Follow Datanami:
August 20, 2015

Riak Taps Mesos for ‘Push Button’ NoSQL Scalability

At the MesosCon conference this week, Basho Technologies and Cisco are demonstrating the Riak key-value store database running atop Apache Mesos, the next-gen cluster manager developed by UC Berkeley’s AMPLab. As Basho CEO Adam Wray tells Datanami, it’s all about making data delivery more scalable.

Basho develops a multi-modal NoSQL database that includes a key-value store and an object store, along with hooks into other big data tools like Apache Spark, Redis, and Solr—with graph and columnar analytic capabilities possibly on the way. While the company is steering its new Data Platform to be a singular store that addresses multiple big data needs, it’s still known primarily for its key-value store, Redis KV.

Basho and its partner Cisco did the work to get Basho KV running atop Apache Mesos, a resource manager that controls how application workloads access the underlying infrastructure resources in a data center. Wray says Basho KV and Mesos will provide a winning combination for scale-out data services.

“When you’re leveraging our platform, not only do you not have to worry about the distributed data tier, but now you don’t have to worry about the infrastructure resources that are allocated against it as Mesos can handle that on-demand component of server allocation,” Wray says. “When we roll it out in GA…we’ll effectively have an auto-scaling capability for their database on premise, based on multiple data centers, all because Mesos handles that infrastructure tier.”

Cisco plans to use the two technologies with Intercloud, the distributed global data delivery platform that Cisco maintains along with more than 40 other federated service providers. Cisco gains a certain level of “push button” scalability that will take a lot of hassle out of ensuring that the distributed database has enough physical computing resources underneath it to do what clients need it to do.

“We’re making it easier for customers to develop and deploy highly complex, distributed applications for big data and IoT,” says Ken Owens, CTO for Cisco Intercloud Services. “

Wray likes what he sees in Mesos, especially as a replacement for certain aspects of Zookeeper, which the key-value store currently uses. “We do use a component of Zookeeper, but we built code around the aspects of Zookeeper for scheduling that weren’t efficient for a distributed environment,” he says. “There’s a possibility we might replace all of Zookeeper someday.”

Basho is certified to work with DCOS, the data center operating system based on Mesos that’s being developed by Mesosphere, the commercial outfit behind Mesos. It’s still early days for DCOS and the products of that ilk, but Wray foresees a day when next-gen resource schedulers are commonplace in the enterprise.

“Mesos is the only thing as a distributed scheduler that I’ve seen that can really address an enterprise-grade offering, which is what we build,” he says. “It’s highly probable the vast majority of our clients won’t be using Mesos initially in an active way within their data centers. However, Mesosphere with their DCOS product is looking to change that in the marketplace.”

The vendors claim the integration of Riak KV and Mesos is the first of its kind for an enterprise NoSQL database. DataStax is also doing some work to integrate the Cassandra NoSQL database with the open source resource scheduler.

Related Items:

Project Myriad Brings Hadoop Closer to Mesos

Basho Goes Vertical with Big Data Stack