Less Time Load Testing, More Time Fixing
Now that you have clear expectations and more tangible goals to base your load tests on, the next step is to build your load tests. LoadComplete makes this astoundingly simple with a record button that launches your browser so you can walk through your website as you normally would. Once you’re done, the requests your browser made to your server are used to create a Scenario. In essence, you can record scenarios for each of the key critical workflows through your site, the exact same way that visitors do.
With one or two scenarios recorded, you can then focus on building out a load test that simulates the volume of traffic you’re targeting in your load testing strategy. This includes settings like test length, how many concurrent virtual users to simulate, which scenarios will be used, and other simulation options.
If you don’t know what the performance behavior of your site is already, or if you are unsure at what volume your performance issues begin to elicit, use a step-wise profile to ramp up into the load. This lets you see where your server response times begin to get worse, and has the added benefit of letting your web infrastructure ‘warm up’ to the test to minimize initial ‘red herring’ results. If you need to simulate spikes or drops throughout the test, much like when a media ad drives unanticipated traffic to your site, you can use the ‘custom profile’ option in LoadComplete to define whatever volume profile you see in your production analytics.
Likewise, you may want to simulate multiple patterns of traffic at the same time, say a read-only search process and a data-entry submission process. LoadComplete makes this incredibly easy, just by adding an additional virtual user group and selecting additional scenarios to test.
If It’s Broken, Fix It, Rinse and Repeat
Upon execution of the test, your results indicate one of two things:
- Your website meets/fails your performance goals
- Your implementation is problematic and needs tweaking
The amount of resources required to actually simulate your volume varies based a number of variables, like hardware and network limits, complexity of the Scenarios in your load test, and how you have distributed your test. By keeping an eye on any ‘dramatic plateaus’ in metrics like Passed Requests, Request Transfer Speed, and Completion Time, you can make sure that you’re not maxing out any simulation hardware (your workstation or any remote agents).
Ruling out issues in your simulation, now you can look to Page Load Time and Scenario Completion Time confidently, and include additional metrics from your servers to obtain more context on where your servers may be bottlenecked. LoadComplete can pull in metrics from any *NIX flavor operating system, Apache web server, and any Windows machine on the market today.For operating systems, CPU and memory are critical dependencies to serving large volumes of web requests, and even if you find yourself sitting with your operations team to monitor ancillary network devices, using LoadComplete to automate your load test with key metrics gives you the power and visibility to find the cause of your issues quickly.
Once you find code or configurations that can be tweaked, you want to make sure that the changes you've made have a positive impact on the original scenario you were testing. Using LoadComplete’s comparison report after re-running a load test, you can see the positive impact on web performance that your changes made, as well as any regression performance problems that it may have introduced.
Try a free trial of LoadComplete to make sure your website reaches its peak performance.
Check out the other posts in this 5 part series!
This Holiday Season Help Your Website Achieve Peak Performance – Part 1
This Holiday Season Help Your Website Achieve Peak Performance – Part 2
This Holiday Season Help Your Website Achieve Peak Performance – Part 3
This Holiday Season Help Your Website Achieve Peak Performance – Part 5