Network architecture is another factor which influences the power consumption of networks. To achieve energy efficiency in networks, most of the existing studies on energy efficient network architectures are based on data transmission, structural layout design etc. In the following sections, the details of these studies will be discussed.
3.4.1 Energy Efficient Data Transmission
It is known that the power consumption of processing switches increases when the data is switched at a very high bit rate. Most of the current switches work in asynchronous mode, and the buffers in the switches need to operate faster than the interface rate in order to enable multiple functions to operate simultaneously to access packets stored in the same buffer. The basic principle of pipeline forwarding was introduced in [113] along with a process where the real-time packet advances one hop in each time frame. The time frame can be considered as a virtual container of a constant number of IP packets, which is a basic time period for synchronising the IP packet switches with pipeline forwarding. In addition, time frames are grouped into time cycles in order to provide the basic unit for a periodic repetition of the reservation which is used for the reserved transmission capacity. There are two applications of pipeline forwarding: Time-Driven Switching (TDS) and Time-Driven Priority (TDP). Due to its ability to transmit traffic in large volumes and reduce the total power consumption, TDS is suitable for high speed optical backbones network.
In [114], a parallel network which has the same fibre infrastructure as WDM networks, but is based on pipeline forwarding that coexists with asynchronous IP technology was introduced (as shown in Fig 3-3). As well as carrying traditional traffic, the required signalling for setting up synchronous virtual pipes can be transported by asynchronous IP routers, in the pipeline forwarding parallel network. There are two reasons that pipeline forwarding can be used to reduce the power consumption of networks. First, compared to asynchronous routers, pipeline forwarding significantly reduces the hardware complexity of the switching fabric controllers, because the content of each time frame is switched and forwarded according to its position within the time cycle and therefore there is no need for header processing. Consequently, less process cycles are needed to handle each packet. Secondly, compared to the reconfiguration of switching fabrics before moving each single packet in an asynchronous router, the reconfiguration frequency of the switching fabric of TDS switches can be reduced by using the whole time frames as switching units.
Fig 3 - 3: Parallel network on the same fibre infrastructure with WDM [114]
In IP over WDM optical networks, IP routers consume most of the energy. Thus, it makes sense to minimise the required number of IP router ports in data transmission as a method to reduce energy consumption. Optical bypass can be used to reduce the number of IP router ports in lightpaths for data transmission where the data remains in the optical domain until it reaches the destination node, thus bypassing the IP ports at all the intermediate nodes. In [58], optical bypass was first evaluated both via MILP and simulations showing that it reduces energy
consumption in IP over WDM networks. In addition, it is also observed in that work that an energy-minimised optical network is also cost-minimised. Similar to the optical bypass, end-to-end bypass traffic grooming, which was proposed in [115], reduced power consumption of intermediate nodes in lightpaths compared to conventional data transmission architecture.
3.4.2 Energy Efficient Structural Layout
Fig 3 - 4: Node architectures of (a) IP/WDM, (b) IP/SDH/WDM, and (c) GMPLS/ASON [116]
Currently, energy saving for the backbone optical network is not just based on the IP over WDM two-layer architecture. There are several other multi-layer architectures, such as IP over SDH over WDM and OTN (optical transport network) over WDM. The energy consumption of these different types of multi-layer optical networks have been evaluated in several studies recently. In [116], the energy consumption of IP over SDH over WDM was investigated. Compared to the pure IP over WDM optical networks, energy consumption can be saved by introducing the SDH layer. IP routers need to process the data packet by packet and IP layer aggregation in routers results in higher energy consumption. In the underlying SDH and WDM layers, the traffic demands are switched at the SDH layer without IP processing, and electronically bypass the IP router. Thus, energy savings can be made. In [117], similar principles have been extended by including a GMPLS/ASON control plane, where some wavelengths can be switched at the WDM layer; thus optically bypassing all the layers above. This will bring further energy savings, as switches in the WDM layer consumes even less energy than that in the SDH layer. Fig 3-4 gives three different node architectures. In [118], energy efficiency and
CAPEX optimality are studied for IP over OTN over WDM networks via MILP model. Under the specific situation studied, the IP over OTN over WDM has a 30% advantage in energy saving over the IP over WDM architecture [118].
In access networks, energy consumption is dependent on the network structure. Passive optical networks (PONs) consume the smallest energy per transmitted bit, attributed to the proximity of the end user to the exchange and the low power consumption of passive network devices. In PONs, Optical Line Terminal (OLT) and Optical Network Units (ONUs) consume a large portion of the overall energy consumption [119]. Typically, OLT chassis comprise of multiple OLT line cards and each of them communicates with a number of ONUs. All of these OLT line cards in the OLT chassis are usually working in power-on mode to avoid service disruption between ONUs and the central office.To reduce the energy consumption of OLT, a novel energy efficient OLT structure, which adapts the number of power-on OLT line cards in the OLT chassis to the real-time incoming traffic, was proposed in [120]. In [121], the energy consumption of several different PON architecture are compared. Among the considered architectures, WDM-PON based on Reflective Semiconductor Optical Amplifier (RSOA), stacked 10G TDM-PON, and point-to- point fibres offer the lowest power per line potential.
3.4.3 Energy Efficient Caching and Mixed Line Rate
Fig 3 - 5: A cache-based On-Demand Service using IP over WDM [122]
In addition to energy efficient multilayer design in networks, the introduction of caching into current data distribution networks can reduce energy consumption
significantly. In [122], an energy efficient cache-based Video-on-Demand (VoD) service over an IP over WDM network was proposed. Different to a conventional IP over WDM network, the cache-based VoD network architecture shown in Fig 3-5 shows that each node in the network is allocated a limited content caching capacity. In this network, the most popular video content in the video servers are stored by the cache in each node. Video requests can be served by a local node cache, rather than directly from the video server. Compared to the energy consumption associated with routing the video traffic demand in the IP layer all the way from video servers, the cache consumes much less energy. Thus, the energy consumption of the whole network is reduced.
Recently, a popular energy saving architecture design has been proposed that uses a different bit-rate transponder and IP port according to traffic demand. With the support of multi-line-rates, physical ports power consumption can be reduced. In [123], an energy-optimisation framework with mixed line rates has been investigated. By using mixed line rates, energy-efficiency can be improved in three different kinds of optical network architectures: transparent, translucent and opaque, compared to that based on a single line rate. In [124], energy efficiency for IPTV program delivery in optical backbone networks with multiple available line rates was investigated. An energy-efficient flow aggregation was proposed to guarantee minimising the total energy consumption in delivering programs.