August 22, 2014 | Pankaj Berde†, Matteo Gerola†, Jonathan Hart†, Yuta Higuchi§, Masayoshi Kobayashi§, Toshio Koide§, Bob Lantz†, Brian O'Connor†, Pavlin Radoslavov†, William Snow†, Guru Parulkar†
The paper presents the development and evaluation of ONOS (Open Network Operating System), an experimental distributed SDN control platform designed to meet the performance, scalability, and availability requirements of large operator networks. The authors describe two prototypes: the first focused on implementing a global network view, scale-out, and fault tolerance, while the second aimed to improve performance, particularly event latency. Key features of ONOS include a distributed architecture, a global network view, and fault tolerance. The evaluation of the prototypes highlights performance issues such as low latency, data model inefficiencies, and excessive data store operations. The second prototype introduced improvements, including a RAMCloud data store, an optimized data model, a topology cache, and an event notification system. These changes significantly reduced latency and improved performance in basic network state changes, reaction to network events, and path installation. The paper also discusses related work and future plans, including the development of use cases, abstractions, resource isolation, and security enhancements, with the goal of releasing a usable open-source ONOS system by the end of 2014.The paper presents the development and evaluation of ONOS (Open Network Operating System), an experimental distributed SDN control platform designed to meet the performance, scalability, and availability requirements of large operator networks. The authors describe two prototypes: the first focused on implementing a global network view, scale-out, and fault tolerance, while the second aimed to improve performance, particularly event latency. Key features of ONOS include a distributed architecture, a global network view, and fault tolerance. The evaluation of the prototypes highlights performance issues such as low latency, data model inefficiencies, and excessive data store operations. The second prototype introduced improvements, including a RAMCloud data store, an optimized data model, a topology cache, and an event notification system. These changes significantly reduced latency and improved performance in basic network state changes, reaction to network events, and path installation. The paper also discusses related work and future plans, including the development of use cases, abstractions, resource isolation, and security enhancements, with the goal of releasing a usable open-source ONOS system by the end of 2014.