An optimization approach for balancing maintenance costs and electricity consumption in Cloud Data Centers
Martedì 12 novembre 2019 ore 14.00
Sala Consiglio (8° piano)
Dipartimento di Informatica "Giovanni degli Antoni" (Via Celoria 18 - Milano)
Prof. Fabio D'Andreagiovanni (Université de Technologie de Compiègne / CNRS, France)
Responsabile: Prof. Roberto Cordone
Cloud Data Centers (CDCs) are data centers that adopt the cloud computing paradigm, according to which applications are virtualized and run in a number of distributed physical servers that may be located very distant from each other, even in distinct continents. Studies on the optimal management of CDCs have commonly focused just on minimizing the total power consumption, deciding how to switch on and off the physical servers composing the CDCs depending on the workload. However, switching the servers causes large temperature transitions in the hardware, which can sensibly increase the failure rates of components and lead to an increase in the maintenance costs.
In this work, we propose a new optimization model for managing CDCs that jointly minimizes the power consumption and the maintenance costs, derived through a material-based fatigue model that expresses the costs incurred to repair the hardware, as a consequence of the variation over time of the server power states. A major objective has been to investigate what is the impact of the maintenance costs on the total costs and whether it is beneficial to leverage the tradeoff between electricity consumption and maintenance costs. Computational results, obtained over a set of scenarios from a real CDC, show that the original heuristic that we propose to solve the optimization model largely outperforms two benchmark algorithms, which instead either target the load balancing or the energy consumption of the servers.
L. Chiaraviglio, F. D’Andreagiovanni R. Lancellotti, M. Shojafar, N. Blefari-Melazzi, C.Canali, "An Approach to Balance Maintenance Costs and Electricity Consumption in Cloud Data Centers", IEEE Transactions on Sustainable Computing 3(4), 2018, DOI: 10.1109/TSUSC.2018.2838338