JCache, Longest Running Java Spec, Finally Ready

News about Java Specification Requests (JSRs) don't usually make it to the front page, but the announcement that JSR-107, the spec request for a Java Temporary Caching API, better known as JCache, has earned final approval should be top-of-the-fold news -- if for no other reason than the time it took to get there.

The JCache project summary explains that the spec standardizes in-process caching of Java objects "in a way that allows an efficient implementation, and removes from the programmer the burden of implementing cache expiration, mutual exclusion, spooling, and cache consistency."

JSR-107 was the longest running spec request in the history of Java and the Java Community Process (JCP). The JSR-107 Expert Group was formed in 2001. But the spec languished for about a decade until Oracle and Terracotta teamed up and began pushing to get the spec into Java EE 7. Terracotta implemented a version of the JCache spec in Ehcache, a widely deployed open-source Java caching solution the company acquired in 2009. But JCache didn't make it into the Java EE 7 release.

"JCache provides a well thought out, standardized API and programming contract for Java caching," said Greg Luck, Hazelcast CTO and co-author of the JCache spec, in a statement. Luck left Terracotta earlier this year to join Hazelcast, which develops, distributes and supports a leading open source in-memory data grid (IMDG). The product, also called Hazelcast, is licensed under an Apache license that allows developers to include the grid in their applications. The company also provides a commercially licensed Hazelcast Enterprise edition, as well as a management and deployment console called Hazelcast Management Center.

Hazelcast is set to release a production class implementation of JCache, Luck told ADTmag in an e-mail. He plans to continue to his work on JCache, "taking up a co-maintenance lead role" in the spec. "Hazelcast plans to deeply bake in JCache, so that Hazelcast is standards based and may be used for caching without vendor lock-in," Luck said.

Hazelcast's senior solution architect Chris Engelbert also severs on the JCache expert group. Oracle's Brian Oliver and Cameron Purdy also serve as spec leads on the JSR.

Last year, Gartner predicted that in-memory computing is "racing toward mainstream adoption." The analyst firm expected the relatively small IMDG subset of the in-memory computing market (IMC) to grow fast and to reach $1 billion by 2016.

Yet Forrester Research analyst Mike Gualtieri now suggests that JCache may be arriving in an environment populated with products featuring their own in-memory database engines that may eliminate or delay the need among some companies for a caching grid. "The primary use case for in-memory caching is to speed up applications that get bogged down by slower database bottlenecks," Gualtieri said. "A hot trend among the database vendors, such as SAP Hana and Microsoft SQL Server 14, is in-memory capabilities that dramatically speed up database access. These databases aren't just using more memory for old-school indexes of record caching. They have added in-memory database engines that take advantage of the hardware characteristics of RAM. Also, distributed NoSQL databases, such as MongoDB, can perform very well and have the same distributed architecture as Hazelcast and Terracotta."

Posted by John K. Waters on March 25, 2014