Red Hat has revised its JBoss Data Grid software package, which now offers the ability to replicate copies of data across different data centers, and also comes with a number of other new features that can limit downtime.
Red Hat hopes that the updates will further entice enterprises to consider using a distributed in-memory data store.
[ Andrew C. Oliver answers the question on everyone's mind: Which freaking database should I use? | Keep up with the latest approaches to managing information overload and compliance in InfoWorld's Enterprise Data Explosion Digital Spotlight. ]
"Most customers are facing challenges in providing access to real-time information, handling huge data transaction volumes, and meeting high uptime expectations. They are unable to solve these problems by scaling a rigid, complex and very expensive traditional data tier," said Christina Wong, who is the Red Hat product marketing manager for the JBoss Data Grid, in an email interview. "So, there is a demand for flexible, high-performance [systems] that can reduce the overhead of interaction with the data tier."
Red Hat bills JBoss Data Grid as a distributed in-memory data store, one that can be spread out across multiple servers for improved reliability and responsiveness. Maintaining a collection of data in working memory can help enterprise applications run more quickly, because they don't have to store or draw data from disk-based database systems. Distributing multiple copies of the data across different locations could also make the application more reliable.
Although not yet a common practice, using a data store instead of a traditional relational database can be a time-saver as well, according to Curt Monash, of database analyst firm Monash Research. "There's often no point in shredding objects into a relational schema. So why not store the objects directly instead?" he wrote in an email interview. It also simplifies the software stack by eliminating the need for ORM (object-relational mapping) tools like Hibernate, Wong said.
Wong maintained that organizations are increasingly becoming comfortable with keeping their data in working memory, rather than committing it to disk, at least initially. Adoption of in-memory technologies such as the Data Grid "is driven by the rapidly evolving business environment, which is becoming more global, distributed, real-time and is seeing huge growth in needs for speed and low latencies," she said. Red Hat has been pitching the Data Grid to financial services, telecommunications, and media and entertainment companies, among other markets.
One new feature in version 6.1 is the ability to replicate data across different clusters, which could be located in different data centers. Spreading the data across multiple locations can improve application responsiveness and increase uptime.
The new version also implements NBST (non-blocking state transfer), which allows nodes to be added or removed from a cluster with minimal interruptions to the application's operation.
"Without non-blocking state transfer, ongoing and queued operations will be put on hold while the re-balancing occurs. However, with non-blocking state transfer, ongoing operations and performance is not affected," Wong said.
Support for Map/Reduce functions have also been improved, allowing long-running applications to operate more smoothly. This version also supports CDI (context dependency injection), a technique favored by Java developers to simplify the architecture of their programs. "The CDI integration in JBoss Data Grid allows users to configure and set up the data grid in a manner consistent with the CDI programming model. Further, JBoss Data Grid components can be injected into CDI beans," Wong said.
Those using Red Hat's Hot Rod cache can enjoy the ability to upgrade JBoss Data Grid without downtime in the new version of the software. The Data Grid software also can respond to API (application programming interface) requests rendered in the Memcache and REST (Representational State Transfer) formats.
Red Hat JBoss Data Grid is based on the open-source Infinispan in-memory data store.