Load testing
Load testing evaluates a system’s behavior, performance, and capacity under specified levels or patterns of workload.
ScyllaDB in Action
Load testing is the technique of applying representative or increasing workload to a system to evaluate its performance, behavior, and capacity limits.
- How it is used
- Determines cluster traffic capacity for sizing and makes monitoring dashboards display realistic load.
- How deep it goes
- Walks through schema-based YAML setup, workload tuning, ramp-up execution, and throughput and latency result interpretation.
- Examples
- Uses an article_summaries_by_date YAML profile, a 1,000-to-1 read/write run, result tables, and a Grafana overview dashboard.
- What you can do after
- Operate/configure cassandra-stress against a schema and evaluate throughput and latency while increasing database load.
Think Like a Software Engineering Manager
Load testing is software performance testing that evaluates behavior under specified levels of concurrent or simulated workload.
- How it is used
- Checks capacity and failure points under normal and anticipated peak traffic as part of software testing before release.
- How deep it goes
- Defines simulated customer load and its diagnostic aims, but gives no procedure for generating traffic or setting thresholds.
- Examples
- Uses sample-customer traffic and a million-customer video-download scenario to distinguish capacity questions from performance measurements.
- What you can do after
- Evaluate/understand whether expected peak traffic exposes capacity limits, bottlenecks, or single points of failure.
Read how Think Like a Software Engineering Manager explains load testing →
Amazon Web Services in Action, Third Edition: An in-depth guide to AWS
A load test is a controlled test that generates workload to evaluate system behavior under specified demand.
- How it is used
- Evaluates how synchronous web traffic or asynchronous queue work drives AWS resources and scaling policies.
- How deep it goes
- Demonstrates workload generation, command parameters, alarms, and scaling observation, without developing broader benchmarking methodology.
- Examples
- Uses a 500,000-request WordPress test and a 250-message URL-to-image queue test.
- What you can do after
- Apply it via Apache Bench and AWS consoles to trigger, monitor, and clean up scaling experiments.
LLMs in Production
Load testing is a performance-testing technique that evaluates how a service or system behaves under an expected or increasing workload.
- How it is used
- Evaluates whether a service can sustain expected or peak workload before production release.
- How deep it goes
- Defines a staging-to-analysis workflow and measures latency, throughput, TTFT, and TPS under increasing demand.
- Examples
- Demonstrates a Python Locust script, CSV metric logging, and a 50-user run that reveals a requests-per-second bottleneck.
- What you can do after
- Diagnose capacity limits and estimate replicas using latency, throughput, TTFT, and TPS measurements.
Build Financial Software with Generative AI (From Scratch)
Load testing is nonfunctional testing that evaluates an application's behavior and performance under simulated user or request loads.
- How it is used
- Uses simulated API traffic to assess responsiveness, bottlenecks, and capacity during development and deployment planning.
- How deep it goes
- Walks through Locust setup, endpoint scripting, dashboard configuration, randomized inputs, and assertions, but leaves richer API data for later.
- Examples
- Exercises a FastAPI file API through Locust tasks, randomized path parameters, response assertions, dashboard runs, and server logs.
- What you can do after
- Operate Locust to script API workloads, run simulated users, inspect metrics, and diagnose failing status codes or routing behavior.
Read how Build Financial Software with Generative AI (From Scratch) explains load testing →
Related concepts
- testkind of · A test is a procedure or check that evaluates behavior using specified inputs or evidence against expected results or criteria.
- Locustuses · Locust is an open-source load-testing tool for defining simulated user behavior and generating configurable concurrent traffic to measure software-system performance.
- performance testingkind of · Performance testing is the evaluation of a system's behavior, such as its responsiveness, throughput, scalability, stability, or resource use, under specified workloads and operating conditions.
- cassandra-stressuses · cassandra-stress is a command-line benchmarking utility that generates database workloads for measuring Cassandra-compatible systems.
- definition of donepart of · A definition of done is an agreed set of criteria that work must satisfy to be considered complete.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.