Fast Hadoop use cases

- [Instructor] So change in any ecosystem is driven by customer need and it's the same in the world of Fast Hadoop. So as more and more data is generated and can be more cheaply and easily stored, whether it's locally or more commonly, in the new cloud based data lakes, whether they're on Amazon NS3, or Google, or Azure or something else. Customers are more and more wanting to have fast analysis of these huge lakes of data. Example use cases are around scenarios such as stream data processing, so streaming, extract, transform, and load, so in other words, taking data that's coming in and processing it showing where there's exceptions or outliers or bad data and responding.

Data enrichment, taking data that's streaming in and maybe combining it with other data sources whether they're authoritative or evaluative like machine learning and putting additional tagged data so that the data can be then further analyzed downstream. Another use case I've seen from my customers is trigger event detection. So this can be around devices, it can be around systems, where thresholds are exceeded. And the idea here is around fast. That the data coming in can be quickly evaluated as an anomaly.

We are probably all familiar with a long time user of a fast Hadoop system, which is commercial credit card services. I know I'm glad when I get that phone call that says, "Hey, this purchasing pattern seems to be "out of normal behavior, has your card been compromised?" This kind of functionality is coming to more and more businesses, as the Haboop ecosystem is evolving. Another set of functionalities around machine learning, specifically around sentiment analysis. So a real common use case that pretty much everyone's familiar with is Twitter sentiment.

Bringing that information via linguistic analysis more quickly to, for example, a commercial client can result in turning around customers who've had bad experiences and fixing those situations to bringing new customers on, people who are looking for a service and can't find it. Another area that I've done some work here is around what's called fog computing. It's kind of a funny name, but the idea is that you have IoT devices in some sort of secondary location. Could be a commercial business or even your home, that are not directly connected to the public internet all the time.

Maybe indirectly connected through an edge layer. In case of your home, that could be your router. So the fog is the edge, it's kind of a strange name, but the idea is to get the information from the devices, the event information from the edge more quickly so that you can again respond. So, for example, if it's raining outside so you can turn your sprinklers off so you're not like wasting water type of thing. So the whole idea around this is to take more information than you're getting and to respond to it more quickly.