Are CSPs ready to make the most of SDN and NFV or is there a big step yet to be taken in properly operating virtualized networks? Spoiler: Yes, yes there is….
The always insightful Tom Nolle of CIMI has blogged about the challenges facing SDN and NFV out in the real-world, serving customers.
It’s well worth a read if you have an interest in the direction OSS needs to go in the next few years.
His article got me thinking, so I’d like to riff on some of his ideas.
A History Lesson
Every network revolution is an operations revolution, or it should be.
The OSS industry has had mixed success in ‘operationalizing’ communications technology. A long time ago there were useful OSS applications for managing plain old telephony and other simple PSDN/copper based services. Those applications, along with physical, workforce processes were ‘good enough’ to operate a telephone network.
In the 1990s/2000s a new breed of more flexible OSS emerged to manage PDH, SDH, ATM, DSL and mobile. It did very well. In the best cases it was able to automate service fulfilment and assurance tasks to enable large volumes of new services, mainly broadband, to be sold. This was achieved using what I call inventory-centric fulfilment: Process automation and resource allocation based around a definitive database of the network and customer services. You know, like Cramer, Granite and NetCracker.
In the 2000s/2010s CSPs turned to IP-based networks to better handle all that data sloshing around. To make IP manageable, hardware vendors offered stuff like MPLS to tame IP. CSPs now had very flexible and fluid networks. But, as Tom notes:
IP convergence broke the 1:1 notion of services/networks because you now had… generalized infrastructure that handled basic connectivity… and a set of service silos that imposed the features per service.
In the IP world service fulfilment is no longer about nailing up a certain amount of capacity across a fixed route. OSS has struggled with this for some years trying to shift its focus to service management while leaving the complexity of the IP network resources to management and planning tools.
Keeping up with the Tech
NFV and SDN only make matters more complex. Not only are service routes dynamic, but the network they run over is (virtually) dynamic. SDN can change the topology and routing policies in seconds. NFV can upgrade, downgrade or relocate device functions without the procurement, planning and engineering cycles of today’s network operations.
The implication is that inventory-centric fulfilment becomes even less relevant if neither the inventory nor the service suits the static resource allocation model any more.
The solution… is no more obvious today than the answer to the question of how CSPs should make best use of SDN and NFV.
Tom suggests:
So there are multiple levels of orchestration implicit in next-gen virtualization-based networks.
For sure. This is what inventory centric fulfilment is missing. Its fixed layers and fixed resource allocation rules need to be abstracted and dynamic.
But that’s the easy bit. The real challenge is in getting OSS to move at the speed of SDN and NFV.
Both in terms of…
- Decision making – Can an OSS application make a decision and configuration change in the sub-second time-scales demanded by SDN? And…
- Business process – Can OSS be usable, secure and precise enough to allow changes to virtualized resources in hours or minutes rather than todays planning cycles of weeks or months?
Yes, the success of tomorrow’s OSS hinges on being able to address CSP business processes, data quality, and system integration issues.
Some things never change.
Finally, thanks to Tom Nolle for his original article and for introducing me to the term Operationalization!