Hazelcast raises $11 million in funds for further NoSQL expansion
This is an exciting time to be in the field of In-Memory Data Grids, explains Hazelcast’s Miko Matsumura.
It comes as no surprise that Hazelcast is now announcing they have raised $11 million in a series B venture capital round. Hazelcast also announced that Greg Luck, who joined the company as CTO earlier this year, would be taking over as the company’s CEO.
Moving into enterprise
Hazelcast’s VP of marketing and developer relations Miko Matsumura told JAXenter will continue to develop its new venture with Hazelcast 3.3, the company’s most enterprise-class product. “We recently shifted our emphasis towards both supporting Hazelcast itself as well as offering Hazelcast Enterprise. So we’re going to be further developing this roadmap.”
The company is also keeping a close eye on new NoSQL trends.
“One of the big shifts that we’re seeing, that we’re going to be investing in, is some of the use-case patterns around the NoSQL key value store. It’s very exciting for us. We’re seeing increasingly CloudBees, Apache Cassandra and CouchBase and others. For NoSQL-style use cases one of the things people are getting excited about is embedability. All of these things offer open source, so that’s a given.
“But people are also enjoying the elastic in-memory performance characteristics. The independent benchmarks that we have show about 10x speed over Cassandra. So it’s definitely trying to be the speed king in terms of distributing data and scale. That’s big stuff for us.”
Speed and growth
But speed is only part of the reason for Hazelcast’s success, as Miko explains. “If you’re in the In-Memory Data Grid business, you’re sort of inherently capitalising on the three orders of magnitude of speed increase that you get just from principally dealing with memory only.”
Most of Hazelcast’s peers can also boast of impressive speeds, but where Hazelcast stands out is its rate of adoption, which is charted by the company’s recent accolades in JRebel Labs and DB Engines rankings.
The mullet approach
Another interesting aspect of Hazelcast is its combination of NoSQL and an In-Memory Data Grid. “In a way, the funny part is that it sort of has a ‘mullet’ quality to it,” Miko explains. “It’s NoSQL on top and In-Memory Data Grid underneath. In a way, IMDG is a deployment pattern that is somewhat meaningful to operational worlds, and as people get more DevOpsy, the top end of it looks very much like a traditional key value store type approach. That’s kind of the mental model.”
Just a key value store?
Miko is reluctant to admit that Hazelcast is a straight key value store, claiming that many of its users are discovering that it has more than meets the average Java developer’s eye. Its topics and queues are becoming increasingly popular for things like distributed messaging, and Miko claims users are discovering more and more ways to get joy out of the product.
“Obviously it is a key value store, and there’s a wonderful interface. It’s just ultra-familiar in terms of Javaland. You’re talking about java.util concurrent map. So it’s just map put key value. That’s a super a Java-ish way of talking to Hazelcast. But obviously Hazelcast also supports multimap queue, it supports topics, and a whole bunch of other kind of data structures that are native java.util, as well as the JCash interface which is the new Java caching standard. So I guess what I’m really saying is, yes – it is a key value store. But there are lots of other fun ways to get into this.”
Whether or not its users see it as a value store or something more, Hazelcast has found itself increasingly pulled into the realm of large-scale caching use cases, an area where the JCache API has delivered a major boost. That’s why Miko believes that this years JavaOne “is going to be a big JCache party.”
Hazel leaves image via Shutterstock