There’s clearly a lot of interest these days in building stateful applications using containers, but not much agreement about how best to go about it. A large number of approaches have been put forward that, to one degree or another, require IT organizations to change how they manage storage.
MapR Technologies, a provider of a distribution of Hadoop, is pursuing an approach to building stateful applications that doesn’t require any change to the underlying storage systems. At the Strata + Hadoop World conference this week, MapR Technologies announced it has added support for microservices-specific volumes to its Converged Data Platform.
In addition, MapR Technologies is adding support for microservices for A-B and multivariate testing to enable rapid machine-learning model development and optimization.
Jack Norris, chief marketing officer at MapR Technologies, says rather than having to acquire a new storage system capable of supporting containers, the Converged Data Platform abstracts away that complexity by providing support for stateful container applications that can be deployed directly on top of Hadoop. The major advantage of that approach is that it provides container applications with access to massive amounts of Big Data that increasingly is stored in Hadoop.
MapR Technologies has been making a concerted effort to marry Hadoop to a broad range of microsevices technologies as part of an effort to woo developers of new applications. Rather than thinking of Hadoop as a data repository, MapR Technologies has been extending Hadoop to support real-time analytics applications. Norris says the next big challenge is integrating those applications with event-driven microservices to create new types of data flows that previously would not have been either practical or even feasible. The real challenge is that many of those microservices are ephemeral in nature, which creates demand for a platform that makes it simpler to dynamically integrate those microservices into a larger data fabric that IT operations can manage using a consistent set of repeatable processes, says Norris.
For all the excitement developers have about microservices applications need to be able to access data to have any value to the business. As a platform Hadoop is rapidly becoming the data lake that organizations are implementing to provide a standard framework for accessing all types of raw data. For developers, that creates a unique opportunity to gain access to data that they otherwise would have to navigate using application programming interfaces (APIs). In a Hadoop platform not only is all the data that need to be access centralized, many developers will discover that much of that data has already been correlated and analyzed. That data is often going to be a lot richer than what developers are likely to discover navigating multiple data sources.
Obviously, there are multiple distributions of Hadoop that organizations are likely to deploy. But as developers continue to exercise more influence over what IT infrastructure actually gets deployed in the enterprise, MapR Technologies is betting that developers looking for the richest sources of data possible will force the platform decision one way versus another.