Saurabh Mishra 32:44
Show? Yeah, so look, Event Stream Processing actually falls in this space called a complex event processing. So si p, right. I mean, so the underlying idea behind the SI P is that it's, it's designed to process complex events that are that are coming to it constantly. So I think the the origin of si P is from capital trading, because the spreads coming high frequency trading happening, and you're kind of kind of processing that and analyzing that to figure out what's going on. So that's that was the genesis, if I remember, right, for these platforms. So we actually invested in this time back right before we even formalized the IoT group that SAS because one of the one of the gentlemen who actually visited SAS for us the RND, behind this actually has done a number of si peas in his previous life. And he he knew that the combination of a CP where we are talking about just even processing with analytics is going to be a killer. Right? So he came to SAS with that vision, and he brought out start form, which because seven Stream Processing ground up. So his goal was to be able to bring world class CP, which is essentially being able to process millions of events per second. and combine that with analytics. So that's what resulted in SAS Event Stream Processing. Now, it's actually and I think of anonymously bias, but I think there's an incredible product, it is it is flexible, are you do actually have an engine, which is small enough, you can actually put it in a small Docker container that can give you an AI can run on a small gateway device. We have use cases that are happening on Raspberry Pi type things more more from a PLC perspective. But we run the same engine on that train example that I talked about, be on the same engine in a cloud scenario. So another customer example is a Volvo Trucks. They have a fleet of their trucks in North America closer to 100,000. Trucks sending telemetry data all the time. And we run this engine at the the back end environment there and analyzing this data all the time. So this is all the telemetry data, this is fault codes. So we're able to kind of understand and figure out like what's going on with these trucks. So when the truck pull in, we have already analyzed the data, and kind of helps reduce the diagnosis and repair time. But But going back to Event Stream Processing, so it's really this, this platform that allows you to bring analytics to data in motion, right. So again, the different paradigm earlier, we used to bring data to rest and then applied analytics on it. And we're turning that around, we are bringing analytics to data in motion. So that's the that's the core idea behind it. And, and it's a versatile platform, it allows you to do simple things like you know, if you wanted to do SQL type logic on a time motion. And so you could have a number of incoming data, data elements and you're doing SQL type logic, math thing, you can do that, you're looking for a particular pattern that in my stream, if event A happen followed by b followed by C, in this time interval raise an alert can do that. But where it gets really interesting is that we're able to combine all of this with, with machine learning now. So for example, we have algorithms that are trained to detect outliers, right. So even if you haven't trained this data, if you have just an A new source of data that started to throw this or these are items will understand the steady state of the asset could say from where the data is coming from. And then once it understands the steady state, it can begin to report against that steady state that I've seen a deviation and how bad the deviation is. Now, in this case, there's enough for you to go and do something with it. But what we truly want to do is that over time, if I have enough data about this asset, then I can actually create a predictive model, that when this asset fails, these are the parameters that are indicative of that failure, then I can actually bring that inside the Event Stream Processing. And that way, as soon as I begin to show signs of that kind of behavior, I can alert somebody. The other aspect where ESP comes in is that if a lot of times when we're going to customers, they have their data science teams working with open source, right, so they're doing Python stuff, they're using other open source frameworks. And which is fine that because that's that's just the nature of what what we live in right now. So we can also take new open source of Gotham, and operationalize that within Event Stream Processing. So we can create a pipeline where we are getting data in, you may be using some SAS ESP functionality to manipulate the data. But in a subsequent step two, you already have an analytic that you built that said, with Python, you can actually use that, and then do whatever else you have to do. So it can combine SAS plus open source and pretty easy way. So that is a platform that has a delicate environment, drag and drop, we also have programming interfaces, if you are python programming, you don't have to deal with the drag and drop UI environment. The engine is his waters, the the core processing, and that's lightweight, so it can go on the edge small devices. And we can also take into the cloud. So it's pretty versatile from that perspective.