Purpose
The performance section is small, so it should carry the basic mental model as well as the case studies. Start with latency, throughput, and utilization. It gives the queueing and bottleneck vocabulary that the other notes assume.
Then read streaming data for a concrete bandwidth-bound workload and cache-line efficiency benchmark for a direct measurement of how access pattern changes throughput.
Notes
- Foundations: latency, throughput, and utilization, tail latency, percentiles, and queueing distributions, Amdahl, Gustafson, and scalability limits
- Measurement: profiling, tracing, and flame graphs
- Bandwidth and streaming: streaming data, cache-line efficiency benchmark
- Case study: efficiently implementing the state pattern on the JVM