dc.contributor.author |
El Zarif, Nizar Rabih, |
dc.date.accessioned |
2017-08-30T13:55:24Z |
dc.date.available |
2017-08-30T13:55:24Z |
dc.date.issued |
2014 |
dc.date.submitted |
2014 |
dc.identifier.other |
b18329792 |
dc.identifier.uri |
http://hdl.handle.net/10938/10510 |
dc.description |
Thesis. M.E. American University of Beirut. Department of Electrical and Computer Engineering, 2014. ET:6160 |
dc.description |
Advisor : Dr. Mariette Awad, Assistant Professor, Electrical and Computer Engineering ; Committee Members: Dr. Ayman Kayssi, Professor, Electrical and Computer Engineering ; Dr. Hassan Artail, Professor, Electrical and Computer Engineering. |
dc.description |
Includes bibliographical references (leaves 59-61) |
dc.description.abstract |
The average power consumption per datacenter is around 1 MW per year, making the power consumption of one datacenter equivalent to that of a small town. It is estimated that the state-of-the-art datacenter consumes around 0.8W to cool down 1W of heat generated by a server. Also, the cooling cost makes up 15percent of the total cost of ownership. A survey by J. Koomey indicated that nearly 5000 MW were consumed by datacenters in the US alone in 2005, costing around 2.7 billion dollars in electric bills. Hence, the need for a better power management arises. Most of today’s datacenters use either Least Loaded First or Round Robin scheduling algorithms which result in and large energy consumption. To reduce the energy cost of operating a datacenter, we look into efficiently scheduling the workload among the available servers. Thus, we modeled the workload scheduling problem as a Variable Cost and Size Bin Packing Problem, and introduced two new solutions based on the Best Fit algorithm. The first solution - Divide and Conquer Best Fit– is a modified version of the Best Fit algorithm optimized for multicore processors. The second solution is the Accelerated Best Fit optimized specifically for a Graphical Processing Unit scheduler. Both algorithms solve VCSBPP and reduce the energy consumed in datacenter servers 7000 times faster than BF. This translates in transforming BF from being a mostly offline solution to an online one. |
dc.format.extent |
1 online resource (xvi, 61 leaves) : color illustrations ; 30cm |
dc.language.iso |
eng |
dc.relation.ispartof |
Theses, Dissertations, and Projects |
dc.subject.classification |
ET:006160 |
dc.subject.lcsh |
Cloud computing. |
dc.subject.lcsh |
Parallel programming (Computer science) |
dc.subject.lcsh |
Computer algorithms. |
dc.subject.lcsh |
CUDA (Computer architecture) |
dc.subject.lcsh |
Computer simulation. |
dc.subject.lcsh |
Energy consumption. |
dc.title |
Energy aware scheduler for cloud computing tasks in a datacenter environment - |
dc.type |
Thesis |
dc.contributor.department |
Faculty of Engineering and Architecture. |
dc.contributor.department |
Department of Electrical and Computer Engineering, |
dc.contributor.institution |
American University of Beirut. |