SDN makes things easier to deal with applications, but it does not negate the need for physical switches and routers to do complex things, says Russell Skingsley of Juniper
BANGALORE, INDIA: Ever since the concept of Software-Defined Networking (SDN) has come up, there has been a lot of debate on whether it will pose a major threat to traditional networking players and even lead to commoditization of their portfolios.
Offering clear explanations to those issues Russell Skingsley, director, Advanced Technology, Cloud and Data Center, APAC, Juniper Networks, says SDNs are a threat to incumbents, whose business model is predicated on the 40-year-old status quo approach to networking. In an interaction with Deepa Damodaran of CIOL, he also says how Juniper's switches and routing gear portfolios are far from being victims of commoditization. Excerpts:
CIOL: What is Juniper's strategy around software-defined networking (SDN) and OpenFlow?
Russell: Juniper is a big supporter of Software-Defined Networking (SDN) and we believe that OpenFlow is only one way of expressing it, albeit an important one.
A few years ago, we announced our strategy for changing the way networking is done with a concept, 'The New Network'. The concept was about making the network a platform rather than just a collection of devices and software configurability.
Juniper’s vision for the New Network and the primary drivers behind SDN are aligned. Legacy networks have grown large and complex, stifling innovation and making them costly to build and maintain. Both the New Network and SDN are about removing complexity, treating the network itself as a platform, and shifting the emphasis from maintaining to innovating.
SDN provides an abstracted, logical view of the network with externalized software-based control and reduced control points for better network control and simplified network operations. Juniper’s vision for SDN includes bi-directional interaction between the network and applications, and a real-time feedback loop to ensure an optimal outcome for all elements and a predictable experience for users.
This capability is transparent, allowing customers to augment their existing network infrastructure to be SDN-enabled.
SDN is still in its formative years. Juniper believes that innovations like SDN require active leadership. We have been engaged with the Open Networking Foundation (ONF), and are a founding member of the Open Networking Research Center (ONRC), a Board member of US Ignite, and we have created our own OpenLab to help drive industry collaboration around The New Network.
CIOL: Do you think that this strategy means that a section of your switches and routing gear portfolios will get commoditized? Why not and how?
Russell: Well we do not, of course! As a network vendor you may think 'they would say that', but the truth is there has always been the threat of commoditization of Ethernet switches and routers and yet it has not eventuated. The fact that a desire for SDN exists at all actually holds the key to why that is the case. SDN in some respect is an attempt to provide an abstraction layer for a fundamentally complex system..
For example, SDN might like to request something like 'provide absolute priority for this voice application. without the need to deal with the switching process underneath. This does not mean, however, that the switch does not need the capability to deliver that priority. The switch will still need to provide priority queuing of the packets even if the SDN layer does not care how it does it.
Now, no matter how much we abstract that functionality, a switch that does not implement a priority queue will not be able to comply with what the software layer is asking for. There is still a need for powerful switching functions and it requires an SDN interface.
In fact, as people start building SDN-enabled applications, I am sure they will request more, not less, functionality over time. Whilst these functions might be described in software simply through abstraction, they may actually be more complex to translate into switch behaviour.
SDN makes things easier to imagine and to deal with for applications, but it does not negate the need for physical switches and routers to do complex things. The value for high quality vendors like ourselves is to abstract that complexity. There will always be a place for such vendors. In fact putting more power into the hands of application developers will only drive the demand for more powerful underlying networks. This is the opposite of commoditization.
CIOL: However, analysts say hardware majors cannot escape commoditization of their switch portfolio with the onset of SDN. If not the core and edge network products, the top-of-rack data center switches will get commoditized. How is this at Juniper?
Russell: Indeed, there is a lively debate underway on whether SDNs will commoditize networking equipment, with many posing valid questions about the extent to which SDNs represent a threat to various networking vendors. Here is the reality: SDNs represent a direct and serious threat to the large incumbents in our industry, whose business model is predicated on the 40-year-old status quo approach to networking.
For Juniper, the dynamics are very different because we constructed our strategy and our portfolio with the purpose of challenging that status quo. If you look closely you see that much of the current focus of SDN lies in the data centre.
We are rapidly gaining momentum in the data centre by using software that has a proven ability to collapse multiple legacy data-center layers into a single, high-performance fabric. The SDN approach and the OpenFlow standard are actually very helpful to Juniper because they will reduce our cost of development, provide us with greater control over other vendors’ hardware, and augment the software-defined approach we are already delivering for customers.
By contrast, this picture looks very different if you are the incumbent in enterprise switching and your main priority is to protect a $10 billion legacy business. This is especially true when that business is predicated on building separate silos of network infrastructure for each different customer function — for example, one network silo for video, another for wireless, another for a campus environment, and so on, with each silo running different proprietary protocols.
It is no secret that this siloed, legacy approach has been very lucrative for the incumbent vendor. It is also not a secret that the standards-based SDN approach poses a grave and serious threat to this decades-old legacy switching franchise.
Read the Second Part of the Interview Here
Why SDN buy-outs do not threaten Juniper