Phil will be speaking at Carrier Network Virtualization 2014 on Wednesday, December 10 at Crowne Plaza Palo Alto. Click here to download the full agenda.
What would you say has had the largest impact in terms of progressing SDN and NFV in carrier networks to date?
I would say that AT&T’s “Domain 2.0” announcement had a very significant impact on the progress of SDN and NFV in carrier networks as it has signaled that the big telco players are seriously seeking innovative approaches to network management and operation. That being said – the actual developments in SDN still lag behind and attempt to implement a Data-Centre solution on a Carrier network, which obviously operates in a different environment. Now it is time for SDN to seek an innovative approach and adapt the earlier, Data-Center centric, models to innovative models that match the Carrier environments. NFV, on the other hand, is progressing nicely and is already implemented in certain parts of the network.
Do you feel that the industry is making as much progress towards implementing SDN & NFV this year as you would like?
SDN adoption lags behind in the carrier networks due to its inapplicability to the Carrier network environment. Now it is time for SDN to seek an innovative approach and adapt the earlier, Data-Center centric, models to innovative models that match the Carrier environments: Long latency, Lack of diversity, High cost of transmission, High port-density, Reliability. NFV, on the other hand, is progressing nicely and is already implemented in certain parts of the network.
How is your company contributing to the realisation of SDN & NFV at present?
PCCW Global is an active participant in several industry SDOs such as MEF, ONF, ETSI-NFV. Through our participation we voice the need to make SDN more carrier-friendly, more agnostic to protocols and more process orientated. We trust standardized APIs are key to allow SDN to evolve from a Data-Centre application into Network-Programmability framework.
What do you think are the biggest hurdles towards implementing SDN & NFV at present and how can they be overcome?
SDN, at present, is still focused on device management, debating OpenFlow vs. NetConf/YANG and others. With a strong drive from white-label HW manufacturers. This model does not address the needs of Carrier networks. Not because it is a bad model, but because it is based on assumptions that were good for the environment where SDN and NFV were conceived: The Data-Centre. In the DC distances (and latency) are short, bandwidth cost is negligible, diversity is easy to implement and resiliency is obtained through stocks of failover devices readily available.
Carrier Networks operate in an entirely different environment where distances are vast, latency is high, diversity is difficult (and expensive) to establish, bandwidth is a significant element of cost, port-densities are very high and device resiliency is a key factor. Thus – carriers tend to stick with their incumbent hardware suppliers that provide devices that have proven track records of survivability, built in resiliency, high port densities and RISC processors that are designed for the sole purpose of pushing packets and frames from one port to another.
SDN must evolve from a Device/Controller paradigm into a Platform/Controller paradigm, and must become agnostic to the SouthBound interface through which the Controller interfaces with the Platform – Be that OpenFlow or NetConf or YANG or others. This will also require an evolution from managing single devices, one by one, towards managing an entire platform or an entire network. The SDN model will need to evolve from the initial three-plane model (Application-Control-Data) into a multi-plane model that includes parts of the device-OS (provided by the HW manufacturer), an Abstraction layer (that represents the vendors-specific hardware and transport resources as abstracted and vendor agnostic resources), a resource catalogue layer (that represents the network resources to higher layers) a Virtualization layer (that assembles network resources into virtual service instances) and a controller that is closer to being an OSS than the current SDN controllers in the market.
NFV is actually progressing nicely. Its applicability in a carrier network is limited because Carrier network gear typically revolves around switching and pushing packets from one port to another. Storage, Applications and processing (other than packet forwarding) are absent from the core of the carrier network. On the edge of the network, however, and at the customer premises, NFV already plays a nice role: Smart CPEs on one side (acting as a mix of router/firewall/VoIP switch) and CPE-elimination on the other side (migrating the functionality of the CPE to a VNF on the Provider-Edge).
Which areas of the network do you think should be prioritised and why?
Our pain-point is Network-Programmability, and that is where we want to see progress. NFV on the CPE is in the right direction and we would want to see more applications there, hopefully standardized.
What are your expectations for the development of SDN and Virtualization in carrier networks in the coming years?
The industry, as a whole, is shifting its efforts towards automation. Managing networks manually is getting out of hand, and new applications and the growth in use of mobile data create a need for on-demand managed services. SDN and NFV are the primary enablers of network automation as they allow configuration and activation of services through software rather than through manual processes. We expect to see SDN evolve to become orchestrated, application aware, Situation-aware, agnostic to SB interface type and capable of interfacing with neighbour controllers for the realization of multi-platform services. This requires shift in efforts of the SDOs from writing code (be that open-source or proprietary) towards defining information models and specifying APIs. It then becomes the role of the hardware and software vendors to adapt their systems to the standardized information models and create APIs per specs.
Which companies do you most admire in the SDN space?
PCCW Global has identified tail-f as an important player that can bridge legacy and innovative platforms. We were not surprised to see it being acquired by a major player, and we think the HW vendors that have initiated a software line of products (e.g. Cisco with tail-f and their own home-grown projects, ALU with Nuage, Ericsson with Telcordia) are worth looking at, as the SDN space as we know it will need to shift closer to the OSS, and the OSS will need to interface closely with the network gear. In addition to that I will also watch companies developing orchestration solutions, especially those who offer expertise in process analysis and information modeling, as they may become critical elements in the migration of carriers from traditional, legacy, network environments towards programmability.