated environments can be used to find obvious errors in the application. The environments should be as close to production, including all applications and databases involved. For any other reasoning, especially statements about an application's expected performance, bottlenecks, or optimizations, performance tests are not helpful and might even lead to wrong assumptions.

The next chapter will cover the topic of application security.