ONF Technical Advisory Group member and ONF Research Associate Jennifer Rexford reflects on the history of programmable networks, in an ACM Queue article she co-authored with Nick Feamster and Ellen Zegura.
During the past few years, Software Defined Networking (SDN) has taken the world by storm. But, many of the key ideas---such as programmable switches, open interfaces to the data plane, logically-centralized control, and network virtualization---have been "in the air" for the last five, ten, or even twenty years. How have these ideas evolved over time, and what factors led SDN to gain significant traction now where earlier efforts stalled?
My colleagues Nick Feamster and Ellen Zegura, both professors at Georgia Tech, and I traced the intellectual history of programmable networks in a survey article published in ACM Queue. We divide the history into three main threads: research on active networks (from the mid-1990s and early 2000s), work on separating the control and data planes (from around 2001 to 2007), and the OpenFlow® API and network operating systems (from 2007 to around 2010). Network virtualization played an important role throughout this history, ultimately taking root as one of the first significant use cases for SDN, so we also give a brief overview of twenty years of work on network virtualization.
As researchers, our focus was on the intellectual history---the evolution of the ideas over time. But ideas rarely arise in a vacuum. As such, we also discuss the application "pulls" and technology "pushes" that drove each stage of innovation, as well as the prevailing critiques of the ideas at the time. Any attempt to record history is necessarily incomplete, and we are happy to hear comments about topics we missed or important issues we did not capture. Since the ACM Queue article was published, we've already received suggestions that we should have discussed the Ipsilon Switch and the MPOA (MultiProtocol Over ATM) effort at the ATM Forum, as well as the ongoing I2RS efforts at the IETF. We hope you find the article useful in shedding light on an fascinating period in the history of computer networking, and that the lessons from the past can help guide the progress of SDN in the future.
- Jennifer Rexford, Gordon Y. S. Wu Professor of Engineering, Computer Science Department, Princeton University