Application/network interconnection is motivated by the observation that network information flows are about to become as important as data information flows. From a business point of view, application/network interconnection based on innovative network designs is expected to result in a win-win situation for all stakeholders. Universal media access with sufficient Quality of Experience (QoE) is expected to be a key incentive in this direction.
Markus Fiedler, (Blekinge Institute of Technology, Sweden & FP7-Euro-NF),
Georgios Gardikis (NCSR "Demokritos", Greece & FP7-ALICANTE),
Martin Waldburger (University of Zurich, Switzerland, FP7-SESERV)
The session was organized in two parts; it opened with brief statement presentations by the five invited panelists, followed by a panel discussion with the involvement of the audience.
Kurt Tutschku (University of Vienna, FP7-Euro-NF) addressed the problem of finding new ways to interconnect applications and networks, mentioning the example of load balancing mechanisms required for handling high-popularity video streaming. He raised the question whether the whole layering architecture has to be reconsidered to allow information flows between applications and the networks. He stated that network information flows may become as important as data information flows.
Frank den Hartog (TNO, FP7-FIGARO) focused on home networks, which can be more easily managed and monitored so as to allow the deployment of network-aware applications. He argued that making applications network-aware through exposure of monitoring information is an efficient and realistic way to optimise their transport over heterogeneous network environments.
Ioanna Papafili (Athens University of Economics and Business, FP7-SmoothIT) focused on optimization of peer-to-peer traffic via Economic Traffic Management (ETM) approaches. She stated that, from a business point of view, joint application/network optimisation should result in win-win situations through market equilibria. She explained that the envisioned win-win situation was shown for a number of ETM trials in operational networks and real end-users.
Csaba Szabó (Budapest University of Technology and Economics, DIGITAL CITY) addressed specifically telecom operators and their view on network design for the case of community networks. Telecom operators are currently mainly restricted in providing plain connectivity services to customers. They typically adopt a top-down, general-purpose network design, which is possible in principle and might deliver new desirable insights and technologies. However, he states, community networks have proven to be successful if they are designed and built with an application-centric (bottom-up) approach. He anticipates telecom operators to be conservative in adopting a bottom-up network design.
Christian Timmerer (Alpen-Adria Universität Klagenfurt, FP7-ALICANTE) discusses scalable adaptive media delivery systems, especially over HTTP, in which Quality-of-Experience (QoE) is the most critical metric. Universal Media Access with sufficient QoE could be the key incentive for deploying network-assisted service delivery mechanisms.
A key issue, as raised by the audience, is how service differentiation and prioritization can be compatible with network neutrality principles. However, while avoiding application degrading, censorship and violation of privacy, some kind of application-aware traffic handling is essential so as to escape from the current “flat” transport paradigm. If we adopt the absolute definition that network neutrality requires the total isolation between the application and networking domains, then this approach is significantly inefficient and certainly counterproductive when it comes to innovation. In any case, many existing architectures in the commercial domain, like Content Delivery Networks (CDNs) cannot be characterized as “neutral” in the strict sense.
Moreover, it was stated that the classic layering approach is not adequate anymore, since the flexibility to adapt to needs is required and that different mechanisms for different types of content are desirable. Programmable flow-based mechanisms such as Openflow are contributing towards fulfilling these requirements. It was furthermore pointed out that cross-domain aspects are extremely important, especially when it comes to Quality-of-Service (QoS) provision and application/network awareness.
Several questions and comments were also directed towards economic aspects. For example, how network intelligence can be paid for and how telecom operators can monetize on novel application/network coupling approaches. Security aspects on payments were raised, along with how revenues could be distributed across several operators in multi-domain communication. It seems that a convincing model has not been found yet, which could push operators to upgrade their infrastructure and adopt new technologies.
Furthermore, the gap between research efforts and commercial deployments was discussed, i.e. the question how research results in this specific field can find their way to the market. Two principles were suggested; first to design mechanisms in an evolutionary manner (non-disruptive) and, second, to start with deployments in restricted environments, such as in home networks or within single domains. It was commented that even old mechanisms, such as IP multicast, while used in single domains, have not survived in multi-domain, Internet-wide scenarios. A sensitive issue for telecom operators is also the degree of exposure of network-aware interfaces, i.e. how much freedom is given to end-users and applications to actually interact with the functionalities of the network.
Last but not least, user experience was once more identified as a key driver towards network innovations. It seems that QoE is a very complex factor, and it sometimes includes aspects beyond service consumption such as the contribution of user-generated content. Many users are willing to pay more, even abandon the flat-rate model, to enjoy a better experience, and this is a need towards which networking innovations and novel business models should converge.