IA: You joined MapR last October after a long stint at Oracle. What attracted to you to the startup?
MM: I was at Oracle for more than twenty years, then took some time off and was trying to figure out what to do next. I was interested in the growth and excitement of smaller companies, after having the opportunity to be involved in smaller companies through the M&A machine at Oracle.
The big data market interested me, and knowing a lot of people I had opportunity to talk to Cloudera and Hortonworks, but they didn’t get me terribly excited. When John [Schroeder, MapR CEO] called my initial reaction was the same, but they twisted my arm in coming out and spending the day with the team and I was hugely surprised at the product and the fact that they had such a different go to market model. MapR truly are a software company- that resonated with me as that’s what I’ve been doing for the past few decades.
While many companies are service businesses, 90% of MapR’s revenues are our product, and 10% are services, so we look and feel much like a traditional software company. We are running 80-90% gross margins, are on a nice path to profitability, and those are things you might expect to see in a software company.
IA: What are the similarities and differences between what MapR and Oracle are trying to achieve?
MM: I oversaw a larger business at Oracle, close to 8,000 folks, but when you’re in the field it doesn’t feel any different. The thing that is terribly exciting is we are in the right place and the right time.
The total addressable market is significant. Looking at some of the bankers it’s upwards of 60 billion dollars, and that market size reflects the differentiated converged data platform we have.
Oracle is certainly playing in the big data and analytics space but I would tell you if you look at our path we are not getting into that heavy duty transactional processing world, that’s not our play. It’s interesting if you look at how our product continues to evolve we are certainly taking on the key tenants you might see in a big OLTP (Online Transactional Processing) database.
But we call it the global namespace. To me it means single global instance, where you can have a single distributed data base sitting around the world and it comes across as a single instance of this data.
That’s something that’s been very important to c-level executives for years and now all of a sudden we have multi-tenancy – it’s something you really have not expected to see in our world up until now, so it’s very different. We don’t compete much with the Oracles of the world.
IA: In February MapR was granted its first patent for its technique for converging open source, enterprise storage, NoSQL and other event streams. How will the patent help MapR to differentiate in the overcrowded Hadoop big data market?
MM: At the highest level we have converged enterprise storage and Hadoop together. Many of the assumptions of Hadoop, it being batch, it being for limited use cases, it being not completely enterprise grade – well if you converge these two that changes everything.
That’s the fundamental piece of our differentiation. And then we converged the NoSQL database and global event streaming to enable real time capabilities, and we continue to do that with our additional innovations.
We are probably in the midst of the biggest change in enterprise computing in decades. A lot of the traditional ways to address these problems is focused on dedicated silos that do special processing. With our converged platform we are bringing a diverse set of capabilities into one platform.
Why that’s important is that we are looking at apps that need the analytics in real time so they can react faster and take advantage of the business opportunity. There aren’t a lot of players out there that can do what we do, that’s why this market is such an exciting prospect.
The underlying details of that patent are significant in terms of how you take data processing and perform that at scale and maintain reliability and data protection. We have chosen to pursue that move in a really unique way, because while we are innovating at that underlying data platform we have done it by supporting industry standards. One of those standards is the Hadoop distributed file system API and one is a standard file system, so there’s no vendor lock-in associated with that innovation.
IA: Do you think Apache Spark is in line to replace Hadoop? What do you advise your customers when deciding which open source distribution to use?
MM: If you look at MapR, they are completely complimentary. A lot of the tech we have done lowers the bar and makes it easy for companies to be successful, and a lot of the things we have done with hadoop have extended to spark as well. We provide both so we are not asking customers to understand it on a deep technical level – they can do either.
There is a lot of excitement around Spark and we are supporting that with our free on-demand training course and our quick start solutions.
Spark simplifies the development process, so it’s easy for companies to come and develop a big data app with Spark. That said, there are some advantages of some of the MapReduce and YARN processing for deeper analytics at scale that are used side by side with Spark. It’s not an ‘either/or’.