the continuity of service of a website with the establishment of an overflow page 2020 Guide -

Under confinement conditions, many websites and applications are currently overwhelmed with users and cannot respond to a rapid change in purchasing behavior, such as drives and e-markets. If you have not anticipated these situations, there is little chance that your infrastructure / application will hold up …

In that case :

google loves me workshop banner
  • Either your site becomes unusable with page loading times in tens of seconds;
  • Either the browser can no longer “join” your server and returns a 503 error.

The best solution to answer this problem: set up an overflow page. But that requires identifying precisely when the conditions are no longer met to allow your site or application to be used under good conditions. To do this, load tests must be carried out, which should not be confused with stress tests.

What is the difference between stress tests and load tests?

The idea of ​​stress tests is to drown the server under HTTP requests to simulate simultaneous connections. Like what just happened unfortunately for a part of the computer servers of the AP-HP which underwent a denial of service attack (DDos) during which, brutally overloaded with useless requests, they were drowned and made inaccessible .

By stressing the infrastructure, we obtain metrics on the hardware part: use of RAM, bandwidth, etc. but no information on the quality of the user’s real experience behind the screen.

Thus, with the stress tests you will identify whether or not your infrastructure can potentially absorb the simultaneous connection of “2000 users”, but you will not be able to judge the quality of navigation on your site for 500, 1000, 1500 users.

What is the point of leaving access to your site if, from 1000 users, the least page requires more than 10 seconds of loading time?

The importance of load testing

Only the ramp-up test campaigns make it possible to obtain this vital information in order to set up an overflow page.

To do this, it is necessary to instantiate virtual machines to reproduce the real behavior of users via connections to a site from browsers. It is the only solution that allows us to observe the navigation conditions, user after user, to precisely identify the key moment when it will be necessary to redirect users to an overflow page.

On the one hand you will have 100% of the Internet users present on your site, who navigate in conditions favorable to conversions, and on the other hand, the users who wait on a dedicated overflow page.

An overflow page to reduce the loss of potential customers

It is this page which has the role of making users wait, while space becomes available on your site. This page is generally static and minimizes the number of HTTP requests to be able to absorb connections, because you will already know that they will not be able to take place in good conditions for your Internet users, if they arrive on your site.

In this context, it becomes relevant to make a stress test to anticipate the ability of this page to absorb traffic, because we are no longer interested in the navigation conditions.

We do not check a route, but the availability of servers in the face of peak traffic.

In summary, setting up an overflow page to ensure the continuity of service for your website requires two types of tests:

  1. Load tests on your site, to identify the key moment when the conditions are no longer met for the user to be able to use your services;
  2. Stress tests, to find out the number of simultaneous users that your overflow page can accommodate.

Once all the data in your possession you will have all the elements to anticipate the next loads on your site or application and ensure optimal conditions of use for all users.

About the Author

Christian SAYEGH : Consulting and strategy in test automation at CloudNetCare, SaaS platform for industrialization of web and mobile tests, I work on projects to improve the quality of Web software and mobile applications through the industrialization of automated tests: scalability, non-regression, performance, website and apps monitoring.

Site – LinkedIn – Twitter