Skip to content

Latest commit

 

History

History
96 lines (70 loc) · 2.79 KB

README.md

File metadata and controls

96 lines (70 loc) · 2.79 KB
Raw

ntu

Network Test Utility.

Usage

Supported Protocols:

  • http/https
  • tcp
  • mysql
  • postgres/postgresql
  • mariadb
  • mssql

CLI

Requires Docker.

Options:

  • --target (-t) – URL or URI
  • --slack-webhook (-sw)– Slack Incoming Webhook URL for reporting results
  • --query (-q) – SQL query for database targets
docker run ghcr.io/polyscale/ntu-cli
docker run ghcr.io/polyscale/ntu-cli --help
docker run ghcr.io/polyscale/ntu-cli latency --target http://host.docker.internal:3000
docker run ghcr.io/polyscale/ntu-cli load --target http://host.docker.internal:3000

Worker

Requires Docker. Use environment variables to configure the worker.

TEST=load
URL=http://host.docker.internal:8080/demo

or

PROTOCOL=http
USERNAME=
PASSWORD=
HOSTNAME=host.docker.internal
PORT=8080
PATHNAME=/demo
SEARCH=param1=value&param2=value
QUERY=

docker run --env-file=.env ghcr.io/polyscale/ntu-worker
docker run --env TEST=latency --env URL=http://host.docker.internal:8080 ghcr.io/polyscale/ntu-worker

How do the tests work?

General

The tests measure different things depending on which protocol you are using for your target and for database protocols wether you provide a SQL query to execute. All tests include network latency.

For a single request to:

  • tcp://...
    • the time it takes to establish a TCP connection is measured
  • [http|https]://...
    • the time it takes to complete an HTTP request is measured
  • [mysql|mssql|mariadb|postgres]://
    • when a query is provided the time it takes to run that query is measured. A single connection is used.
    • when no query is provided the time it take to establish a database connection is measured

Latency Test

The latency test fires 100 queries in sequence and provides metrics based on the collected execution times.

  • median – median of execution times
  • P50 – 50th percentile of execution times
  • P90 – 90th percentile of execution times
  • P95 – 95th percentile of execution times
  • P99 – 99th percentile of execution times

Load Test

The load test fires an increasing amount of queries in parallel ([1, 5, 20, 50, 100, 250, 500, 1000]) in two iterations each. The test collects execution times, successful and failed request counts for each concurrency level and provides metrics based on the data.

  • successMedian – median execution time of succesful requests
  • successAmean – average execution time of succesful requests
  • successRange – range of execution times of succesful requests
  • successRequests – count of succesful requests
  • errorMedian – median execution time of failed requests
  • errorAmean – average execution time of failed requests
  • errorRange – range of execution times of failed requests
  • errorRequests – count of failed requests
  • totalRequests – count of requests overall