Last night, at the peak of rush hour, my husband and I decided to go to one of our favorite Mexican restaurants here in Raleigh, N.C. This was not an easy decision because the restaurant is located on Glenwood Avenue, which at 6:00 pm is a traffic nightmare. We opened the Waze app and it confirmed what we had expected: miles of ominous red lines. Nevertheless, sometimes chips and queso (and a good margarita) is worth the effort! But as I drove onto Glenwood, I was really surprised. The road was not congested at all and, humorously, the Waze app actually warned me that I was exceeding the speed limit! But at the same time, the Waze app still showed the red lines of heavy bumper-to-bumper traffic.
Analytical limitations leave Waze users on edge
This led me to think how one application could find itself in such a data conflict. I could see that the edge analytics that captured my speed in real time was accurate, but when you consider the amount of this specific data that was being sent to the Waze app, I couldn’t help but try to diagnose the problem. I thought that perhaps some of the information was only within the app itself and other data was flowing into the cloud-based servers. But my research on Wazeopedia (did you know about that?) made it clear:
“Other Wazers shown on the map in the app are delayed by between two to five minutes. This is both for security concerns, server capacity and network data capacity. It would be a huge amount of data required to update every Wazer’s location every two seconds.“
In other words, the analytics platform that integrates a massive amount of vehicle speed data, as well as all user-reported, satellite and traffic data, while providing an integrated view in real time could not handle the performance and scale of real-time analytics. Given that Waze is owned by Google, I’m sure that there is a corporate protocol to use internal technologies, which might explain the performance and scale limitations.
This is such a real-world example of the volume of data that surrounds us every millisecond of every day, and how some applications and enterprises struggle to handle it. This is especially true in IoT use cases where some data is initially analyzed at the edge but then consolidated with a variety of other data sources in the cloud and enterprise data centers. This next step means the central analytics platform must not require any compromises in performance or limitations in scale. The Waze app clearly is facing these issues.
Delivering 15 million rides per day, Uber requires extreme analytics
But, this is not the case for all IoT analytics uses cases. Consider the ride-sharing company that we all know and use. After 10 years of unrivalled growth and the unimaginable volume of data associated with that growth, Uber is currently active in 65 countries and 600 cities worldwide, managing the supply and demand of 15 million rides every day. Fifteen million rides means 15 million active app users, 15 million requests for rides, 15 million evaluations and assignments based on the most efficient time to arrival, and 15 million determinations of pricing based on both time and demand. So, how does Uber manage this amount of data and still deliver an average ETA of under seven minutes for more than 90% of its customers? One answer to that question — and let’s be fair, there are other elements as well — is that the analytics platform that Uber relies on for geospatial, pattern matching and time series analytics can handle the volume of data in the near-real-time subsecond time required.
Managing billions of events and recommendations: AdTech runs on no-compromise analytics
But Uber is not alone in achieving this level of performance on this magnitude of data. Consider the AdTech industry and companies like Criteo, Taboola, Cardlytics and Simpl.fi. Criteo logs 50 billion total events per day and makes 15 million predictions every second. Taboola serves up 450 billion recommendations of articles, blogs, videos, products and apps to more than 1.5 billion unique users every month. These are examples of companies that don’t allow technology which forces limitations for their customers.
All of these disruptive data-driven organizations are drawing value from the spectrum of solutions available to them across the IoT ecosystem.
All IoT Agenda network contributors are responsible for the content and accuracy of their posts. Opinions are of the writers and do not necessarily convey the thoughts of IoT Agenda.