Energy- oriented Internet (The Network)

Continuing with the previous class, Sergio Ricciardi talks about the energy consumption in the Networks. He explains that bandwidth has incremented by 1000 in 10 years and energy consumption by 10. An Optical Cross-Connect node (OXC) with micro-electro-mechanical system (MEMS) switching logic consumes about 1.2 W per single 10 Gb/s capable interface, whereas a traditional IP router requires about 237 W per port, so that’s obvious that power consumption of electronic traffic is higher than optical traffic

.

In order to design an Optical infrastructure, have to be taken into account some things. 3R regeneration should be avoided as much as possible in planning, designing and managing new paths. EDFA are more performing (higher gain, lower insertion loss, noise and crosstalk effects) than SOA but have also higher energy consumption. The use of dispersion compensation fibers will reduce the dispersion of the optical signal traversing the fiber and reduce the number of required optical amplifiers.

The Energy consumption is currently dominated by the access network because of the high number of end-point devices. With rising traffic volume, the major consumption is expected to shift from access to core networks. Energy consumption also grows in backbone networks. In conclusion access networks dominate at low rates and network routers dominate at higher rates (reduce hop count, improve router efficiency (technology), employ energy-aware algorithms & protocols, manage routers better (sleep states), develop better network architectures using and manage distribution and replication of contents).

However, current router architectures are not energy-aware. One method to reduce energy consumption is to be focus on energy-aware architectures that can adapt their behavior, and so, their energy consumption, to the current traffic loads (advocated both by standardization bodies and governmental programs and assumed in many literature sources.  The router power consumption has two parts: a fixed and a variable. The fixed part due for the device to stay on and the variable part are somehow proportional to the traffic load.

In an attempt to reduce the energy consumption, the study of the network energy consumption becomes important. We have different energy models: Analytic (Parameters + Mathematical description of the network; unambiguous formula + abstraction + generalization; modeling difficulty + complexity) Experimental (Energy consumption of real world devices, experimentally measured + inter/extra-polating data, cannot be used for future energy-aware architectures) and Theoretical energy models (Theoretical predictions of the energy consumption as functions of the router size and/or the traffic load, simple and clear, predictions may substantially differ on the long run from the real energy consumption values).

Because of much of the time our systems are idle but on, one solution possible is the sleep mode. There are different possibilities to implement this mode: per-node (Downclocking and Energy proportional computing) and per-interface (Adaptive link rate, Low Power Idle, STOP-START).

Per-Interface energy saving techniques have to had into account that faster interfaces require lower energy per bit than slower interfaces and that there are low utilization periods, but the energy consumption  is throughput-independent. To solve the problem (throughput won’t be equal ) we implement the idea of temporarily switching off or downclocking unloaded interfaces and line cards (per interface sleep mode): Adaptive Link Rate (ALR) and Low Power Idle (LPI). ALR dynamically modify the link rate according to the real traffic needs and LPI has the particularity of transmission on single interface is stopped when there is no data to send and quickly resumed when new packets arrive.

Other solutions are implemented to solve the problem of energy-aware.  The OSPF-TE protocol that wants to minimize the GHG emissions by routing connection requests through green network elements. RWA Algorithm has as an objective to minimize GHC emissions, power consumption and costs.