ab

Additional Notes

ab stands for ApacheBench. It sends many HTTP requests to a URL and reports timing, throughput, failed requests, and concurrency behavior. It is useful for quick baseline checks of a web server, reverse proxy, API endpoint, or local development service.

Despite the name, ab can test any HTTP server, not only Apache. It is a simple benchmark tool, not a full load-testing platform. Use it for quick measurements and comparisons, then use stronger tools when you need realistic user journeys, multiple endpoints, login flows, or distributed load.

Installation

On Debian and Ubuntu systems, ab is usually provided by apache2-utils:

sudo apt-get install apache2-utils

On RHEL, CentOS, Fedora, Rocky Linux, or AlmaLinux systems, it is commonly provided by httpd-tools:

sudo dnf install httpd-tools

Syntax

ab [options] http://host[:port]/path
ab [options] https://host[:port]/path

Parameters

• url: Full HTTP or HTTPS URL to benchmark. Include the scheme and path, such as http://127.0.0.1:8080/.
• requests: Total number of requests to send, set with -n.
• concurrency: Number of simultaneous requests, set with -c.
• options: Headers, POST data, timeout, output, and authentication settings.

Common Options

• -n requests: Total number of requests to perform.
• -c concurrency: Number of requests to run at the same time.
• -t seconds: Run for a time limit instead of stopping only by request count.
• -k: Enable HTTP keep-alive so multiple requests can reuse connections.
• -H "Header: value": Add a custom HTTP header.
• -C name=value: Add a cookie.
• -A user:password: Send HTTP Basic authentication credentials.
• -p file: Send a POST body from a file.
• -T content-type: Set the content type for POST data.
• -i: Use HEAD requests instead of GET.
• -s seconds: Set socket timeout.
• -q: Quiet progress output during large runs.
• -e file.csv: Write percentile data to a CSV file.
• -g file.tsv: Write data suitable for gnuplot or spreadsheet tools.
• -v level: Increase verbosity for debugging responses.
• -V: Show version information.

Examples

ab -n 500 -c 10 http://127.0.0.1:8080/

Send 500 total requests with 10 concurrent requests.

ab -n 1000 -c 50 -k https://example.com/

Benchmark with keep-alive enabled.

ab -n 200 -c 20 -H "Authorization: Bearer TOKEN" https://api.example.com/status

Add a custom authorization header.

ab -n 300 -c 15 -p payload.json -T application/json http://127.0.0.1:3000/api/items

Send POST requests using a JSON payload file.

ab -t 30 -c 25 http://127.0.0.1/

Run the benchmark for about 30 seconds.

ab -n 1000 -c 25 -e percentiles.csv http://127.0.0.1/

Save percentile timing data to a CSV file.

Reading Output

• Requests per second: Approximate throughput for this run.
• Time per request: Average request time. ab prints both per-request and concurrency-adjusted forms.
• Transfer rate: Approximate response data throughput.
• Failed requests: Requests that did not complete as expected.
• Non-2xx responses: Responses outside the 2xx range. These may be normal if the endpoint redirects or denies access.
• Percentage of the requests served within: Latency percentile table.

Practical Notes

• Always include a trailing / when benchmarking a site root, such as http://localhost/.
• Start with small -n and -c values, then increase gradually.
• Do not benchmark public or third-party systems without permission.
• ab sends repeated requests to one URL. It does not model real browser behavior, assets, sessions, or user workflows.
• High concurrency can exhaust local client limits before the server is actually saturated.
• Use -k only when you want to measure keep-alive behavior.
• For production-grade testing, consider tools such as wrk, hey, k6, or JMeter.