Research Article
An Energy Saving Mechanism Based on Vacation Queuing Theory in Data Center Networks
@INPROCEEDINGS{10.1007/978-3-319-90775-8_16, author={Emna Baccour and Ala Gouissem and Sebti Foufou and Ridha Hamila and Zahir Tari and Albert Zomaya}, title={An Energy Saving Mechanism Based on Vacation Queuing Theory in Data Center Networks}, proceedings={Mobile Networks and Management. 9th International Conference, MONAMI 2017, Melbourne, Australia, December 13-15, 2017, Proceedings}, proceedings_a={MONAMI}, year={2018}, month={5}, keywords={Power consumption Vacation queuing theory Data center networks}, doi={10.1007/978-3-319-90775-8_16} }- Emna Baccour
Ala Gouissem
Sebti Foufou
Ridha Hamila
Zahir Tari
Albert Zomaya
Year: 2018
An Energy Saving Mechanism Based on Vacation Queuing Theory in Data Center Networks
MONAMI
Springer
DOI: 10.1007/978-3-319-90775-8_16
Abstract
To satisfy the growing need for computing resources, data centers consume a huge amount of power which raises serious concerns regarding the scale of the energy consumption and wastage. One of the important reasons for such energy wastage relates to the redundancies. Redundancies are defined as the backup routing paths and unneeded active ports implemented for the sake of load balancing and fault tolerance. The energy loss may also be caused by the random nature of incoming packets forcing nodes to stay powered on all the times to await for incoming tasks. This paper proposes a re-architecturing of network devices to address energy wastage issue by consolidating the traffic arriving from different interfaces into fewer ports and turning off the idle ones. This paper also proposes to attribute sleeping and active periods to the processing ports to prevent them from remaining active waiting for random arrivals. Finally, we use the vacation queuing theory to model the packets arriving process and calculate the expectation of vacation periods and the energy saved. Then, we strengthen our work with a simulation part that validates the analytical derivations and shows that the proposed mechanism can reduce more than 25% of the energy consumption.