Performance measurement for Acceptance and Production

After completing your implementation of SDL Web and before going live, measure the performance of the various areas of your implementation in the Acceptance and Production environments to find any performance problems.

Publishing

The publishing area tends to be the least predictable and the most demanding area of your implementation. Many factors can impact publishing performance. Refer to the troubleshooting section for more information about these factors.

In general, a publish cycle of two minutes or less and a publishing volume of roughly 500 items per hour would be reasonable performance goals to pursue.

To identify possible performance bottlenecks in your implementation, note the following:

  • In a system without any customizations, performance is typically most influenced by queuing, both on the Content Manager and Content Delivery sides. The various queues need to be fine-tuned and adjusted to each other carefully to ensure a steady publish flow, and fit the hardware set up for this purpose.
  • If you customize the Content Deployer by adding custom Modules and/or Processors, deployment may slow down noticeably, especially when these customizations interact with back-end systems.
  • If you create long or complicated templating code, rendering and resolving of simple items can slow down. (Slow rendering and resolving is not always a problem: for example, a new navigation may take 10 minutes to render, but it is a rare and high-impact change to the Web site, and so will not slow down publishing overall.)
Content Delivery

On the presentation side, performance is often easier to predict than publishing. Use your actual or expected number of visitors and Web site load as a reference point, as well as your existing infrastructure for your Web site.

Another helpful approach is to create a reference implementation, publish a Web site and measure its performance on a single server, then extrapolate to find out how many servers you expect to need.

You can experiment with having multiple servers running side by side, or with running multiple instances of your Web and application server on the same machine.

It is recommended to run your Content Deployer on a separate server. Combining it with your Web server can harm your performance and introduce risks.

Content Manager

One way to measure performance is to automate GUI interaction, repeating basic steps such as opening a Page, opening a Component, modifying and saving the Component, etc, and measuring the actual duration of such operations in milliseconds.

Another way to measure performance is to have actual users interact with the product and report back on their performance experiences.

SDL Web has various extension points, such as the user interface, workflow and the Event System and many other parts of Content Manager. Therefore, if you have built extensions, measuring the performance impact of each of those customizations is important. To optimize performance, zoom in on performance bottlenecks by switching off various customizations you have built and examining how they affect performance.

Also, your security framework may have a noticeable impact on Content Manager performance. Be sure to test not only with an administrator account, but also with a realistic end user account.