1 DENS: Data Center Energy-Efficient Network-Aware SchedulingDec 20, 2010 DENS: Data Center Energy-Efficient Network-Aware Scheduling Dzmitry Kliazovich University of Luxembourg Pascal Bouvry University of Luxembourg Samee Ullah Khan North Dakota State University

2 Why energy is important?Increased computing demand Data centers are rapidly growing Consume 10 to 100 times more energy per square foot than a typical office building Energy cost dynamics Energy accounts for 10% of data center operational expenses (OPEX) and can rise to 50% in the next few years Accompanying cooling system costs $2-$5 million per year December 20, 2010 Pascal Bouvry

3 Distribution of data center energy consumptionDecember 20, 2010 Pascal Bouvry

4 Data center architecturesThree-tier data center architecture Most Widely Used Nowadays Access, Aggregation, and Core layers Scales to over 10,000 servers December 20, 2010 Pascal Bouvry

5 Data center componentsServers’ Energy Model memory modules, disks, I/O resources CPU Idle server consumes about 66% of the peak load for all CPU frequencies December 20, 2010 Pascal Bouvry

6 Data center componentsSwitches Most common Top-of-Rack (ToR) switches typically operate at Layer-2 interconnecting gigabit links in the access network Aggregation and core networks host Layer-3 switches operating at 10 GE (or 100 GE) Links Transceivers’ power consumption depends on the quality of signal transmission in cables and is proportional to their cost 1 GE links consume 0.4W for 100 meter transmissions over twisted pair 10 GE links consume 1W for 300 meter transmission over optical fiber Supported power management modes DVFS, DNS, or both December 20, 2010 Pascal Bouvry

7 Pascal Bouvry ([email protected])Simulator components Switches’ Energy Model Chassis ~ 36% Linecards ~ 53% Port transceivers ~ 11% December 20, 2010 Pascal Bouvry

8 Pascal Bouvry ([email protected])DENS methodology DENS achieves balance between Energy consumed by the data center Individual job performances Job QoS requirements Data center traffic demands DENS is architecture specific Data Center Architecture December 20, 2010 Pascal Bouvry

9 Pascal Bouvry ([email protected])DENS methodology Avoid Overloading Computing server selection Penalize Under-loaded Servers Favor High Server Utilization December 20, 2010 Pascal Bouvry

10 Pascal Bouvry ([email protected])DENS methodology Computing server selection Favor Low Congestion Levels Penalize Congested Queues December 20, 2010 Pascal Bouvry

11 Pascal Bouvry ([email protected])DENS methodology Computing server selection Maximize Server Load, Minimizing Network Congestion December 20, 2010 Pascal Bouvry

12 Performance EvaluationGreenCloud simulator is developed Three-tier data center topology 1536 nodes, 32 racks, 4 core and 8 aggregation switches December 20, 2010 Pascal Bouvry

13 Performance EvaluationServer workload distribution Green scheduler Redistributing Computing Load DENS scheduler Round-robin scheduler December 20, 2010 Pascal Bouvry

14 Performance EvaluationNetwork workload distribution Network congestion Redistributing Load to Satisfy QoS Green scheduler DENS scheduler Round-robin scheduler December 20, 2010 Pascal Bouvry

15 Performance EvaluationTop-of-Rack (ToR) switch load Network congestion Green scheduler Green scheduler DENS scheduler DENS scheduler December 20, 2010 Pascal Bouvry

16 Performance EvaluationEnergy QoS Data center energy consumption Parameter Power Consumption (kW·h) Round Robin scheduler Green Scheduler DENS Data center Servers Network switches 417.5K 353.7K 63.8K 203.3K (48%) 161.8K (45%) 41.5K (65%) 212.1K (50%) 168.2K (47%) 43.9K (68%) December 20, 2010 Pascal Bouvry

17 Pascal Bouvry ([email protected])Conclusions We acknowledge Funding form Luxembourg FNR in the framework of GreenIT project Research fellowship provided by the European Research Consortium for Informatics and Mathematics (ERCIM) Related publications "GreenCloud: A Packet-level Simulator of Energy-aware Cloud Computing Data Centers," in Journal of Supercomputing, special issue on Green Networks, 2011. “GreenCloud: A Packet-level Simulator of Energy-aware Cloud Computing Data Centers,” IEEE Global Communications Conference (GLOBECOM), Miami, FL, USA, December 2010. GreenCloud December 20, 2010 Pascal Bouvry

18 Pascal Bouvry ([email protected])Thank you! December 20, 2010 Pascal Bouvry