Performance Engineering: Measure Before Optimizing
Use workload shape, queueing, profiling, and query evidence to locate the dominant cost before changing code.
Curriculum (9 steps)
Latency measures how long one operation takes; throughput measures how many operations complete per unit time.
An interactive concept flow for latency, throughput, and queueing are linked, from its assumptions through the main operation and boundary checks.
Profiling attributes resource use to parts of a running program; benchmarking compares defined workloads under controlled conditions.
Follow the key state changes and boundary checks involved in profile before rewriting the slow path.
Measure a Cache-Locality Change
practiceUse a repeatable matrix workload to compare loop order, profile cache behavior, and report what the measurement can support.
Database performance depends on how many rows and pages a query touches, how much data moves between operators, and whether the working set fits in memory.
Follow the key state changes and boundary checks involved in find the data access that dominates query cost.
An interactive concept flow for find the data access that dominates query cost, from its assumptions through the main operation and boundary checks.
Explain the model, execution steps, complexity, and limits of profile before rewriting the slow path.