The Linux operating system provides many well-developed tools that support the concept of energy-efficient networking. This paper outlines the results of research focused on the design and implementation of a control system reducing power consumption of IP-traffic processing in a network of Linux-based software routers. It is demonstrated how the standard ACPI-compliant system components can be adjusted to meet the requirements of adaptive energy-aware network control and what performance tradeoffs may there be expected. In particular, it is demonstrated how the abstraction layers provided by the Linux kernel can be used to exploit energy-saving mechanisms of packet processing servers. Formulations of the routing optimization and service rate control problems are presented and discussed. The results of the extensive experimental studies are presented as well.
1 March 2016 in International Journal of Applied Mathematics and Computer Science
The issue of energy-aware traffic engineering has become prominent in telecommunications industry in the last years. This paper presents a two-criteria network optimization problem, in which routing and bandwidth allocation are determined jointly, so as to minimize the amount of energy consumed by a telecommunication infrastructure and to satisfy given demands represented by a traffic matrix. A scalarization of the criteria is proposed and the choice of model parameters is discussed in detail. The model of power dissipation as a function of carried traffic in a typical software router is introduced. Then the problem is expressed in a form suitable for the mixed integer quadratic programming (MIQP) solver. The paper is concluded with a set of small, illustrative computational examples. Computed solutions are implemented in a testbed to validate the accuracy of energy consumption models and the correctness of the proposed traffic engineering algorithm.
Energy-efficient traffic engineering for wired network poses a number of difficulties due to equipment and protocols diversity and complexity. If emerging optimization are to be formulated exactly, they are intractable. This is why numerous heuristics for approximate solutions are proposed. They differ in assumptions on the routing scheme in the network, and also on the availability (and usefulness) of traffic-related or measurement-based power consumption data. Those most interesting w.r.t. authors' work are presented and classified. The authors propose a number of strategies for switching the links on or off, which are still worthy consideration - while the routing scheme stays in control of some shortest-distance mechanism, like OSPF. Operation of the algorithms is verified for two standard test cases, resulting in proposed further improvements and hints on implementation details.