As online businesses prepare for the holiday shopping season, it is imperative for them to understand the amount of traffic their website can handle. We spent last week’s Google+ Office Hour Hangout with Ragnar Lonn and Robin Gustafsson of Load Impact, a Cloud Tools Marketplace partner, to discuss their load testing tool and how it can help you prepare for high traffic events. Here are some highlights from the talk along with links to the part of the video that we discussed them. If you are looking for more information, I’ve embedded the video of the entire Hangout at the bottom of the post.
- Robin showed us a live demo on how you can simulate consumer behavior on your site by recording sample user behavior and saving it as a “scenario.” Using that information, the Load Impact software can use each walkthrough scenario to create scripts that can load test your site from a variety of geographical destinations.
- You also have the ability to install an agent on your servers that can send back information about how the server is performing during the load test. This agent can provide stats on things like CPU and disk utilization for each tier of your application (web, app and database).
- Load Impact has two types of users it can simulate: the Virtual User (VU) and the Simulated Browser User (SBU). The VU simulates one connection per user at a time, meaning that everything happens sequentially; this is a great option to run tests against an API. However, if you are simulating a normal website, you want to load things in parallel such as images, style sheets, et cetera. You can put your site under the load of 100,000 SBU load test users using Load Impact.
- The team went over some of the results that users get after the load tests are completed. You have the ability to graph user load time, TCP connections active, CPU and RAM utilization among others. Check out the video for more types of metrics and results you can get from Load Impact.
- Ragnar points out that you should utilize the information from Google Analytics to help simulate user behavior as you create scenarios to input in Load Impact. The benefit of load testing is that you can amplify those typical customer behaviors on your site. This gives you the advantage of understanding how actual user behavior can affect performance rather than performing a rudimentary load test by “ping flooding” your servers.
- So what can you do if your site can’t handle the expected load? Racker Drew Cox advises one of two things: (1) you can either fix your code or (2) enhance your architecture. While adding more resources can be a short-term fix to help you out in a pinch, the optimal solution is to examine your code and make it more efficient. Having clean, efficient code will enable you to have a lighter weight solution; using less compute resources can result in major financial savings.
We will be off air this Thursday as we enjoy some turkey and pie, but you can check out Rackspace Google+ page for the next Hangout invite!