How to Fix WiFi with Crowdsourced Monitoring

A student calls their university IT helpdesk. "My internet is slow.” Here are the steps to troubleshoot.

1: Send someone to the room to test the wifi after the lecture is already over and everyone has cleared out.
2: Give up.

There has to be a better way. DCU and HEAnet are working with GÉANT to find it.
Normally, when we think about measuring performance, we think of network probes. Excellent work is being done with cheap, cost effective probes which give you very reliable information on the network as they see it. But we're also trying something else. Wouldn't it be great if you could troubleshoot rooms that didn't have a probe in place? And wouldn't it be REALLY great if you could find out, at a glance, if the problem is really on the user's laptop?

So in addition to looking at gathering reliable data from dedicated probes, we're also seeing what we can do with data crowdsourced from users' machines. To do this, we'd need to run frequent, noninvasive tests inside the user's browser, and coordinate the results with the access point they are connected to. But most of the technology we need to do this is already in place. We can pull information from wifi controllers and RADIUS servers for Eduroam users, and run simple tests in
Javascript on frequently-visited webpages, in order to build up a picture of the average quality of the network in a given location at any given time.

Then, when the helpdesk needs to answer the question "How is the wifi performing in lecture room 1", they can look at the average performance over time in that room and see if there is a deviation.

But not just that. Since the reports come from live users over the course of the day, it becomes possible to see if there are recurring problems at particular times. And we believe that the same information can be used to help with capacity planning.

Best of all, when the user arrives with their complaint, the helpdesk can see how their device performs compared to the average - and if they notice a problem, then instead of sending someone out to test a room that’s working fine, they might instead be able to ask “is your virus checker up to date?”

In this talk we’ll discuss the method we’re using, its success in the various iterations so far, and how it can be applied at other HEAnet clients - with an interesting side discussion about how to take an agile, iterative approach to developing something brand new in an environment where both operations and funding very often lead to a conservative approach.

All talks:

If you require information about any aspect of the conference please contact: Fi Coyle: / +353 1 6609040