1 Comment

Summary:

Red Hat is expanding its set of cloud capabilities by announcing the JBoss Enterprise Data Grid. The product gives customers an in-memory data grid that scales along with the server infrastructure and provides a high-performance cache to offload the demand on the primary database.

infinispan

Red Hat is again expanding its set of cloud capabilities, this time by announcing the JBoss Enterprise Data Grid (EDG). Data grids extend the type of scalability often associated with cloud computing resources to the data layer, and also speed access to data by storing it in a server’s memory instead of in a away from the application in a traditional disk-based database.

The product is based on a JBoss Community project called Infinispan, and gives customers an in-memory data grid that scales along with the server infrastructure and provides a high-performance cache to offload the demand on the primary database. Red Hat is pushing EDG as a complement to its JBoss application platform and Makara platform-as-a-service offerings, which puts it in good company along with VMware, Oracle and GigaSpaces.

According to the press release:

JBoss Enterprise Data Grid is designed from the core to support cloud-scale computing with concepts such as multi-tenancy, elasticity and distributed code execution. Enterprises will have the opportunity to harness those capabilities to deploy highly available, massively scalable and highly performant shared data grids to accelerate applications and curtail data-tier costs.

Put simply, EDG will cache data within servers’ aggregate volumes of memory, giving any applications connected to the grid low-latency access to the distributed pool of data. EDG, like all data grids, scales automatically when new servers are added, and redistributes data across the rest of the pool when servers go down. In the future, says the Infinispan web site, the product could evolve from targeting primarily transactional applications to supporting applications that require some degree of MapReduce-style processing of data.

Data grids are key components of many cloud-platform and PaaS software products — including VMware’s SpringSource (via the 2010 acquisition of GemStone Systems), Oracle Coherence and GigaSpaces eXtreme Application Platform — because they scale along with applications, and easy scalability is a key tenet of cloud computing.

  1. The scalability of these Java caches is limited only to horizontal part. The only way to scale these caches up on a 384GB server box is to launch 50-100 JVM processes with 3-6GB of memory heap each (to avoid long GC pauses). I am not kidding. There is something called BigMemory by Terracotta, which can help in this case but the performance according to some independent tests is not so good, but at least it solves GC long pauses issues.

    Share

Comments have been disabled for this post