At last week’s Big Telecom Event in Chicago, Caroline Chappell, senior analyst with Heavy Reading (the sister organization of Light Reading) moderated a panel discussion on SDN/NFV Management and Orchestration. Readers of this blog will know that’s a subject near and dear to us at Packet Design, and Cengiz Alaettinoglu, our CTO, was a member of the panel. He was joined by speakers from Infoblox, Overture, UBIqube, and NTT America.
Ms. Chappell opened the discussion by posing the question, “How real is SDN/NFV, and how quickly will network operators move from proofs of concept to production deployments?” In keeping with what we had heard in earlier conference keynotes and panel sessions, Doug Junkins with NTT America, the only operator on this panel, stated they are already implementing SDN/NFV in 50 data centers globally. He said the primary business drivers are: (1) driving down the cost of provisioning services and (2) new services creation. When asked how NTT views SDN in the context of NFV, he explained that they use the term network automation in place of SDN. Network automation enables customer self-provisioning and real-time provisioning—very much in support of NFV and mentioned in the last blog post.
We agree that automation is a key component, but SDN is much more. Orchestration technology enables intelligent automation—creating an infrastructure that can adapt autonomously to changing demands. Cengiz explained that this network programmability introduces three management challenges:
- Speed of network changes: When network configurations can change in seconds, it is imperative that management systems stay in sync. Polling the infrastructure periodically (every 15 or 30 minutes, say) is clearly inadequate, and polling more frequently creates unacceptable overhead. Traditional SNMP management technologies are marginalized.
- Visibility into changes made programmatically: Today, when a network change is required, humans are involved in the planning, modeling and configuration. With SDN, however, these changes occur without operator involvement. When things are running OK, that’s fine, but when there is a problem (and there will be), where do humans go to understand the current network state, identify what changes were made, and troubleshoot problems?
- Impact to existing services of requested programmatic changes: SDN allows application requests for network resources to be provisioned programmatically. What governs these actions? How does the network know if the requested changes are a good idea? What will be the impact on existing services? What if historical traffic profiles indicate that the network load will increase in ten minutes? Should the changes still be made or will that create contention?
These challenges and how they can be addressed are discussed in more detail in a white paper, authored by Cengiz.
Echoing these comments and others made in earlier sessions by speakers from Deutsche Telekom, Google and Time Warner, Mr. Junkins said that the biggest challenge NTT faces is the way management is done—that managing static inventory through OSS systems has to change to accommodate dynamic environments. There are big concerns: Maintaining FCAPS, new processes, tools, training, SLAs, etc. present challenges, but NTT’s operations group sees long-term opportunity and is embracing the change.
Clearly, the operational impact of SDN/NFV is significant and will require operators to rethink their management systems, retool and retrain. But the baby won’t be thrown out with the bath water, as too much has been invested in OSS systems. So these systems will have to adapt and yes, new management technologies will be integrated with the legacy ones. In discussing the expected migration path to SDN/NFV, the panelists agreed that new management technologies will be delivered in chunks, driven by customer demands, including open APIs. The good news is that the speakers expected little impact to existing application performance management (APM) tools, which sit above the network infrastructure. These tools will work with and provide input to the SDN/NFV orchestration system.
Network operators will not accept lesser functionality, so their management tools – existing and new – must meet that requirement. According to NTT, management vendors must involve operations in product definition and design processes. At the same time, operators must adapt from managing “boxes with blinking lights” to working with virtual machine user interfaces.
These are interesting times.