Somewhat ironically, a significant number of smaller dense wave-division multiplexing (DWDM) equipment vendors — sellers of optical transport gear — could easily find themselves moving in a software-defined networking (SDN) direction more aggressively than larger suppliers of such gear in the long term. The folks at the bigger companies have already spent a good deal of money on MPLS technology, and replacing the entire control plane with SDN overnight may not be so appealing in terms of getting an adequate return on investment. More importantly, for small WDM companies, systems-engineering work with SDN has been found to be much simpler and superior to working with the control plane of MPLS.
For DWDM gear vendors, the MPLS data plane is not relevant, while the SDN control plane promises less complexity. For data-center customers, offers improved control-plane and application-plane functionality as well as a higher level of interoperability that will allow for a true multivendor network equipment environment.
The lower-end optical firms have to be more practical in looking at the big picture, as they cannot afford to concentrate their R&D dollars on every emerging solution. Skipping straight to the next generation often has the advantage of encompassing the interim approach more efficiently, such as using 40 Gb/s as just a service, and not as an uplink, in going right to 100 Gb/s.
It has been demonstrated that a DWDM vendor’s network management system (NMS) can accommodate SDN functionality to provide A-to-Z routing of wavelengths between any nodes. Of course, the major problem is that SDN is now standardized only around switches.
SDN does not seem to be practical unless there will be complete standards for the DWDM layer as well as sufficient third-party devices that can be managed, which meet those requirements. Thus, DWDM firms will likely have their own NMSs or proprietary SDN variants for at least the next two to three years.
Certainly, the use of a single vendor, such as Cisco Systems, which can provide the same management system from Layers 1 through 3, is an option for some customers. However, one supplier cannot provide the myriad elements that are necessary in the implementation of data centers — involving SANs, WANs, switches, routers, load balancers, optical devices, etc. — and it is in this sector where SDN will ultimately make a dramatic difference in the movement to a cloud-based, virtualized infrastructure environment.
There will be numerous cases in U.S. and Western European enterprises in which a 100-Gb/s core router will be purchased with the need to interconnect data centers. A pizza box will often be preferable over a 10U-to-12U rack because only single or dual ports of 100 Gb/s will be necessary to tie the centers together. At the same time, if the manufacturer is not sufficiently far enough out on the learning curve on SDN, it is hardly inconceivable that it could lose out on winning business in this space well down the road.
In looking at the history of standardization in the industry, when it is all said and done, clearly not all of the promises of SDN will be fulfilled. On the other hand, there is room for a large amount of optimism far into the future, because SDN is arguably a relabeling of the direction the industry has been heading toward for quite a while anyway in terms of programmability of networks.