Sramana Mitra: Are there other use cases?
Shaun Connolly: Let me tell you about the healthcare use case. What it does is show the sensor data use case well. In the healthcare industry we work with healthcare providers as well as university hospitals – the University of California Irvine Medical Center is one example. In many respects, they are looking to optimize patient outcomes at scale. You might have a lot of electronic medical records in existing systems or new systems. The healthcare use cases bring some of the traditional data that might be in systems like EPIC, which is a healthcare-specific database. But they also have sensors on patients that go home. Those sensors collect vital statistics, temperature, etc. So providers are able to monitor patient even after they have left. That information is not only good for keeping track of how patients are doing after they left the hospital, but it can also be useful for clinical trials.
The point is that these new ways of collecting information about the patient have been fundamentally changing over the past few years. Hadoop provides ways to not only capture some of the existing data, but also the new forms of data. So you can begin getting the idea of a large population of patients that have similar escalations, so you can prescribe a similar regiment or a different one – all that has evolved around sensors on patients, providing the information. That is another use case on optimizing patient outcomes and showing the power of the platform.
The last use case is around enabling Hadoop to be a big data platform that drives efficiency. Typically customers will start off with a lot of these targeted use cases. Once they see the power of the platform, they begin to pour more and more data for longer periods into the platform. With a broader perspective of the modern data architecture, Hadoop can be the place where you cost-effectively keep your data for long periods, and you are able to keep your data warehouses focused for however long fits into perspective and serve up that information in a very high-performing way to the business. On a platform like Hadoop, you can do analytics on data that is 10 to 20 years old, which you couldn’t historically do in a data warehouse because either the economics or the size of the data would prohibit it.
There is a strong optimization angle introducing Hadoop into a broader modern-day data architecture, and it enables other data systems being used by the business to stay focused on what they do well and delegate some of these less responsive needs to the Hadoop platform, still retaining the ability to access some of the older data for longer term analytics. Sometimes it is harder to uncover if you only keep 30 to 60 days worth of data. Patterns become apparent when you look at six months, one year, or 10 years of data. Hadoop really shines when you are talking about this older data, uncovering patterns you may not have seen before.