 Hi everybody, we're back. This is Dave Vellante of Wikibon.org and I'm here with my co-host Jeff Kelly. This is theCUBE SiliconANGLE's continuous production. Two days wall-to-wall coverage of Hadoop Summit. We're here at the San Jose Convention Center. Lot of action going on today, lot of innovation. We're hearing some great keynotes. Muhammad Elmala is here. He's the manager of enterprise applications and architecture at Children's Hospital in Los Angeles. We're going to talk about how they are using Hadoop to solve healthcare problems. Muhammad, thanks very much for coming on theCUBE. It's really a pleasure having you. My pleasure. So, tell us a little bit about the hospital. When we start there and in your role. Right, so Children's Hospital of Los Angeles has been around for about 100 years plus and we've implemented our Electronical Medical Record in 2004 for the inpatients. And then we're currently actually, as we speak, we're rolling out the outpatient, so VMR for our 30 plus clinics. So we're motivated, of course, with the success that we had with our inpatient EMR and with everything that is happening in the healthcare around us. So the EMR is definitely not the only thing in a hospital. There is hundreds of other applications that we maintain and introduce every couple of weeks. There's a new app. But definitely EMR is the main core piece of the hospital operations. Yeah, I was just saying, so the hospital's application portfolios are very diverse. Lots of small applications, some very large applications. Things change quite frequently. You guys are regulated in the number of ways. And you're under severe budget constraints. So those are some of the drivers. What are the big drivers that lead you to a place like this conference, Hadoop Summit, and Hadoop in general? Right, so there are actually multiples, but the one that I was actually speaking about today was related to the research department. So we have the Saban Research Institute. And what does the Saban researchers need is continuous access to large amounts of data. Not only our patient data, but also patients from around the world that their data is de-identified, we remove the protected health information from them. And then researchers around the world can use these to find data sets for their medications, research for trials, and what have you. So they have been coming to us for years, you know, to get them access to the EMR, to give them access to our data sources. And we've always been challenged by the ability to give them that access because these systems have accumulated large number of data. And they usually need all this data, all this history for again, continuous amount of time or for a long time. So it's not something that we can do without putting a lot of serious money in it. When Hadoop came about, or at least when I heard about Hadoop a year plus ago, I was intrigued because I felt that there is a match between the two. The need from research for this data, and as you've mentioned, the limited budget. So instead of allocating space on sands that are expensive or creating an enterprise data warehouse that like we used to think before, this idea came and says, okay, there's another way cheaper and we'll give you what you need. Yeah, when you talk to the original people that developed Hadoop and some of the applications around Hadoop, they always talk about the container, how much money used to go into the box, whether it was a storage array, you mentioned sands. So you're saying that the economics of Hadoop really are what attracted you and enabled you to develop these new applications for your research. True, but not the only reason. Definitely the, actually the good match was also because the researchers in nature, they are also intrigued about the data. They are data scientists by default, many of them at least are. So they were actually, when I first mentioned the Hadoop and the idea of having a Hadoop in-house, they were really amazed. They really thought that this is something that only a researcher would have thought about. They didn't think that a health IT or an enterprise department of the hospital would be also interested in having a Hadoop. So once we found the mutual interest, we started the application. Do you feel like you're unique in that regard? Somebody who brought this innovation to the quote unquote line of business. I mean so often you hear about IT, skeptics, no we can't do that and so forth. So are you unique in that regard? Or do you think that Hadoop is of the nature that anybody in your position sees the potential? And what are you hearing from your peers? I wouldn't say that I'm unique in this. I think there are other hospitals and other care providers that have actually stepped into the plate and started with Hadoop before we do. So by any means I don't think we're unique. I think it just matter of timing. Some people are already a little bit earlier than others and it's all about all the other factors that plays in. So you have to be in the right place in the right time. My moment came a few months ago. Other people actually did it a year ago and I'm sure other people will follow in the years to come. So Mohamed you talked about the economic factor in terms of storage and how Hadoop is a solution that allows you to store large volumes of data at a really what is essentially a fraction of the cost that you would pay to some other types of solutions. But I wonder if we can move up the stack a little bit if you will. So you're obviously storing this data so that you can provide access to it to your researchers. Talk about a little bit about how you do that and the actual applications you're building on top and then maybe we can go into some of the real, we usually say business value, but I guess maybe patient value or research value of the data and getting that value out of the data that's stored in Hadoop. Right, so definitely the financial factor and the HDFS being relatively cheaper. I just don't want to dismiss that as the only factor. There are definitely other factors and I started on the note that our researchers are data scientists or at least they play or put that hat on in many times. So the fact that we give them that platform that again, that was a perfect match for them because they were all the type of developers that they will jump in and use Java or other scripting or open source language to run their patch processing or queries or what have you against the Hadoop cluster directly. They don't need us really to build a whole tier above Hadoop, but nevertheless we did actually provide them with a simple UI, nothing fancy or complicated due to the nature of it. And that UI with a simple HTML page was a little bit of JavaScript that makes web services calls directly to the HDFS. So the idea is it's relatively not meant for our physicians or end users that are not of a technical nature. It was targeting a specific community which is other researchers. So it's a little bit dry compared to the EMR or other applications that you will find. So you've got some fairly sophisticated users that it sounds like. So tell us about some of the problems they're tackling. How are they using the data? What are the use cases? Right, so the main use case was that they wanted to, they had a grant already secured that needed them to look at vital signs of patients as they are inpatient or inside the hospitals, stay in the hospital. So the way that we are doing it now or before actually was that the nurse or the care provider will take a snapshot of the vital signs. So for example, after the nurse administers the medication, she wants to see if any affected the blood pressure, for example. So it takes a snapshot of that and that only snapshot of that moment will go to the EMR. That data is usually very unique. The nurse of nature, they don't have the time to do that a lot. So if they do it three or four times a day, that will be the most that they can afford. The researcher from the other side wanted to see a whole history, continuous curve of the blood pressure, for example. They go up and down with the sleeve, with the medication so they can find patterns basically that is related or can be correlated to the events that are happening to the patient while at the hospital, for example, as we mentioned, the medication. So that continuous storage of the data with high frequency because we're not getting this every hour, we're getting it every few seconds. Couldn't be achieved just with a SQL server which we have actually as the first hop to aggregate the data, we put it in a SQL server and then from the SQL server we scoop it into HDFS. So this is data that these vital sign data is, so this is essentially streaming off machines? Exactly. Talk about that a little bit and the impact that that is going to have on your industry. We brought theCUBE to General Electric had an event last week talking about the industrial internet. Use cases, of course, is all data generated by industrial equipment in the healthcare field, so MRI machines, could be blood pressure, any number of machines. How do you see that kind of developing and impacting the way doctors, nurses, under clinicians deliver care? So there are two pieces to point of views. The one point of view that we're looking at right now and have been describing is the research which is to a certain extent you can think of it as after the fact. So the patient most likely will be out of the hospital by the time we look at this data and analyze it. There is also the dimension of, as you've mentioned, the things that we can do with this data on the spot. So I know of companies out there that provide some monitoring of that feed that comes in and basically ability to alert nurses and care provider if there is a certain jump in the heartbeat or any of the other numbers that are coming in compared to the threshold of the normal for someone in, for example, in their age or so on. So there is, again, there is the aspect of the unlimited things that we can do it for research because this data used to be deleted before. Nobody could afford to store this data earlier in an expensive storage. Right now we have all this data for as long as we want. The other piece of it that I would like to highlight is before we used to, if there is a specific alerting feature or server that we wanted to do, they used to hook their cables and their hubs directly to the device itself to be able to get this data out. Sometimes it was stored on premises, sometimes we ship it on the cloud and then they analyze it and provide us with some analytics on it. But that was not going far because no matter how large or scalable the device is, you have four ports, you have eight serial ports, you have a limited number of ports. So we couldn't use that model for long. We had to stop at some point and say, this cannot continue doing this. Let's get this data one time only, get all the data one time only, put it in a centralized place and then from that, you can use analytics from different vendors, from different provider as much as you want. So it's not only provided the ability to solve the existing problem at hand but offered us a scalable solution for the future. So as you look to extend access to some of this data to kind of the frontline clinicians and doctors and nurses on the floor, how do you see that evolving? Now we talked a little bit earlier about, your research is fairly savvy when it comes to working with data. Doctors and nurses generally are not, don't consider themselves data scientists. So how do you plan on or do you plan on extending access to some of this data and analytics to them and how will you tackle that issue of kind of meeting them where their skill set is? Right, so thankfully the Electronical Medical Records systems are sweet. Nowadays are relatively open. They are not like a few years ago which was very hard to change and customize or add pages to it. So right now the EMR that we use currently is relatively open which allows us to customize the user interface, the middle tier web services to feed data from any source really, not just the Hadoop ecosystem but any source which we are currently doing for other purposes. So for example we have contact information that we feed from our resource directory into the EMR. So while the physician or the nurse inside the EMR they can look up the contact information of other providers without leaving the EMR like it used to do, like they used to do before. So given the same model which is relatively decoupling or lightly decoupling and at the same time using service oriented architecture will be able to get the data out of the Hadoop feed it into the EMR transparently for our end users to consume. So it sounds like so the EMR application is where the doctors and nurses spend the majority of their time when they're interacting. So the idea is to kind of embed that analytics into the environment rather than having them have to go to a different application or system to see the analytics. Exactly but definitely it's not the only way. We also provide them with mobile access. Not to all of the data we're still not there yet but we definitely want the physicians and the care providers to be able, if they wish, if they prefer to, they can access it from a mobile device and by logging in they can get this data as well. Excellent, Mohamed. Well listen, thanks very much for stopping by the Cube and sharing your insights. Good luck with your initiatives. I'm very impressed with your forward thinkingness and hope to see you around the Hadoop shows in the future. Pleasure is mine, thank you. Keep it right there everybody. I'll be back after this. This is Dave Vellante, I'm here with Jeff Kelly, John Furrier is also in the house. We'll be back with our next guest. We're live, this is the Cube from Hadoop Summit 2013. Right back.