Network performance versus the storm

September 2017 saw several devastating hurricanes in the Caribbean. Our team in the Boca Raton, Florida office watched anxiously as Harvey slid by, Irma passed overhead, and then Maria swirled just offshore. As you can guess, our team paid close attention by following the weather on TV and online. Frustration mounted when the team had to wait for weather news pages to load, so we thought it was time we took a closer look at weather-related websites to see exactly how they compare.

We set up 12 Full page check screens to load a real browser every 20 minutes and check the performance of each of 12 different weather-related websites from all thirty-nine US checkpoints. Full Page Check reports on each element’s time to complete:

• Solve
• TCP Connect
• Http’s handshake
• Send
• Wait
• Receive
• Timeout

The report also provides average and total time for each of the above as well as total download time for the entire page. Let’s take a look at the results.

How well did the weather pages perform?

We looked at the sites for a total of 65 days to see if weather events affected performance. Although six hurricanes occurred during the observation period, we did not see any patterns that might have coincided with heavy on-site traffic due to weather-related events. As you will notice in the graph below for the top and bottom five pages, none of the patterns indicate any significant changes in performance due to a weather news event.

Performance over time

As we suspected from our own experience, there was no real competition for the best sites. As the pink and light blue lines indicate in the chart above, WeatherBug and Weather Spark maintained the highest performance throughout the test interval. Looking at the week from 20 to 27 September when there were no significant weather events, the performance hierarchy becomes more obvious.

The graph below shows the daily average performance of the top five and bottom five sites. WeatherBug holds up pretty well in the two-second range, Be Sparkwhich did equally well during the first few weeks of testing, suffered a performance loss in late August. Weather Sparks performance dropped by another three seconds at an average of five seconds due to the number of requests rising from nine to 39.

Performance 20.-27. September

The chart below shows the average and median times for all the sites tested.

according to Alexa’s ratingsthe three most popular sites for weather news all finished in our bottom five. With load times higher than 14 seconds, these sites more than exceed the expected two-second rule. If you’ve ever experienced these pages yourself, you may have found them just as frustrating. So let’s take a closer look at these pages and see what these sites can do to improve the user experience.

Download speeds of course affect page load times, and the more data to download, the longer it will take. When we map time against download size, although you can see a correlation, download size is certainly not the primary factor affecting page performance.

It seems like websites keep trying to cram more and more stuff into a single page. Based on our findings with the weather stations, we see a clear connection between the number of elements and performance. In the request versus time chart below, you can see a clear correlation between the number of elements and page performance.

When we take a deeper dive into the connection times, we see that the slowest sites have significantly longer average latency than the other sites, and the slower sites take longer to handshake. The long waiting and handshake times multiplied by the high number of requests contribute to the poor performance.

Improving website performance

The path to a better performing website is to first know your current performance levels. You can’t get those numbers behind your firewall, and you can’t get that information from a single test. Understanding your performance means detailed controls from your users’ location. Uptrends has both synthetic and real user options for you.

Synthetic monitoring

Synthetic monitoring is a proactive approach that actively engages with your site and alerts you to page and performance issues. Uptrends Full Page check uses a real browser to connect to your website and load your page and track the performance of the entire page and each element on the page. Your results for each check are displayed in a handy waterfall report where you can see the entire page load progression. Using Uptrends’ 168 worldwide checkpoints, you can select test locations based on the geographic location of the site’s users.

Real user monitoring

While synthetic monitoring is an active approach, Real user monitoring is a passive monitoring method that waits for user interaction to track page performance. Real User Monitoring records the performance of your pages based on the experience of actual site visitors. A small script file collects the performance data and sends it to the cloud. Uptrends aggregates the data and reports on performance based on device, operating system, location, browser and pages viewed.

Once you have information about your site’s performance, you can start analyzing your site based on the connection and the elements. Using the example of weather websites, the biggest contributing factor to poor performance is the number of requests. Reducing the number of HTML elements, combining CSS and script files, and caching images and other page elements can drastically reduce page load times.

Solve any waiting times due to location. True user monitoring will tell you exactly where your users are located. By using CDNs (Content Delivery Networks), you can reduce the distance between your content and users. By reducing the distance, you improve waiting times.

Optimize images and content based on user environments. Why deliver desktop-sized image files to mobile users? Knowing the users’ environments can facilitate content planning and guide synthetic testing.

Conclusion

The reliability of weather forecasts is often erratic at best, but weather websites don’t have to be. With proper monitoring and implementation of a few performance-enhancing changes, weather websites can build their reputation and foster brand loyalty. Show your users you care by improving their experience. Start with a free 30-day Uptrends trial or schedule a demo today.