Skip to content

Load Testing Blog

  • in JMeter
  • 11 min read

Performance Testing and Artificial Intelligence (½)

Summary

AI is becoming central in software delivery, but relying on it alone for performance testing can limit accuracy and real insight.
A balanced approach is needed: AI can help, yet human expertise remains essential for defining requirements, assessing risks, and understanding real system behaviour.
This first part of a two blog post serie compares the methodology of performance testers with the output produced by ChatGPT, focusing on requirements gathering and risk assessment.
A fictional application is used to evaluate how both approaches differ in depth, relevance, and business awareness.
The analysis shows strong overlaps but also highlights where AI oversimplifies or lacks contextual judgment.
Used wisely, AI strengthens performance testing—but replacing human reasoning is not realistic or effective.

Table of Contents

AI in Performance Testing: MCP server Integration with OctoPerf

Some topics are just too trendy to overlook, and AI in testing is definitely one of them. A few weeks ago, we shared a blog post introducing the integration between an MCP server and OctoPerf, highlighting the many benefits it brings.

To illustrate this in action, we recently hosted a webinar led by Thomas Pitteman, performance testing expert at Adeo and OctoPerf power user. In this short 20-minute session, Thomas walks through the concept behind the integration, demonstrates several prompts to launch a test directly from Claude, and shows how to query the AI for insights and recommendations on the results.

The webinar wraps up with a Q&A session, which we chose to keep, so you can hear answers to questions you might be wondering about too.

Sharing variables between Virtual Users

Summary

Sharing dynamic values across Virtual Users is often required in realistic load-testing scenarios. AMQP provides a simple and reliable way to exchange data between scripts by pushing values into a queue and retrieving them later. The tutorial demonstrates how one Virtual User can publish generated IDs to RabbitMQ, while another consumes them on demand.

CloudAMQP is used as the example broker, but the method works with any AMQP-compatible system. OctoPerf integrates easily with these HTTP endpoints, allowing both insertion and retrieval of queued values. Combined with JSON extractors, this approach makes cross-VU coordination straightforward and scalable.

Table of Contents

OctoPerf Spira Integration tutorial

Summary

This guide shows** how to connect Spira and OctoPerf to streamline load-testing workflows**. The integration lets teams launch OctoPerf scenarios directly from Spira test cases and automatically receive results back in Spira. Setup involves enabling the OctoPerf app, configuring server credentials, and adding custom properties to Spira test cases.

Running a test from Spira triggers an OctoPerf execution using the referenced scenario ID. A webhook combined with OctoPerf notifications pushes the final test results into Spira for centralized reporting. The process improves collaboration by keeping performance test execution and results within the same toolset used for QA management.

Table of Contents

How to load test OpenID/OAUTH

Summary

Load-testing applications using OpenID Connect requires simulating real authentication flows and validating token handling under stress. The guide walks through the OIDC “code” flow, explaining tokens, JWT internals, and the roles of the application, user, and identity provider. A full JMeter scripting workflow is detailed—from recording traffic to extracting dynamic values like state, nonce, session codes and authorization codes.

The process includes generating tokens, decoding them, managing headers, and preparing for token refresh logic. Keycloak is used as a reference example, but the same approach applies to any OIDC-based system. Guidance is provided on load modeling, login frequency, token expiration, and designing realistic test policies.

Table of Contents