Building resilient applications is not easy. We expect our servers to have decent predictable latency even in the face of high loads, unexpected load/latency spikes; we want graceful degradation in face of failures. This unconference session provides a space for practitioners to discuss experiences of designing and building applications with these concerns in mind. The agenda will be decided by the participants on the day, but examples of topics relevant to the conference are:
- Queueing delays
- Sizing queues, thread pools, concurrency limits
- The effect of outliers (high percentiles) on system performance and capacity
- Load shedding
- Circuit breakers
- Cascading failures
- Isolating interdependent systems