Traffic in a Storm

by Marco Di Vittorio
Manager, Application Administration Academic & Collaborative Technologies (ACT

Most traffic moves slower during a snow storm. Pedestrian, vehicle and airline routes typically grind to a halt when large white flakes flood the ground in high volumes. Web traffic on the other hand, can exhibit a very different pattern during the same extreme events. Over the course of the recent snow storms that hit Toronto, the University homepage experienced significant spikes in request rates driven primarily by visits to the campus status page. The charts below tell the story.

Top Referrer URLs:
Monday Jan 28 1:00 PM to Tuesday Jan 29 1:00 PM

Beginning on Monday, Jan 28th, as the weather worsened over the course of the afternoon, requests to the U of T homepage increased and dramatically spiked at approximately 6:00 PM. It’s reasonable to assume the source was primarily students checking to see whether their evening classes were cancelled due to the weather. The traffic soon subsided only to spike to an even higher level at 6:30 AM the following morning as the U of T community woke up and checked the website to find out if campuses were open for business.

Total HTTP Requests per 5-minute interval:
Monday Jan 28 1:00 PM to Tuesday Jan 29 1:00 PM

High traffic alarms were sounding but the hybrid on-prem and cloud infrastructure responded and the website continued to smoothly serve content during the increased periods of demand. Roughly 200 GB of data was served over the 24-hour period. Site uptime was 100% and the average page response time was 431ms (as measured by

A third event occurred on Wednesday, Feb 6th beginning at 1:40 PM as the campus status page was updated with information that UTM, UTSC and STG were closing once again due to poor weather. As demonstrated in the graphs below, by 1:50 PM total request volume over a 5-minute interval peaked at almost 70,000 and then sharply declined. At the high-water mark, request volume was approximately 20x greater than the 6-month average measure. This data might also serve as a rough indicator that it took approximately 10 minutes for the closure message to disseminate across a significant portion of the university community.

Total HTTP Requests per 5-minute interval:
13.7K requests on Feb 6 at 1:40 PM

Total HTTP Requests per 5-minute interval:
68.0K requests on Feb 6 at 1:50 PM

While these metrics reinforce the benefits of the current infrastructure design, the most important outcome from our perspective is that we were able to reliably and rapidly deliver timely information to the institution’s students, faculty and staff.