SPEAKER: Pedro A. Aranda Gutiérrez, TELEFONICA I+D
NFV and SDN are promising technologies for networking. However, their real potential is currently overlooked. As a matter of fact, many traditional networking guys are prone to see Network Functions as the “open” version of their legacy middle-boxes and the SDN controller as the delightful replacement of their patch panels. NFV/SDN bring far more opportunities than deploying network services while comfortably sitting at the own desk!
The fine-grained control over packet forwarding enabled by SDN is very useful to optimize service graphs in NFV and to compose applications and services. NFV strongly relies on the concept of recursive composition, starting from elementary software components, to VNFs and Network Services. The notion of application composition goes farer than mere service chaining, envisioning the possibility to perform in parallel several tasks (e.g., firewalling, natting, monitoring), to assemble the result and to apply it to the packet.
From another perspective, SDN is enabling DevOps development cycles which benefit from prior experiences in software development, such as Agile development. The standard software cyclic lifecycle made of design, implementation, test, deployment could be easily adopted for network services and applications, with great benefits in terms of release times and development costs.
Finally, another issue in NFV/SDN converge concerns packaging. NFV is already defining its packaging formats (e.g., OpenMANO, TOSCA), while SDN applications are usually shipped in the controller-specific format (e.g., jar file, python script). There is still a missing gap in bringing these technologies together and converging to a common and overarching standard.
Going further, the last remark is for current description languages. For example, current way of describing Service Graphs in NFV-MANO is human-readable, but.. is it human-understandable? “Intent” is the magic word in this context; it indicates the possibility to specify what is expected from the network, rather than how to configure the underlying infrastructure. Current efforts in such direction are still at the early stage, and with several drawbacks (e.g., FlowObjectives, Boulder, NEMO).
The final message is a warning to avoid being “seduced” by the buzz recurrent words in NFV/SDN: these are very powerful concepts, though the implementation of their full potential is very challenging and many aspects still need investigation.