Performance

Measurement, profiling, resource use, latency, throughput, and optimization.

Entries (6)

Bound candidate retrieval before expensive rankingLimit candidate generation before applying a richer ranking algorithm outside the storage engine.

Bounded workPlace explicit limits on batch size, concurrency, queue depth, and scan range so one operation cannot consume unbounded resources.

CachingReuse a previously computed or fetched result to avoid repeating expensive work, at the cost of managing staleness and invalidation.

Latency vs throughputDistinguish how long one operation takes from how much work a system completes over time, because optimizing one can worsen the other.

Measure before optimizingIdentify the actual bottleneck with measurement before changing code for performance, so effort targets a real, verified cost.

Performance budgets and profilingSet explicit resource limits for a system and use profiling to find where those limits are spent before changing code.