Plenty of people in the industry will welcome some healthy competition between OpenDaylight and ONOS, but the projects will also need to find some common ground in order to avoid further fragmentation in the market.
ON.Lab has been working on ONOS for two years, said Guru Parulkar, the lab's executive director and co-founder. The project's leaders will aim to bring ONOS, which is a distributed SDN controller, to network service providers. But the project also has aspirations to pivot the technology toward cloud providers and enterprises later.
"We want to transition the world of networking away from the way it works today -- where vendors are using proprietary router operating systems to sell more hardware -- to an open SDN that enables service creation and scale on any hardware, including white boxes," Parulkar said.
The nuts and bolts of the ONOS controller
ONOS is a modular, scalable and resilient open source SDN controller that network operators can deploy across multiple x86 servers, said Ram Appalaraju, a strategic advisor to the ONOS project. The controller can currently handle up to 1 million path intercepts per second and respond to a network event in 10 milliseconds.
"It's a distributed core that runs on multiple servers," Appalaraju said. "Each instance is identical and they cooperate together to form a single system. If you need more control plane capacity, you add more servers. It also has high availability. If an instance fails, the workload is seamlessly distributed to other systems."
ONOS also has carrier-grade persistence. The state of the entire control plane is stored on every instance at once, which enables hitless updates.
Like OpenDaylight, the southbound abstraction layer of ONOS uses multiple protocols, including OpenFlow, to interact with network infrastructure. It can discover, map, configure and control the network. On the northbound side, ONOS contains an "application intent framework" for policy enforcement, conflict resolution and other operations.
"With the application intent framework, we want administrators to be able to specify what they want to do [without caring] how it's done," Appalaraju said. "If you want to provision a 10 Gb link from data center one to data center two, you can specify that at a high level. Then, the framework translates that down to the infrastructure."
The high-level service abstraction of the application intent framework could prove to be a differentiator for ONOS, said Eric Hanselman, chief analyst of 451 Research LLC. Many projects are trying to achieve the same thing, including OpenDaylight and OpenStack.
"The nirvana we're heading toward is an environment that fully integrates networks with the needs of the applications that are consuming that capacity," Hanselman said.
ONOS vs. OpenDaylight: Scalability, vendor independence
With OpenDaylight on the market already -- and the majority of the networking vendor community involved with the project -- some might wonder whether the world needs another open source SDN controller, while others will welcome its arrival.
"Competition is always good," said Nick Buraglio, a network engineer at a global research network. "OpenDaylight was the first really big one. I like OpenDaylight -- that it's easy to use -- but I would never deploy that in my WAN. For one thing, it's Java. And two, I don't know that scalability is [emphasized in OpenDaylight yet]," said Buraglio. "[The ONOS project] has thought about scalability and security. I'm a wide-area guy: not being able to handle huge scale is a problem in that world. You have to be extremely battle-tested and armored and to be able to support geographical redundancy."
ON.Lab also argues that OpenDaylight is a vendor-driven project, whereas its ONOS project is led by researchers who are trying to build the best open source SDN controller without looking at how to protect legacy business models. Instead, ONOS will look for contributions and partnerships with service providers. NTT Communications and AT&T are already on board, ONOS's Appalaraju said.
"Service providers have agendas, too," Buraglio said. "They will steer it toward whatever is best for them, which boils down to hardware support. It has to be perfect for vendor X because that's what they've committed to. Plus, the oversight of OpenDaylight is the Linux Foundation. At the end of the day, vendors can try to shove it around, but it's open source."
Moreover, OpenDaylight and ONOS will have plenty of overlap in terms of contributors and supporters. AT&T has committed $5 million to ONOS, spread out over five years, to help develop the platform. But that's not the only SDN project AT&T is supporting.
"AT&T is collaborating closely with OpenDaylight, as well as a handful of other groups working on SDN and NFV, including OPNFV, OpenStack and ETSI ISG NFV," AT&T said in a statement. "The aim is to help the ecosystem thrive so that the best solutions come forward."
Open source SDN controller competition is good but fragmentation is not
Now that the industry has two competing open source SDN projects, the possibility of fragmentation is greater than ever.
Many network operators want an open SDN platform that allows them to mix and match, to move from one vendor to another, to migrate high-level abstractions and northbound SDN applications from one controller to another. OpenDaylight has brought together a lot of different vendors, but very few of them have embraced a commercial approach, which would allow customers to jump from one vendor's controller to another. Now with ONOS in play, fragmentation could increase.
"The danger is that we end up with two divergent sets of northbound interfaces, the market fragments, and we don't get critical mass behind either of them," said Joe Skorupa, vice president and distinguished analyst at Gartner Inc. "For ONOS to be important, they need to actively support the same northbound APIs and work very closely with OpenDaylight to make sure that applications are portable. Whether you are an enterprise or a service provider, the greatest disruption is caused if you have to change out an application [because it's incompatible with a new controller]. [The SDN application] is what gets integrated into operational policy and procedures."
ON.Lab may believe it has an opportunity to "blow OpenDaylight out of the water," because many network operators have determined that OpenDaylight "is not quite there yet," Skorupa said. Given that ONOS is the more robust controller today, the project's leaders may elect to just put OpenDaylight out of business, rather than worry about cooperating with the project.
"But there is an awful lot of inertia around OpenDaylight," Skorupa said. "What happens if two or three vendors involved in OpenDaylight jack up [the software], pull out the poor code and replace it with something that runs at a million transactions per second?"
In that case, an arms race would emerge between two very strong open source SDN projects. Cooperation is the smarter path to take, Skorupa said.
Here are five more open source SDN controllers.
Cloud provider adopts OpenContrail open source SDN controller.
Open source hardware: Facebook's Open Compute campaign.
Find out how OpenDaylight puts openness above vendor interest.