Network Virtualization and Energy Efficiency University of Passau Gergö Lovász, Andreas Fischer, and Hermann de Meer Outline 1. Power Consumption of ICT 2. Economic Principle and Energy Efficiency Benchmarks 3. System Virtualization 4. Network Virtualization 5. Energy-efficient Protocols and Services 6. Conclusions and Future Works 2
Power Consumpiton of ICT > Dramatically increasing power consumption of ICT 5 billion kwh in 2000 vs. estimated 55 billion kwh in 2010 (Germany) > Energy costs exceed hardware acquisition costs > ICT senvironmental limpact is directly linked to its energy consumption CO2 emission caused by ICT = CO2 emissions caused by aviation Energy-efficient HW alone is not sufficient to stop or even to reverse this development 3 Economic Principle and Energy Efficiency Benchmarks > Economic Principle Maximization principle (fixed input, maximum output) Minimization principle (fixed output, minimum input) > Change in paradigm Replacing the maximization principle by the minimization principle fixed benefit (e.g. a determined QoS) is made available by using a minimal set of resources (hardware and energy) > Measurement of energy efficiency through energy efficiency benchmarks measurement of performance per watt instead of performance alone major energy efficiency benchmarks: JouleSort (based on external sort) Applicable for a wide range of computer systems Energy optimizations from a whole-system perspective SPECpower_ssj2008 (java-based) Measures energy efficiency of volume server class and multinode class computers Uses different loads: 100% to 0% in 10% steps 4
System Virtualization > System Virtualization virtualizes an entire computer system A virtual machine monitor (VMM) provides virtual machines on top of a real machine Virtual machines (VMs) provide virtual CPUs, virtual RAM, virtual Harddisks, > Very prominent in data centers today VM VM VM VMM Real Machine Consolidation of virtualized services leads to lower hardware cost and higher utilization of existing hardware Energy-efficiency achieved by turning off unused hardware 5 Network Virtualization > Extending System Virtualization to the network Virtual router Software with routing functionality Possibly multiple instances on real machines Virtual link Logical interconnection of two VRs Properties can change dynamically Virtual network Virtual routers connected by virtual links Phys. Link > How to achieve energy-efficiency in the network? Virtual Links Consolidation of routers possible, too but what are effects? 6
Energy-efficient Protocols and Services > Energy-efficient communication protocols needed Synchronization of communication Enabling of dormant cycles Delegation of services and functions From energy consuming devices to energy efficient devices On demand mechanisms Get rid of heart-beats, power on network device on demand Energy efficient signaling Out of band signaling on separated energy efficient paths > SOA-based application-layer services with energy properties Accounting based on power consumption 7 Conclusions and Future Work > Virtualization of networks is a major enabler to achieve an energy-efficient Future Internet Consolidation of virtual routers possible Savings similar to those seen today in data centers should be possible > Detailed implications still have to be investigated Protocols need to be adapted Stability and scalability of virtual networks need to be investigated 8
References Wuppertal Institut für Klima, Umwelt, Energie GmbH, 2000. Koomey. Estimating total power consumption by servers in the us and the world, Technical report, 2007, http://enterprise.amd.com/downloads/svrpwrusecompletefinal.pdf Fraunhofer ISI. Der Einfluss moderner Gerätegenerationen der IuK-Technik auf den Energieverbrauch in Deutschland bis zum Jahr 2010, Studie. http://www.isi.fhg.de/e/publikation/iuk/ Fraunhofer-IuK- Kurzfassung.pdf Rivoire, Shah, Ranganathan, Kozyrakis. JouleSort: A Balanced Energy-Efficiency Benchmark. ACM SIGMOD International Conference on Management of Data (SIGMOD), 2007 Hans Albert. Ökonomische Ideologie und politische Theorie. Verlag Otto Schwartz & Co, Göttingen 2. Aufl. 1972. pp.13ff. Koomey et al. Server energy measurement protocol, Version 1.0, Following Energy Efficiency Server Benchmark Technical Workshop, Santa Clara, CA, 2006, http://www.energystar.gov/ia/products /downloads/finalserverenergyprotocol-v1.pdf Bundesverband Informationswirtschaft, Telekommunikation und neue Medien e.v., Energieeffizienz im Rechenzentrum Band 2, 2008, p. 10 A. Berl, A. Fischer, and H. de Meer. Using System Virtualization to Create Virtualized Networks. Workshops der Wissenschaftlichen Konferenz Kommunikation in Verteilten Systemen (WowKiVS2009), Kassel, Germany, March 2-6, 2009. vol. 17, EASST, 2009. 9