Will there be a market for SDN network operating systems? The answer will be critical to the evolution of SDN standards and the development of applications that leverage SDN capabilities.
The SDN industry is split on this question. Many SDN proponents and vendors, including the Open Networking Foundation, Big Switch, ADARA and Vyatta (Brocade), believe the network operating system is a unique (and critical) component of the SDN architecture. Other leading network and IT providers, such as Cisco and VMware, disagree, saying their SDN and network virtualization offerings do not have a distinct network operating system.
The rise in popularity of software defined networks (SDN) has led to a resurgence of the use of the term "network operating system" or NOS. From an IT history perspective, the initial period of NOS popularity was in the 1990s when Novell's Netware, Banyan's VINES and Microsoft's LAN Manager battled for market leadership. These NOSes ran on standard server hardware and enabled PCs combined with local area networks (Ethernet) to become distributed systems running client/server applications (e.g., email, file sharing, directories, security, database and network management).
The NOS circa 2013 is typically described as the software (middleware) layer that provides the platform for SDN applications. It provides abstraction of the transport layer and network virtualization services, offering north-bound APIs that enable applications and orchestration systems to program the network and request services from it. The SDN NOS is synonymous with the control layer in most SDN architectural diagrams.
A classic definition of a SDN NOS is software (sold separately from switching or routing hardware) that runs on standard server hardware (e.g., x86) and offers APIs as a platform for SDN applications. Current examples of a NOS include OpenFlow with enhancements (e.g., from Big Switch, NEC, IBM, HP, etc.) and software from startups such as ADARA, LineRate, Midokura and Brocade/Vyatta.
Network incumbents such as Cisco with IOS/ONE and Juniper with JUNOS plus its SDN software, can be a considered a NOS, but are dependent on the underlying network infrastructure. The SDN market is in its very early stages and there is wide variety in the capabilities of these SDN "NOS" product offerings. One of the few topics that Cisco and VMware agree on with regards to SDN is that there is NO such thing as a SDN network operating system!
The key question is whether there will be an independent market for SDN NOSes along the lines of the traditional network OS a la Novell Netware. Or whether the capabilities of the SDN NOS will be incorporated in a defined SDN control layer that is tightly tied to a specific cloud or network hardware architecture. This question will impact the emergence of SDN standards (e.g., northbound APIs) and ability of ISVs to develop SDN applications. Ultimately, the marketplace will decide the benefits (and drawbacks) of the various SDN NOS technologies and whether or not the SDN NOS market will emerge as a significant component of the overall networking market.
Lee Doyle is principal analyst at Doyle Research. Doyle Research provides targeted analysis on the Evolution of Intelligent Networks: SDN, OPEX and COTS. Lee Doyle has over 28 years experience analyzing the IT, network and telecom markets. For more information please see doyle-research.com and email me at firstname.lastname@example.org.