To improve pipeline speed and manage costs, you must first understand how your resources are being used. This is done by analyzing pipeline throughput and agent utilization metrics. These measurements show you how many jobs are completing over time and whether your build agents are sitting idle or are overloaded. By reviewing this data, you can identify the best level of parallelization—running multiple jobs at once—without wasting money on unused capacity or creating bottlenecks.
You can control how many jobs run simultaneously by adjusting concurrency settings, either in YAML pipeline files or through the Azure DevOps web interface. This involves setting limits for the minimum and maximum concurrent runs for a workflow. When you enable these settings, you allow more instances of a workflow to run in parallel, which can significantly speed up processing. However, increasing concurrency can sometimes lower other system limits, such as the SplitOn property, so you must consider these trade-offs.
To balance performance needs with budget, you can dynamically adjust the number of available build agents. Using auto-scaling allows the system to automatically add or remove agents based on the current workload, ensuring you pay only for what you use. For self-hosted agents, you can scale out by adding more machines to the agent pool to spread the load. In cloud-hosted scenarios, like with an Azure integration runtime, you can scale up by allocating more Data Integration Units (DIUs) to increase the power of a single agent. These adjustments ensure pipelines run quickly when demand is high while reducing costs during quiet periods.
Seeking the thrill of transformative tech? Explore the art of designing and implementing DevOps solutions on Azure. Master the shift towards CI/CD, testing, and delivery, while preparing for the Designing and Implementing Microsoft DevOps Solutions exam!
Prepare and test your skills
Prepare and test your skills
Pipeline throughput and agent utilization metrics should be analyzed to understand resource usage. These measurements show how many jobs are completed over time and identify whether build agents are overloaded or sitting idle, allowing you to set parallelization levels that prevent bottlenecks and avoid paying for unused capacity.
Concurrency limits are configured in YAML pipeline files or through the Azure DevOps web interface by defining minimum and maximum concurrent runs. Increasing these settings allows more workflow instances to run in parallel to speed up processing, but it can also reduce other system limits such as the SplitOn property.
Scaling out involves adding more physical or virtual machines to an agent pool to distribute the workload across self-hosted agents. Scaling up applies to cloud-hosted scenarios, such as an Azure integration runtime, where you allocate additional Data Integration Units (DIUs) to boost the processing power of a single agent.
StudyCerts
StudyCerts