HANDBOOK ON GREEN INFORMATION AND COMMUNICATION SYSTEMS

1. HANDBOOK ON GREEN INFORMATION AND COMMUNICATION SYSTEMS Chapter 22: Green Optical Core networks Mohammad Ali Mohseni and Dr. Akbar Ghaffarpour Rahbar Sahand University of Technology, Tabriz, IRAN Sahand University of Technology, IRAN

2. Outlines • Introduction • Energy consumption in optical core networks • Network layers overview • Power consumption models • Energy efficiency metrics • Energy consumption of network equipment: Switches, Amplifiers, Transmitter, Receivers • Energy efficient optical network design • Modular structure of switch • Static and dynamic node structures • Single line and multi-line rate network • Optimum repeater spacing • Energy-aware optical core networks • Power-aware routing and wavelength assignment • Power-aware grooming • Selectively switching sleep network elements • Power-aware survivable networks • conclusion Sahand University of Technology, IRAN

3. Introduction • Green attribute refers to component, system, or network which consume energy efficiently • Why greening the networks? • Growing the number of users and services leads to increase power consumption of ICT • Decreasing cost • Lack of energy resources • Environmental effects: decreasing emission of green house gas in the air Sahand University of Technology, IRAN

4. Introduction • Green research approaches • Energy efficient approach: • A network is energy efficient if its power consumption is minimized without degrading its performance. Energy efficiency should be considered when designing a network • Energy awareness approach: • Networks are usually designed for peak traffic load; however network traffic varies during time and in wide margin time is less than peak. • A network is energy aware if its power consumption change according to changing offered traffic load Sahand University of Technology, IRAN

5. Energy consumption in optical core networks • Network layers Sahand University of Technology, IRAN

6. Power consumption metrics • There are few metrics for evaluating energy efficiency of network equipments. Most simple and common metric is ratio of total power consumption to capacity of system • Higher value means the system is more energy-hungry • If system is a switch then the system capacity is summation of bit rates of line cards, this ratio estimates a lower bound for energy consuming per bit for a typical switch • In practice, a system does not work at high capacity because traffic changes during time. So it is more acceptable to replace equipment capacity with equipment throughput Sahand University of Technology, IRAN

7. Power consumption models • Power consumption models express power consumption of equipment versus offered traffic load: • Linear power consumption model • Theoretical power consumption model • Combined power consumption model • Statistical power consumption model Sahand University of Technology, IRAN

8. Energy consumption of switches • Switches are classified in two distinct classes: • Linear analog switch • Digital switch • Input and output interfaces can involve optical demultiplexers/multiplexers, transponders, O/E or E/O converters, wavelength converters components • control unit configures switch fabric for an input wavelength adjust to appropriate output if switch fabric is active ones Pswitch=Pinput+Pswitch_fabric+Poutput+Pcontrol_unit Sahand University of Technology, IRAN

9. Energy consumption of optical amplifier • Optical amplifiers • Power of optical signal is attenuated while passing through  optical fibers. • Attenuation increases exponentially by length of fibers • Attenuation should be compensated by optical amplifiers or repeaters • Amplifiers perform re-amplifying, re-timing and reshaping signal in all optical domain so optical signals remain in optical domain during passing through links Sahand University of Technology, IRAN

10. Energy consumption of optical transmitter and receiver Optical transmitter Optical receiver

11. Energy efficient network design • A network is energy efficient if its power consumption is minimized without degradation of its performance. Energy efficiency should be considered when a network is designed • Modular switches • Static and dynamics network • Single rate and multi rate network • Optimum repeater spacing Sahand University of Technology, IRAN

12. Modular structure of switches • A modular switch has several chassis and each chassis has some line cards • Each line card has several ports for establishing connections with other nodes • Modular structure provides possibility of load-less components to go to sleep mode for power efficiency purposes Sahand University of Technology, IRAN

13. Energy efficient network design • Static and dynamic network • In static network, allocation of wavelengths is performed statically and do not change during time • In dynamic network, allocation may change during time based on traffic volume and network topology • Dynamic operation needs lower number of wavelengths than static due to wavelength conversion possibility Static classic optical node(SCON) Static low consumption optical node(SLON) Dynamic optical node(DON) Sahand University of Technology, IRAN

14. Static and dynamic network • Two parameters determine which structure is more energy efficient • ratio of energy consumption of LR (Long Reach transponder) and SR (Short Reach) transponders • ratio of power consumption of a device in sleep state and active state Sahand University of Technology, IRAN

15. Single rate and multi line rate network • Current optical backbone networks supports 10/40 Gbps data rates and 100 Gbps data rate is in way • In SLR (Single Line Rate) network equipments operate at the same bit rate • In MLR (Multi Line Rate) network wavelength channels have variety of capacities (10/40/100) • Optical impairments limit the usage of high data rates (40 and 100 Gbps) to short distances Sahand University of Technology, IRAN

16. Single rate and multi line rate network • There are some facts which determine power consumption of network which should be considered: • More number of wavelengths needs more number of transponders • Long distance requires more amplifiers • BER threshold limits the distance on which a lightpath with data rate r on wavelength λ can reach • Low data rates need more fibers to support high capacity requests Sahand University of Technology, IRAN

17. Optimum repeater spacing • Optical signals are attenuated while passed through fiber • Amplifiers are applied to compensate attenuation • Amplified spontaneous emission(ASE) noise • when an optical signal is amplified by an EDFA, ASE noise is added to original signals • For long fiber, it is required to use electronic repeaters Sahand University of Technology, IRAN

18. Optimum repeater spacing • Optimum repeater spacing lopt is optimum distance between repeaters which minimizes power consumption of link L Sahand University of Technology, IRAN

19. Power aware optical core network • Networks are usually designed for peak traffic load • However, network traffic varies during time and is less than peak in wide margin time. • A network is energy aware if power consumption of network changes according to changing offered traffic load • Power aware routing and wavelength assignment • End to end versus link by link traffic grooming • Selectively switching sleep network elements • Power aware survivable WDM network Sahand University of Technology, IRAN

20. Power aware routing and wavelength assignment(PA-RWA) • RWA establishes a minimum cost route between source and destination nodes and assigns an appropriate wavelength to the established path • The objectives of conventional RWAs are to minimize routing cost and blocking probability • Route cost : path length, hop count, and etc. • PA-RWA aims to adjust network power consumption with offered traffic load • PA-RWA is modeled as an ILP formulation. • Since there are usually many constraints for an ILP problem, these problems are classified in NP-hard problems • Heuristics • Least Cost Path (LCP) • Most Used Path (MUP) • Ordered-Lightpath Most Used Path (OLMUP) Sahand University of Technology, IRAN

21. Weighted power aware routing and wavelength assignment (WPA-RWA) • PA-RWAs decrease the power consumption of network but also decrease network performance • PA-RWAs set cost of links as link power consumption so routing algorithm may select minimum power consumption path which is not shortest path. • Selecting long paths decreases some QoS parameters such as BER and delay • Weighted PA-RWA (WPA-RWA) shows the impact of energy minimization on path length and blocking probability Sahand University of Technology, IRAN

22. Traffic grooming • Traffic grooming is a traffic engineering approach, i.e., process of aggregating small traffic in bigger units • Traffic grooming is performed by three strategies: • End to end traffic strategy • Link by link traffic strategy • Hybrid strategy • In hybrid strategy, a lightpath partially can be aggregated with previously established lightpaths or partially established new routes. Intermediate nodes of a new route operate in bypassing strategy. Sahand University of Technology, IRAN

23. Traffic grooming • End to end strategy • Pend_to_end=PDS+PEO+POS+PTX+Pamp+PRX+ POS+ PTX +Pamp+ PRX+ POS +POE + PDS Sahand University of Technology, IRAN

24. Traffic grooming • Link by link grooming • Plink_by_link=PDS+PEO+POS+PTX+Pamp+PRX+ POS+POE + PDS +PEO+POS+PTX+Pamp+ PRX+POS+ POE + PDS Sahand University of Technology, IRAN

25. Traffic grooming • Traffic grooming • Evaluation of the power consumption over 5 years for both the link-by-link and end to end grooming scenario, following a traffic increase of 50% per year Sahand University of Technology, IRAN

26. Traffic grooming • Traffic grooming is modeled as an ILP problem. • Like other hard problems, it is solved by heuristics: • Path based heuristic • Link based heuristic • Request size based • Link utilization based Sahand University of Technology, IRAN

27. Traffic grooming • Differentiate power saving traffic grooming • Grooming degrades some QoS parameters such as delay and BER because it may route some light-paths in longer routes than minimum available routes. • Some users not eager to use unproven technologies: • Solution: divide demands by two classes: • Red class: Red class demands are served as traditional routing algorithm such as shortest path objective function • Green class: Green class demands are routed with energy saving objectives • Differentiate power saving • Red demand base heuristic • Total demand base heuristic Sahand University of Technology, IRAN

28. Selectively switching sleep network elements • Network consumption is more than required because: • Over provisioning • Traffic is time varying • Protection resources are always active • Solution: • Switching off or sleep additional network equipments • Selecting the elements that should go to sleep mode • Rerouting • Reactivation Sahand University of Technology, IRAN

29. Selectively switching sleep network elements • Switching off network elements • Which network elements go to sleep mode? • Load-less • Load-less than threshold • Most power • Least flow • Random • Least link (node degree is least) • ILP • Rerouting: • Reactive rerouting • Proactive rerouting • Reactivation Sahand University of Technology, IRAN

30. Survivability • Survivability is an essential necessity to protect links and nodes against failures. Protection can be provisioned at different levels: node, link, and path • Path protection schemes: • 1:1 protection • 1+1 protection • 1:N protection • M:N protection • Protection resources of networks, always are active due to rapidly recovering of failures. These resources consume energy, whereas unused for a long time. • Solution is switching exclusively protection resources to sleep mode. Sahand University of Technology, IRAN

31. Energy aware shared path protection • In conventional networks, load balancing improves performance parameters because it distributes traffic over network resources and utilizes them as equal as possible. • Load balancing reduces congestion in links and nodes, and also it decreases requirement protection resources • Traffic aggregating increases requirement protection resources Sahand University of Technology, IRAN

32. Energy aware path protection • Aggregating primary and secondary resources separately can lead to more energy conservation Sahand University of Technology, IRAN

33. Energy aware dedicated path protection • Separating working and secondary paths increases power saving and also blocking probability since resources are divided for different purposes, working and protection. • Energy aware dedicated path protection with differentiation of primary and secondary path(EA-DPP-Dif) • Initially, cost of link is set based on required power for activating amplifiers • After establishing the first lightpath over a link, other lightpaths can be established with lower cost because links are turned on now. Path cost degradation is performed for aggregating lightpaths on turned on links. • EA-DPP-Dif considers two different penalties: one for differencing between primary and secondary paths and another for aggregating primary and secondary paths separately Primary path provisioning phase Protection path provisioning phase Sahand University of Technology, IRAN

34. Energy aware dedicated path protection • Energy aware dedicated path protection with mixing secondary path with primary path(EA-DPP-MixS) • Initially, cost of link is set based on required power for activating amplifiers. • After establishing the first lightpath over a link, other lightpaths can be established with lower cost because links are turned on now. Path cost degradation is performed for aggregating lightpaths on turned on links. • EA-DPP-MixS permits mixing primary and secondary paths, especially when routing the secondary path Primary path provisioning phase Protection path provisioning phase Sahand University of Technology, IRAN

35. Conclusion • Energy conserving in ICT is inevitable due to • Current networks are not energy efficient • Rapid growth of networks infrastructures • High cost: economical, environmental, … • Lack of energy resources • Guaranty of current increasing growth of these networks • Solutions • Develop networks energy efficiently • Modular structure of switch, static and dynamic node structures, single line and multi-line rate network, optimum repeater spacing • Networks adjust their energy consumption by traffic loads (energy aware networks) • PA-RWA, Power-aware grooming, selectively switching network elements to sleep mode, power-aware survivable networks Sahand university of technology, IRAN