Performance & Scalability

Core: start here

• N+1 Query ProblemOne query for a list, then one more query for every item in it.

• Percentiles (p50, p95, p99)The value below which a given share of measurements fall; the right way to read latency.
• AutoscalingAdding and removing capacity automatically as load changes.
• BatchingGrouping work to reduce per-item overhead.
• Cold StartThe delay when a serverless function or container starts fresh.
• BenchmarkingMeasuring performance under controlled conditions.
• PerformanceHow fast and efficiently a system does its work.
• Performance TestingMeasuring speed and resource use; load, stress, soak and spike tests.
• BottleneckThe component that limits overall throughput.
• Moving Work Off the Request PathDoing slow work in the background to keep responses fast.
• Payload SizeSending fewer bytes over the wire.
• ProfilerA tool that measures where time or memory goes.
• Load TestingMeasuring behavior under expected traffic.

Core: start here

• Capacity PlanningPredicting and providing the resources needed for future load.
• Tail LatencyThe slowest requests (p99), which users notice most.

• Performance vs CostDeciding how much speed is worth paying for.

• Universal Scalability LawWhy adding nodes eventually makes a system slower.

• Percentiles (p50, p95, p99)The value below which a given share of measurements fall; the right way to read latency.

Core: start here

• Percentiles (p50, p95, p99)The value below which a given share of measurements fall; the right way to read latency.

• AutoscalingAdding and removing capacity automatically as load changes.
• BatchingGrouping work to reduce per-item overhead.
• Cold StartThe delay when a serverless function or container starts fresh.
• BenchmarkingMeasuring performance under controlled conditions.
• PerformanceHow fast and efficiently a system does its work.
• Performance TestingMeasuring speed and resource use; load, stress, soak and spike tests.
• BottleneckThe component that limits overall throughput.
• Moving Work Off the Request PathDoing slow work in the background to keep responses fast.
• Payload SizeSending fewer bytes over the wire.
• ProfilerA tool that measures where time or memory goes.
• Load TestingMeasuring behavior under expected traffic.

• Capacity PlanningPredicting and providing the resources needed for future load.
• Tail LatencyThe slowest requests (p99), which users notice most.
• Continuous ProfilingProfiling live services to find hot paths.
• PrecomputationComputing results ahead of time.
• Flame GraphA visualization of where a program spends its time.

• Performance vs CostDeciding how much speed is worth paying for.

• Universal Scalability LawWhy adding nodes eventually makes a system slower.

• LatencyThe time a single operation takes.
• N+1 Query ProblemOne query for a list, then one more query for every item in it.

• Percentiles (p50, p95, p99)The value below which a given share of measurements fall; the right way to read latency.
• BatchingGrouping work to reduce per-item overhead.
• Cold StartThe delay when a serverless function or container starts fresh.
• BenchmarkingMeasuring performance under controlled conditions.
• PerformanceHow fast and efficiently a system does its work.
• Performance TestingMeasuring speed and resource use; load, stress, soak and spike tests.
• BottleneckThe component that limits overall throughput.
• Payload SizeSending fewer bytes over the wire.
• ProfilerA tool that measures where time or memory goes.

• Load TestingMeasuring behavior under expected traffic.
• Tail LatencyThe slowest requests (p99), which users notice most.
• Flame GraphA visualization of where a program spends its time.

• Performance vs CostDeciding how much speed is worth paying for.