Together with our client we determine the different successive user actions and average think time between each action that will occur in everyday reality. After defining these, we simulate this reality with a minimum amount of simulated users on the clients’ hardware setup.
NedBox is a free website to practice Dutch through television clips and newspaper articles. This website was developed by University Leuven and industry partners, including Televic Education http://www.televic-education.com/en/. Sizing Servers Lab stress tested this case.
Example of 1 successive user action on NedBox: a student surfs to the NedBox website, selects a clip and watches the clip. We simulated this reality by letting a minimum amount of simulated students perform the successive user action on the clients hardware setup.
Next, we gradually increase the number of user actions per second and measure the effect of each increase on response times and throughput, while monitoring and correlating this with the resource-usage (CPU, disk, memory and network usage). At Sizing Servers Lab, we always monitor this with our in-house developed tool vApus. Because vApus was designed as a research tool, it’s very customizable and applicable on a multitude of desired platforms and scenarios. The result of this baseline stress test is a clear bottleneck overview of your service.
In the NedBox case, we gradually increased the number of times a simulated student surfed to the NedBox website per second, the number of times the student selected a clip per second and the number of times a student watched a clip per second. We measured the effect of these increases on response times, throughput and resource-usage.
Based on the bottleneck overview from the baseline stress test, we’re able to offer you advice on how to optimize your IT or web service and which hardware & software will work out best for your case. We test these recommended hardware-software combinations with new stress tests and offer a comparative overview of all suggested combinations.
Looking at the Nedbox baseline stress test results, we pinpointed the main NedBox bottlenecks. For this specific case we advised some modifications at infrastructure and application level. In the end report we described how to convert these advices into reality.