ega d de E 2014
Accepted 3 November 2014
Available online 10 November 2014
Energy efficiency ty an energy efficiency, when network links face low traffic situations or even active/passive (active/active configuration) is nowadays crucial to corpoancing and redundancy between two nodes through the use of a discrete number of single links. energy efficiency extensions whilst providing bot e LAG to s wake them up individually (sleeping capabilities) another one is to adapt the operation rate of the l the LAG as a function of each link traffic load (rate a tion capabilities). In both cases, the energy consumption derived from the link aggregation could be reduced.
For the former case, the IEEE 802.3az standard (EEE,
Energy Efficient Ethernet) could be applied . This http://dx.doi.org/10.1016/j.adhoc.2014.11.005 1570-8705/ 2014 Elsevier B.V. All rights reserved. ⇑ Corresponding author.
E-mail addresses: email@example.com (J. Galán-Jiménez), alfonso.gazo@cs. ox.ac.uk (A. Gazo-Cervero).
Ad Hoc Networks 25 (2015) 595–605
Contents lists available at ScienceDirect
Ad Hoc Ne journal homepage: www.elsrate and data center networks . To this extent, link aggregation techniques are used to provide active load balavailability and load balancing features.
One approach is to put the links of thh high leep or whilst inks in dapta-It is well known that reduction of energy consumption of network equipment is an open issue in wired networking research  due to the environmental impact of producing energy associated with the operation of current
ICT equipment . Moreover, provisioning of high availability (active/passive configuration) and load balancing amount of energy consumed by the active links in the LAGs (Link Aggregation Groups) when traffic load is low or in absence of failure.
In the last few years, several works have been proposed to improve energy efficiency in single links . By exploiting the use of these techniques, LAGs could incorporateLink aggregation groups
Equivalent energy function 1. Introductionconfigurations, because they still consume almost the same amount of energy as if they were operating at their full rate. Most of the network equipment currently deployed is not yet able to take advantage of saving energy in scenarios where network links are not used at their full capacity. By exploiting the use of existing energy efficient techniques for single links, this work presents a study in which an impact is quantified in terms of energy saving opportunities when link aggregation techniques are in use. Optimal link aggregation configurations are determined and energy savings are assessed using real network traffic datasets. Results show that adding links capable of using sleep mode to link aggregation groups is a more energy-efficient technique compared to using less links in link aggregation groups, even if they had more advanced energy-saving features like rate adaptation capabilities. 2014 Elsevier B.V. All rights reserved.
However, these methods can be considered as inefficient from the energy efficiency point of view due to theReceived 6 February 2014
Received in revised form 22 September data centers networks. Network link aggregation techniques are quite commonly used to provide those features. However, these techniques often come with a cost in terms ofDesigning energy-efficient link aggr
J. Galán-Jiménez a,⇑, A. Gazo-Cervero b aComputer Science and Communications Engineering Department, Universida bDepartment of Computer Science, University of Oxford, United Kingdom a r t i c l e i n f o
Article history: a b s t r a c t
Link high availabilition groups xtremadura, Spain d traffic load balancing features are key requirements in today’s tworks evier .com/locate /adhoc of these two devices fails. To prevent this possible situaproposed to reach the optimal solution. As energy savings achieved by the three methods are similar, authors encourage network operators to use the simplest one.
Differently from , the work of  propose a method to decide how to act individually over a particular link depending on its utilization, i.e. by putting it to sleep or waking it up. A fixed link utilization threshold (90% during simulations) is used to determine the link to activate or switch off at every moment (one single link each time).
With data from real traces collected during one month of observation, simulation results achieve up to 86% of energy savings for a threshold of 90%, although there is a low risk of traffic overflow.
In order to improve their method and reduce the risk of traffic overflow, the authors propose another approach, which considers both the link utilization threshold and the number of active links in the LAG (1–3) . In this 596 J. Galán-Jiménez, A. Gazo-Cervero / Ad Hoc Networks 25 (2015) 595–605tion, different link aggregation techniques focused on device-level redundancy are also used. Several protocols, such as VRRP (Virtual Router Redundancy Protocol)  or STP (Spanning Tree Protocol), defined in 802.1D standard  have been proposed. However, network equipment vendors have recently shown interest in incorporating alternative mechanisms to solve some of the problems identified when using these techniques, such as the problem of potential reduction of available bandwidth derived from blocked ports by STP. IETF TRILL protocol (Transparent Interconnection of Lots of Links)  and
SPB (Shortest Path Bridging) technology, specified in IEEE 802.1aq standard  are two examples which have already been deployed by most of the network equipment vendors despite their youth.
Fig. 1 shows the relationship between the different features provided by link aggregation techniques and the protocols developed in this scope. Four features have beenstandard is based on the action over sending and idle intervals to be able to connect or disconnect some of the components of a single network interface. For the latter case, the hardware-based Rate Adaptation method could be used . With this technique, each single link in the LAG can be configured with a particular discrete number of energy levels .