 Try to go to Alex Williams. Let's go to see what he's got here, and Alex bring it on with Lars Okay, you hear me I'm here with Lars George and Lars I'm not sure if they can hear us out there, but we'll see what we can do here just to talk for a few minutes We are we're here at the Hadoop world Big event and you're here giving out your book you tell us about your book a little bit Yeah So that's the first book on Apache Hage base I felt back then when we discussed that between the I'm a committer with H base as well so I'm a developer by nature and back then when we talked about Probably is a need for for any good project to have a book and so when That's like more than a year ago I got talked into by another colleague of mine if I wouldn't be interested because I've written blog posts on H base So they they approached me and said why not you going ahead and write the book and there it is basically just someone had to do it and I did it and Yeah, it's going going good. I guess I'm French president in general. Yeah, we've seen lots of talks Yeah, yeah, what's the ten of the conversation on H base? I you know project is seen in there Then that was discussed in the keynote and how they're using H base there and they said that there are still some There's there are still some hurdles with it. Yeah, so I try to what we usually have an eye when I talk to People interested in H base. I'm trying to make the point or state the point that it's still an early version of a product Like a 0.90 at the moment we work on 092. So it's not a version one So it's not like a shrink-wrapped software that you can just install and use right? It's way way way behind for example my square way Just install it and you use it and you scale it when you build up as you go So with H base there's a lot of upfront complexity that you need to understand and appreciate so it's a bit of a learning curve and and It's important for a company who wants to be successful with H base that I have a champion in the company that That really takes it on and makes it their own and works with H base And even with the community getting feedback and giving feedback So that it's successful. So if you look at like Cassini or for example there There was a lot of Discussions and a lot of talks going back and forth in the end. It's usually quite okay and It's it's coming comes down to like schema design and a few things But it needs a lot of upfront expertise and that's not too easy to come by. Yeah, that's a cultural change Yes, yes, because it's not an enterprise tool Especially because I work out there. I'm not selling H base as a product I mean we have it a part of our platform So we're giving this like this part of Hadoop we're giving H base as a tool for big data But the point is that this is nothing that we have produced. We're part of the community We're helping the community giving back and working on issues, but we're not really This is not our tools not like a proprietary software So this happens at the pace of open source. Therefore, it's not comparable with we can't compare this with any enterprise Software, so the release dates are slower You can't really take anyone up for missing features because it's up for anyone else help out making It's a better software system. So it's a it's a it's a it's a lot of companies who raise this down I mean you see those face big at that and a lot of other companies or publicly saying we're using H base and But it's took a while for them to regain the Confidence to say so so it's but I think it's it's going going strong. Well, it's a model of the new storage requirements Isn't yes, and now, you know, and really what's increasingly required Not just for companies like Facebook, but for companies and financial services Yeah What we see is sort of the transition from Hadoop, which is the batch-oriented nature So you're building a system that can take in lots of large amount of data, but you're having always a batch-oriented Layer in between so you have to deal with some sort of delay for The data to arrive at the at the at the end So you're running jobs and they they have to run for a while some jobs can take hours to finish now That is great So that's if that's a huge amount of problems for for our customers and for for companies around the world to take The next step forward in terms of storing data, but then you're missing the last bit, which is sort of the real-time component That's why H base comes in a place sort of fills that gap and sort of helps bringing Data and information directly to the end users driving websites as opposed to just generating models that You can use for other things H base is the one that sort of bridges that gap I Hear that criticism about the dupe and its lack of real-time capabilities I wrote a post last week You're talking to a force is to James probably as you gave provide a five Potential distributions that you could do a real-time with you. Is it getting better? How would you how would you mark the state of real-time? Analytics with with you this interesting topic. I think that's it's going to I think that part of our dupe and H base on top of that is going to really really Very much an area of interest for for a lot of companies If you look at if you hear the Facebook talks for example, you can see that a lot of real-time for them It's like half an hour. So when they say real-time analytics They're actually taking 30 minutes from the event on the website to occur or to happen or to show up in the database On the other H base site to drive To their customers. So real-time is a is a is a form of making data available more Current or more immediate, but it's never at the moment. It's not at that scale Where you click on the web and immediately see the result on the on the delivery side So it's currently when most companies talk about real-time. It's this from a few hours down to a few minutes But I'm sure that over the next couple of months. This is going to transition into a discussion of okay How can we actually bring this to actual real-time which means someone clicks on the web and I have the event in H base right away? So I get the counters all updated within milliseconds and I'm really close to what's happening right now But can you do that with a batch oriented? No, the H base or how do is doing still doing the batch oriented building of Models machine learning is a typical issue where you use the data that you have aggregated in how to do Recommendation engines or so we building mathematical models You need to do this batch on and on the entire data set and H base What just stores the last 30 days or the last 30 hours or something? I guess of the data never the entire data set So you're still using the batch oriented huge amount of storage in Hadoop and use H base to bridge The whatever two hours that it needs for the job to run to bring this real-time So what I see in the future is that you have to build the models and you use H base to update the model as it goes live But every two hours every and hours the sting wakes up and recalculates the model updates H base So then the error rate that H base introduces because it doesn't have all of the data available It's going to going to be covered by H base But you won't see this as an as a user You basically get real-time data with a slight increasing error But the bench basically doing the the recomputation Fills there or catches up every few hours And therefore you're having something that looks like it's real-time But you're still using the large capacity of Hadoop to do the actual number crunching and H base will be sort of the The bridge between and keeping things alive and responsive great well large George Thank you very much for taking some time to talk with us. You're welcome. Good luck in your book Thank you very much for to keep the touch. Thank you very much. Hey guys Alex Talking to folks here back to you