Tuning continuously

Continuous performance tuning is a natural benefit of the implicit scalability and high observability of serverless solutions because we can safely deploy with reasonable defaults and tune as we learn from the metrics. The resource and work metrics show us how our services actually perform, so we can make informed decisions about optimizing performance and throughput. For example, in Chapter 4, Trusting Facts and Eventual Consistency, we covered various techniques and parameters for optimizing the throughput of our stream processors. We will set these parameters upfront based on educated guesses and reasonable defaults. We can perform load testing to double-check these hypotheses. But it is ultimately the real observability metrics that dictate proper tuning.

Of all the autonomous service patterns, performance is of the utmost importance for BFF services, because they are user-facing, and users have high expectations. The work metrics are the place to start...