What is the performance cost of using a cloud-based comment system for your blog ?
Outsourcing blog comments to cloud services such as Disqus or Livefyre offers plenty of advantages: spam protection, analytics, already signed users and deep integration with social networks such as Twitter and Facebook.
Those features are appealing. But outsourcing your comments comes with a price: it will slow down your site. It’s been demonstrated time and time again that even an extra 500ms negatively impacts user engagement. In 2009, for instance, Google demonstrated this effect by adding artificially an extra 400ms delay which led to 0.6 percent fewer queries. Having a slow site can also negatively affect its search ranking.
The obvious question, then, is: how much does using these cloud-based commenting systems affect your site loading time? To try to find an answer, I loaded the four main cloud comment systems — Disqus, LiveFyre, Facebook and IntenseDebate – into a blank page. To make this test as real as possible, I loaded comments associated with popular posts from various popular sites such as Mashable. On the “client” side, I used the awesome service webpagetest which allows you to evaluate loading time using real browsers from various locations in the world.
I choose the following locations to get good world coverage: Moscow, Tokyo, Paris, Sydney, and Washington, D.C. For each location, I ran the test 10 times and took the median time for each service.
The first test I ran evaluated how much latency is added when your browser’s cache is empty. The results are summarized in the graph above. As you can see, for the “onload” event that indicates when the basic page is rendered but not necessarily complete, using a cloud comment system adds between 800 ms and 3.5s. Facebook is the fastest and IntenseDebate is the slowest. Looking at the result country-by-country, it appears that the Intense Debate is far slower in Europe (5.4s from Paris) than in the United States (2.07s from Washington, D.C.).
In terms of the full page rendering, adding one of these cloud commenting systems increases the latency between 2.5s and 5s. Again, Facebook is the fastest. This time, though, the others are roughly equivalent.
I designed the second batch of tests to evaluate how performance improves for returning users. Returning users, at least in theory, already have in their browser cache some data that should reduce the loading time. As visible in the graph above, these four cloud commenting systems do leverage caching to improve loading time. Disqus and Facebook make a good use of caching to make the page load well under one second. Surprisingly, however Disqus total loading time does not decrease as much as the other one. One reason might be that I chose a very active page and the content was changed between runs. Once again, IntenseDebate appears to be slower that the other contenders.
What to conclude ?
Using a cloud comment system clearly impacts loading speed in a negative way. This isn’t necessarily an indictment of these commenting systems. Rather, it means that you’re in trading off functionality for speed. In this context the clear winner, at least for me, is Disqus. It is almost as fast than Facebook, but offers more features, including the fact that it display social reactions from various networks including Twitter and Facebook. Facebook is the fastest, but I don’t think the functionality provided are worth the speed penalty. At the other hand of the spectrum, IntenseDebate is a suboptimal choice. It’s far too slow for what it provides. Interestingly, the observed speed is faithfully reflected in the page speed score and to a certain extent by the YSlow score as visible below. Don’t forget to run page speed on your site, and work hard to improve it, as speed will help increase user engagement.