Because getting things done quicker makes life better
Having a common dataset that meets the expectations of both Ops and Performance Engineers is the basis for good collaboration. Tests that use loads similar to production loads provide a common layer to build upon. The quality of performance test results relies on the quality of the loads used in performance tests.
While performance engineers work with development teams to improve the performance of the applications, they work with Ops to analyze the server infrastructure and platform sizing and tuning for performance.
Test results are only as good as the load profiles used.
Load patterns in performance tests can be quite different from the patterns of app usage in production, which ultimately affects the accuracy of the test results. Comparing load profiles is as challenging as it is time-consuming.
The app being tested in Performance & Load (P&L) testing or Pre-Prod stages is always newer than that found in production environments. The newer version of the app will not have the same URLs as the app in production making it hard to generate test loads that match production.
The load or request rates in test environments are typically lower than those received by production servers. Adding to that, the server sizing and infrastructure in test environments are smaller than production infrastructure. If the load intensity is not proportional to the scaled down test infrastructure, the insights into infrastructure performance will be skewed
The duration of load profile patterns in test loads may not be similar to that experienced by apps in production environments. The order of the load patterns in the test loads may not be similar, which affects the behavior of the app as well the infrastructure performance. The duration of load profiles need to be proportional to that experienced by apps on production servers.
Machine learning at your service
Sign up to be invited when a spot opens up
System load has patterns but detecting and quantifying them is challenging due to their changing nature. Our machine learning algorithms detect patterns and quantify them statistically.
Visually identifying similar load patterns is not practical as patterns as are never exact. Our algorithms match production load patterns with patterns in test load so that you can visualize similarities without having to guess.
The loads on production servers can be high, and the patterns can last for a long duration. Because server size and infrastructure in the P&L stage is often smaller than production, you’ll greatly benefit from our algorithms’ ability to compare the ratios between the test load and production load. Now you can scale a test on more modest infrastructure and easily compare it to production load.
Patterns found in Reference Load and Compare Load are listed in the order they were observed. Color-coded patterns make it even easier to compare. Typically, production dataset is the Reference Load against which the test dataset (Compare Load) is compared.
Load and time ratios for each pattern make it easy to compare the scale of your test load to the production load.
The order of pattern change affects system infrastructure performance as the system need to respond to every change. Automatic pattern order detection reveals and matches pattern orders.
Sign up to be invited when a spot opens up