A new version of the Open Network Operating System (ONOS), named Blackbird, has been released (the first version of ONOS was out in December 2014).
ONOS features a highly available, scalable SDN control plane featuring northbound and southbound open APIs and paradigms for a diversity of management, control, and service applications across mission critical networks. It is architected as a distributed but logically centralized control plane to achieve high performance, scale-out and high availability. ONOS’ high availability characteristics include full recovery from events such as switch and link failure, node failure, entire ONOS cluster failure, single node cluster failure, cluster partitioning and device-node communication failure.
The ONOS Blackbird release defines the following set of metrics to effectively measure performance and other carrier-grade attributes of the SDN control plane.
Performance Metrics
Topology – link change latency
Topology – switch change latency
Flow operations throughput
Intent (Northbound) install latency
Intent (Northbound) withdraw latency
Intent (Northbound) reroute latency
Intent (Northbound) throughput
Scalability
Ability to scale control plane by adding capacity
High Availability
Uninterrupted operation in the wake of failures, maintenance and upgrades
ONOS aims to achieve extremely high target numbers of 1,000,000 flow operations per second and less than 100 ms (and ideally under 10 ms) latency. Most of ONOS Blackbird release’s measurements meet these targets; the ones that do not will continue to be optimized in the coming releases and in conjunction with use case and deployment requirements.
The Blackbird release also addresses the challenge of effectively determining “the carrier-grade quotient” of the SDN control plane. Metrics currently used to measure performance, including simplistic ones such as “Cbench,” do not provide a complete or accurate view of the SDN control plane capabilities thereby highlighting the need for a more indicative set of measurements.
“Achieving the high availability required to deliver network resilience at the necessary scale without compromising performance as you add controller instances has been an elusive goal for open source SDN solutions and a barrier to adoption—until now,” said Guru Parulkar, Executive Director for ON.Lab. 0″Architected as a distributed system, ONOS is the first open source SDN solution to achieve linear scale-out while maintaining high performance and availability. As the size of your network grows, ONOS instances can be added to scale the SDN control plane, and seamlessly deliver the needed throughput. This ability not only breaks down barriers to real-world deployment but also future-proofs your network.”
A comprehensive explanation of these metrics and Blackbird performance assessment using these metrics is published on the ONOS wiki at http://bit.ly/1GhIr3X
