Sramana Mitra: Talk to me a bit about the computer architecture side of what you do.
Adam Wray: Riak Core is focused on distributed systems. All of our algorithms are trying to make sure that we have places in multiple locations so that when you have a knottage within the Internet, which is a common thing, or if your servers fail, there’s always another point at which that data is accessible. That’s one thing. That’s a really tough thing to solve. We use what’s called the ring architecture.
The second aspect is the accuracy of the data. The availability component along at large scale is very tough to do. Most of our competitors come tumbling down when faced with that challenge. We have an operational overhead challenge in trying to manage all of that. That’s one half of it. The other half is keeping everything accurate and available. When we started, we were always thinking about enterprise and production level needs. One of the things that will always work for is you is to make sure that everything is accurate regardless.
We’ve spent the last five years building things like conflict resolution devices with the idea that the database is accurately ensuring whether the data being written is actually the right data. I’ll give you an example. You might have three different people across the world writing to a database. Most NoSQL databases use last write went. That might get you the right data but it might not as well. In high probability, let’s say the person who actually was the last one gets there first because the other two people took longer to get through. All of a sudden, you’ve got incorrect data being written down.
We’ve built a series of ways in which you can do conflict resolution so that there’s intelligence within the database to decide what actually is accurate. What everybody in the industry does is, on the NoSQL unstructured side, they say, “This is the data. You got to tell me whether it’s right or not.” This makes it harder for the developers and the business users trying to get value. Of course, in the relational or structured database side, they have the advantage of keeping it all in one location. When one guy starts to write, nobody else can put anything in there until he’s done. It might be very painful when you have a lot of distance involved which distributed systems are designed to handle.
Those two areas of accuracy and availability is the core foundation that we built on. Then we deliver the models and simplification of architecture so the clients don’t have to focus on how to make all these data models work.
This segment is part 3 in the series : Thought Leaders in Big Data: Adam Wray, CEO of Basho
1 2 3 4 5