The OpenDaylight Project's fourth release, Beryllium, became available today, adding some features aimed at bringing the software-defined networking (SDN) framework up to speed in production environments.

In that sense, Beryllium provides a nice point for stepping back and seeing that OpenDaylight is starting to be shaped by user input, which in turn is prompted by real plans for networks. So says Neela Jacques, OpenDaylight's executive director.

"We have a broad platform that can be like Linux — something that can support a broad range of use cases," he says.

"People are asking big questions instead of using it like a toy," says Colin Dixon, chair of the OpenDaylight Technical Steering Committee.

Among the use cases emerging is network visibility. It's a lower-level job than the network virtualization OpenDaylight aims to do, but for companies that are skeptical about using the framework, network visibility is a way to get something out of the technology while entailing minimal risk.

"It's a simple use case one or two people can put together," Jacques says. "Think of it as a 'starter' OpenDaylight deployment."

Here are a few highlights of the Beryllium release:

Beryllium includes the most serious work yet regarding the ability to cluster OpenDaylight for high availability and for general scalability. (Clustering refers to having multiple OpenDaylight controllers behave as one controller.)

There's also been a lot of work toward stability. OpenDaylight has spent $1 million on test and integration, not including the volunteer work that went into those activities. It's a level of test "uncommon for open source projects," says Dixon.

Scaling has also improved. With the previous release, Lithium, it was a big deal for OpenDaylight to be able to say everything worked reliably in deployments of at least 500 switches. "We now have that just pegged," Dixon says. "This is exactly the kind of stuff you'd have to get right as you move to production."

Finally, Beryllium brings OpenDaylight up to speed with four methods of group-based policy: the NeMo network modeling language, Application Layer Traffic Optimization (ALTO), Group Based Policy (GBP), and Network Intent Composition (NIC). These, says Dixon, are "more user-friendly parts" that are "having huge adoption in non-hyperscale environments."