Sometimes Load Testing is just beautiful

Date: 12th September 2013
Author: Deri Jones

Can there be beauty in a graph? It depends on the context, of course, but the way our client’s face lit up at the real-time information they were seeing was a beautiful sight.
And the distillation of so much information into an easily understandable format was to me a thing of beauty in itself.

The Challenge

We’re working on a really interesting eCommerce project for a retailer’s online store. Our
challenge was to help them understand just how resilient their website is and its capacity to
handle a large number of site visitors. This is where the graphs came in.

 Click the image for full size.

We were able to show them in real time the sort of critical information they need to make informed decisions while the testing was on going – information that would have taken days to manually produce without our technological nous. They could also see exactly what the online store’s capacity looked like through using realistic load patterns mapped across an extensive web estate consisting of multiple URLs.

In a normal Load Test, the main number the eCommerce team need, is the throughput in terms of ‘Things the User can Do’ per hour:  most criticially of course, ‘Orders per Hour’, or ‘Itens added to Basket per Hour’.

But due to the complex mix of technolgy systems behind the scenes, of this projec the client had in addition  clearly defined ‘page types’ and how they should be grouped when measuring throughput. What made the process more challenging, however, was the complexity of page types included in any given user journey. And we had to break down our load test findings and group the traffic against a variety of major system functional blocks.

We created multiple user journeys to plot against the client’s main matrix. And to generate the extra Page_Type breakdown, it entailed building a list of URLs visited during the test and grouping pages, regardless of the URL, category or sub-category, into specific Page_Type categories. While this was straightforward for most page types, some pages within the user journey had to be identified using other data, such as the previous page visited, to determine its grouping. So, all in all, a complex challenge, but a necessary one if we were to provide the client with all of the information they needed.

Throughput - completed User Journeys










Note: Complex page-breakdown matrix
Click the image for full size.

The outcome

It was when we presented the team with a simplified matrix that clearly illustrated how multiple user journeys could impact the client’s eCommerce functionality that the smiles broke out. In all, we replicated 9 user journeys, across 150 pages based on the web analytics data from the last peak traffic period for a truly accurate reproduction of potential traffic. We got the feeling they were happy with the knowledge we’d provided.

So a graph, like a satisfied customer, can be a thing of beauty.