Real-time sentiment built on the Azure IoT Stack
Learn how Microsoft Pulse, a cloud-based sentiment collection platform, leveraged the Azure stack to enable hundreds of thousands of users to provide real-time feedback and see real-time results during major broadcast events like State of the Union and the 2016 Election Debates. Talking points include Quick Pulse Voting UI/Walkthrough, with an explanation of the product, why scalability and real-time responses are important. We describe the technical problem – the need to scale to accept data from hundreds of thousands of concurrent users engaging (pulsing) every five seconds, quickly tally the data by demographics, and report back results so that users can see the sentiment response to what is being talked about. We also discuss actual numbers (peak votes/time series, total round trip time, concurrent users). We talk about the pipeline of the pulse round-trip (i.e., user clicks the button, request goes to WebAPI front-end, event hubs, picked up by the aggregator services, tabulated, cached in Redis and reported back down to the client via SignalR). Finally, we cover other architectures that were considered, and discuss what worked, what could have worked better, and what might work.