SwiTching And tRansmission (STAR)

The Internet power consumption has continued to increase over the last decade as a result of a bandwidth growth of at least 50 to 100 times. Further bandwidth growth between 40% and 300% is predicted in the next 3 years as a result of the growing popularity of bandwidth intensive applications. Energy efficiency is therefore increasingly becoming a key priority for ICT organizations given the obvious ecological and economic drivers. In this project we adopt the GreenTouch energy saving target of a factor of a 100 for Core Switching and Routing and believe this ambitious target is achievable should the research in this proposal prove successful. A key observation in core networks is that most of the power is consumed in the IP layer while optical transmission and optical switching are power efficient in comparison, hence the inspiration for this project. Therefore we will introduce energy efficient optimum physical network topologies that encourage optical transmission and optical switching at the expense of IP routing whenever possible. Initial studies by the applicants show that physical topology choices in networks have the potential to significantly reduce the power consumption, however network optimization and the consideration of traffic and the opportunities afforded by large, low power photonic switch architectures will lead to further power savings. We will investigate a large, high speed photonic switch architecture in this project, minimize its power consumption and determine optimum network physical topologies that exploit this switch to minimize power consumption. We will design new large photonic switch fabrics, based on hybrid semiconductor optical amplifiers (SOA) / Mach Zehnder interferometers as gating elements to minimise the switching energy per bit, and plan to optimize the network architecture making use of these new switch architectures and introduce (photonic) chip power monitoring to inform higher layer decisions.
Networks are typically over provisioned at present to maintain quality of service. We will study optimum resource allocation to reduce the overprovisioning factor while maintaining the quality of service. Protection is currently provided in networks through the allocation of redundant paths and resources, and for full protection there is a protection route for every working route. We will optimize our networks to minimize power wastage due to protection. The power savings due to optimum physical topology design, optimum resource allocation, optical switching instead of IP routing, more efficient photonic switches and energy efficient protection can be combined and therefore the investigators and their industrial collaborators BT, Alcatel Lucent and Telekomunikacja Polska, believe that an ambitious factor of 100 power saving in core networks can be realised through this project with significant potential for resulting impact on how core photonic networks are designed and implemented.

The development of new network architectures, protocols and hardware to reduce the energy consumption of communication networks is of particular importance currently, and is one of the most exciting areas in the communications field. Optical networks play a crucial role in networking today, however, a new generation of energy-efficient, ubiquitous, high capacity and flexible optical networks is required. The proposed work will contribute to the research community on many fronts, firstly in the design of new energy efficient large photonic switching architectures and their use in optical networks to maximise energy efficiency. Here, design guidelines, and rules for dimensioning the network and its components will be derived. The results are of particular interest to equipment manufacturers and network operators seeking to redesign their network in the most energy efficient manner. Secondly, energy-efficient resource allocation, traffic engineering and routing strategies will be formulated and their performance assessed using analytic (eg. mixed integer linear programming optimisation) and simulation studies. This will enable the network operator to choose the most appropriate strategies for their network. Finally, a study of energy-aware survivability mechanisms will enable the network operator to design the most appropriate energy efficient survivability strategy that maximises the use of optical transmission and optical switching at the expense of IP routing.
As can be seen from the attached letters of support, there is considerable backing for this proposal from the collaborating industrial consortium. Whilst the programme has direct exploitation, we are delighted that there are several areas within the research activity which are at a basic and generic level and require technical breakthroughs. As a result, we believe that the project will be of benefit to the wider networking academic community, and allow the creation of links with other universities in the communications, networking and computing fields, both in the EU and internationally. Since the project will develop new technologies, components, sub-systems, architectures, protocols and algorithms it will also be of considerable interest to industry. For example, the architectures and protocols work will be of benefit to companies generally involved in network design and operation, carriers, Internet service providers, data centres and large ICT users. The hardware developments are of particular interest to network and communication equipment manufacturers, and indeed in wider applications where energy consumption is a major concern. The technical developments will be of particular benefit to the standards bodies (e.g. IEEE), government departments and the green ICT community. The programme will have tremendous economic benefits to the EU and internationally and fundamental transformational ecological benefits to our planet.

We will focus on maintaining and indeed growing links with our existing industrial collaborators as we believe that their markets will continue to grow. For example, Ethernet services are expected to grow substantially, with estimates of sales in the US alone being $27B over the next few years. Some industry analysts are anticipating a tripling of sales within 5 years despite the cost reduction pressures that will occur. The potential for the work in our proposal also to impact the $14B networking field is considerable. The energy reduction measures introduced in this project will have far reaching impact and their potential will be of direct benefit to the collaborating networking and equipment companies and beyond. This project will lead to exploitable concepts suitable for licensing, for example in energy efficient components, where we have an excellent track record in licensing to world leaders such as Avago with components already in production or in software, e.g. Xen or VNC.