The Somerville, Mass., company is announcing Tangosol Coherence 3.0, with new features such as WAN clustering support, clustered JMX (Java Management Extensions) support, read-ahead caching, and custom partitioning and partition affinity.
Cameron Purdy, president of Tangosol Inc., said Coherence leads the market for Java-based, in-memory distributed data management, clustered caching and data-intensive grid computing.
"Coherence is a self-forming, dynamically organizing, transactional, highly resilient, self-healing and lossless data fabric," Purdy said. "That means it doesnt lose data when servers die, even though that data is being managed in memory."
Purdy said Tangosol pioneered the use of peer-to-peer clustering for data grids, but with Coherence 3.0, the company extends that architecture to support disaster-recovery sites and WAN environments.
Meanwhile, support for JMX means Coherence can manage large-scale grid deployment from any application management tool, the company said.
The read-ahead caching mechanism enables the use of data-eviction policies that load fresh data into the cache, enabling the application to be ready for the next request, the company said.
And custom partitioning enables the fine-tuning of data load balancing in a data grid, while partition affinity works with custom partitioning to make sure that related objects are load-balanced to the same physical location, Tangosol officials said.
"Our first focus has always been on HA [high availability], and our second focus on scalability," Purdy said. "Nothing in Java touches the HA scalability that Coherence has shown. Nothing has even come close. We measured over half a million HA messages per second in one production application."
Bruce Tate, a Java consultant and head of his own Austin, Texas-based practice known as J2Life LLC, said in a statement: "With Coherence, you can maintain a simple code base and still get extraordinary caching. You can run inside or outside an application server, depending on the needs of your application. You can even reduce your risk by clustering across two completely separate data centers, and provide high availability combined with in-memory data access for any Java application."
Moreover, Purdy added, "If we could find a demanding enough application, it would easily scale out to thousands of CPUs. Our resource load per node stays constant, regardless of the number of nodes. Thats one of the ways that we achieve linear, horizontal scale-out."
And Tangosols solution is able to deliver this functionality "without any single points of failure," Purdy said. "Machines come and machines go, machines lock up, but the data grid not only keeps running, but it does so without losing any of the data that it is managing, even when machines are dying."
Tangosol is the specification lead for the Java caching API known as JCache or Java Specification Request 107.