 Hello everyone, my name is Andrew Stoikis and with my partner Aditya, we make up the spring 2020 Office of the CTO's co-op group. Today, we're going to be doing a demo on edge video streaming and real-time analytics Why is this important? Well, there's a bunch of use cases specifically we wanted to highlight the ones involving a smart city where we can make decisions based on a Video camera at a traffic light such as oh someone ran a red light. We need to give them a ticket or Oh, there's a blind person at the crosswalk. We may need to trigger the light There's also equivalent use cases in IOT health care retail Agriculture and many other industries around the world So we're going to go into the different main components of our demo. The first was the messaging component For our main core messaging we use Apache Kafka, which is a great distributed streaming platform that provides High throughput and low latency. We deployed it to our OpenShift cluster using the Strimsy operator Which made it really easy and streamless. We also deployed two other Kafka pieces an HTTP Kafka bridge that takes in HLS HTTP live streaming data from our simulated edge video source Which is simply pulling video from a YouTube live stream specifically one in Jackson, Hawaii I'm in and sending it to the Kafka bridge. The other piece was a Kafka Knative source that allows us to connect our Kafka messaging system to a Knative serverless service The next main components were the analytics with TensorFlow serving the main service architecture with Knative and then also our persistent storage architecture with Ceph essentially the first video service allows us to take in video from our Kafka messaging component and Run real-time analytics with a TensorFlow serving deployment Then we have persistent storage done with Ceph object storage, which acting as the back end for an HLS server That is completed by our video streaming service The best part about this whole thing is that the video service and the video streaming service can scale up and down Automatically because they are serverless functions So now let's go into the actual demo This is the overall system diagram for our demo. I'm just going to show the end video streaming service Deploying and being used So as you can see we've already been sending during this presentation data To our Kafka HTTP bridge. So now we're ready to go ahead and Deploy our video serving service great. It's deployed Let's look at our cluster You can see the services up Let's get the link for that service with OC get KSVC We're going to follow this link and then type in slash video and Then the name of your video that you want to watch So in this case it's just out m3 u8 because we're actually using the HLS playlist file And as you can see we get a live stream up that is doing real-time analytics showing cars showing traffic lights and Any other discernible object in the video? One of the coolest parts about this demo is that we have the ability to rewind as well Since we are using Ceph as persistent storage so I can go backwards in time and still view analytics Thank you so much for tuning into this demo Stay tuned for future demos and please let me know if you have any questions. Thank you