Thought Leaders in Big Data: Adam Wray, CEO of Basho (Part 4)

Sramana Mitra: How does all this fit in the competitive landscape? Can you help me build an ecosystem map of where you sit?

Adam Wray: I’ll take a high-level map and break it down. It starts with the first two trends. The first trend in unstructured data was Hadoop. That trend was driven by the thought process that commodity hardware is cheap. Data, as a strategic enabler, is becoming critical. Hadoop became wildly popular over the last five years. You might even have heard the term data lake bounced around at times. That’s effectively the thought process of putting all your data in one central place.

What people are waking up to in the last two to three years is that it’s actually still expensive. I need to get business value out of it. Right now, the architecture is not designed to be real time but designed for post analysis. The challenge should be thought of from the perspective, “What is the business value you’re after?” Therefore, how can you have accessibility to that data in near real time such that you get to that value and apply it to your end user’s needs.

What NoSQL has sprouted up is the thought process of, “We expect the data to be distributed and we expect they’re going to have a real-time need set.” From the industry perspective, you’ll see that difference. By the way, both have a play. There are times when collection and post analysis is right. I’m just saying that we’re fixated on active workloads and I think our competitors are as well.

If you look within the NoSQL evolution for unstructured data, you have these steps where things are going through. It started with point solutions. Companies like MongoDB have built a very simple-to-use document data store because document profiles are simple to access. People can think of them in the same way that they might have put their data into the SQL database. It allows developers to excel very quickly.

Then, there are other models. There’s Key Value, which we started delivering first. There’s an in-memory component. You could go down this list. What’s happened within the enterprise environment is that once they started to use these, they found that one model works for certain workloads. If I have a lot of document-type structure, I’d want to use a document store whether that’s MongoDB or CouchDB. If I have an open and unstructured environment, I would think that I would need something more open-ended like Key Value. No one size fits all.

They’ve become multi-model in their environment. They’re having to support multiple models versus just one model. This creates complexity. You have to manage the operational overhead of the model, the developer overhead of the model, the synchronization, replication, and cluster management of the data. Usually, that data might be tied to each other and have to talk to each other. Lifecycle-wise, we’re at that stage.

This segment is part 4 in the series : Thought Leaders in Big Data: Adam Wray, CEO of Basho
1 2 3 4 5