Abstract

With the rapid growth in user base, wide adoption of cloud-computing-based services and the enormous growth in the overall bandwidth requirement, the energy consumption in data centers (DCs) and network infrastructure is rapidly increasing. The increase in energy consumption increases the operating expenditure of networks and affects the environment by emitting an enormous quantity of greenhouse gases (GHGs). GHG emission may be reduced by reducing the consumption of non-renewable energy (NRE) [even if promoting the usage of renewable energy (RE)]. In this study, we explore an operational strategy to provision downstream services (also referred to as connections) from DCs to the access segment of the nodes in a core network. To reduce NRE consumption, we allow a connection offered from a DC at one time interval/period of a day to be migrated and offered from another DC at the next time interval (based on the time-varying nature of RE supply at the nodes and the time-varying bandwidth of the connection). However, a connection migration (i.e., service migration) involves overhead (due to signaling) and may disrupt ongoing service. We allow an optimal connection migration that takes into consideration the savings in NRE consumption and the cost (i.e., penalty) to be incurred due to the migration. Even though a connection migration at a given time interval may offer reduced NRE consumption, it may result in prohibitively large end-to-end propagation delay (due to large path length). Thus, we judiciously allow connection migration. We solve the dynamic optimization problem on-line (based on real-time information) using a mixed integer linear programming (MILP)-based optimization model and an auxiliary-matrix-based heuristic. The MILP model offers the optimal solution, whereas the heuristic efficiently provides a solution for practical scenarios with large networks. Through exhaustive simulation study, we show that the proposed heuristic offers significant savings in NRE consumption with reference to the scheme that does not allow migration.

© 2016 Optical Society of America

Designing an Energy-Efficient Cloud Network [Invited]

Burak Kantarci and Hussein T. Mouftah
J. Opt. Commun. Netw. 4(11) B101-B113 (2012)

IP-Over-WDM Network Design Methodology to Improve Efficiency in Overall Expenditure Due to Cost and Energy Consumption

Suman Kr. Dey and Aneek Adhya
J. Opt. Commun. Netw. 7(6) 563-577 (2015)

Delay-Aware Bandwidth Slicing for Service Migration in Mobile Backhaul Networks

Jun Li, Xiaoman Shen, Lei Chen, Jiannan Ou, Lena Wosinska, and Jiajia Chen
J. Opt. Commun. Netw. 11(4) B1-B9 (2019)

Energy-Minimized Design for IP Over WDM Networks

Gangxiang Shen and Rodney S. Tucker
J. Opt. Commun. Netw. 1(1) 176-186 (2009)

Anycast-Based Optimizations for Inter-Data-Center Interconnections [Invited]

Molka Gharbaoui, Barbara Martini, and Piero Castoldi
J. Opt. Commun. Netw. 4(11) B168-B178 (2012)