SDN without Analytics is like Taking Reservations

Without analytics, SDN is only automated, not intelligent. So far, the promise of SDN has been realized in its utility regarding automation. The argument goes like this: by separating the control and routing planes and implementing network changes via a centralized controller, enabling and disabling services becomes much faster.

This, as far as I know, is true but it’s a little bit like the Seinfeld car reservation scene below. “You know how to take the reservation, you just don’t know how to hold the reservation and that’s really the most important part of the reservation, the holding. Anybody can just take them.”

Today’s SDN knows how to take the reservation. But because it lacks analytics, SDN doesn’t know how to hold and satisfy the reservation request. Today’s SDN isn’t aware of inventory beyond those elements provisioned by the SDN controller, doesn’t know current usage and/or state, and doesn’t comprehend what will happen if it tries to fulfill the reservation.

SDN without analytics delivers automated networking but falls woefully short of intelligent networking.

To be clear, there is immense business value (e.g. opex savings) in automating network changes. If one person can do ten times as much with SDN as without then companies will need fewer people to fulfill service requests, freeing them to work on tasks that cannot be automated. However, automating changes without knowing whether the network can support or fulfill the request creates tremendous risk. For example, when user-demanded services are provisioned over links and paths which are already at or near saturation points the automated service will impact existing customers negatively. Without analytics, SDN becomes an extremely efficient way to fail.

And as much value as exists in automating network changes, imagine how much more value can be created with intelligent SDN analytics. In such a world, the state of the network will be used as a reference point to determine the best (i.e. most cost-efficient or with the least delay) routes when new services are requested. Service providers will have new sources of revenue as they deliver bandwidth on demand services. Such services will be purchased, provisioned and torn down (a) at the touch of a button and (b) without interrupting or impacting service of existing customers. Service providers will have ways to delay, defer, or cancel unneeded capital expenditures as they optimize traffic across existing infrastructure.

But all this presupposes an analytics-enabled SDN solution. Short of this solution, SDN misses the potential of being dynamic, responsive, and flexible. This point seems to be [at least part of] what Craig Mathias is describing in his article “SDN analytics offers key to smarter, more adaptive networks.” Unless and until SDN incorporates intelligence into the framework, it will be able to only take reservations. Only by understanding the current state of a network can service providers hold reservations. Only by implementing analytics-enabled SDN can service providers realize the full potential of SDN as a fast and efficient tool to satisfy emerging market demands.