Chris Wright, VP and Chief Technologist at Red Hat, and a Board Member of OPNFV and OpenDaylight discusses NFV, use cases, and open source at OPNFV Summit.
What is your long-term vision for NFV?
NFV in the long run is a full network transformation: shifting from proprietary, function-specific hardware to an open, flexible infrastructure hosting network services and applications that will drive a new wave of digital innovation. Central offices become data centers, and the edge provides low-latency, high-bandwidth access to tremendous computing power. Trends like IoT, based on simple devices, are a beginning. More exotic futures – with self-driving cars and augmented reality – will all be built from the hierarchical distribution of compute, network, and storage at each successive network hop from a device.
How have you been using or viewing NFV? What are your key target use cases/PoCs?
I have been viewing NFV as a huge opportunity for cloud adoption and a good technology demonstration showing the power of SDN. It’s a different use case for a cloud and a great way to show the innovative power of the open-source development model. Our key target use cases and PoCs are focused on virtualizing network functions such as vCPE, vEPC, and vIMS. The vRAN use case is following, with some new challenges in terms of predictable latencies and real-time-type requirements.
What is the key role for open-source from your perspective?
Open-source is a development model that allows an industry to quickly collaborate on core, non-differentiated functionality. This creates a form of de facto standardization that often outpaces the traditional standards model – because the implementation is rapidly iterated on, and the duplicate effort inherent in black-box interoperability can be eliminated. It starts with a form of commoditization and can lead to remarkable innovation. I see open-source giving service providers the flexibility to build their networks from best-of-breed components.
What hurdles remain for service providers in order to attain the ideal network?
The software needs to continue to grow necessary functionality to allow for the complete network transformation that service providers are looking for. We’ve focused a lot on raw performance – next is management and operations. We have to convince the network operations teams that this is a safe technology transformation. At the same time, this is a cultural and business model shift. We can’t underestimate the challenge that people present. We see this in the IT shift to DevOps, and we’ll have similar issues here on the network side. Change is hard, even when it’s rife with opportunity.
What is your perspective on the evolving relationships among standards, open-source, and vendors?
Standards are incredibly useful for building interoperability across independent implementations of the same functionality. The network simply wouldn’t work without basic protocols like TCP/IP. This is especially true when you are building functionality while paced with a standards body. Open-source can implement standards, but it can also drive ahead and work with standards bodies to define trailing standards. This works well during bursts of innovation, where the typical, slow standardization process can stifle innovation. Ultimately, I believe we need to combine the iterative open-source development process with a feedback loop to standards orgs. We are striving to do this between OPNFV and ETSI, for example. The process of industry collaborating to build a standard, vendors building to that standard, and customers requesting that standard – it’s simply too slow. The power of open-source is that customers and vendors can collaborate directly in building the software.
What new technology or trend in the networking space are you most curious about and why?
I love to watch how hardware and software evolve to provide network functionality. White-box switches are one step forward. At the same time, modern servers are capable of providing more and more network functionality. The line between network device and server continues to blur, and I expect that we are just at the beginning.