A new power profiling method and power scaling mechanism for energy-aware NetFPGA gigabit router

Abstract

Today the ICT industry accounts for 2–4% of the worldwide carbon emissions that are estimated to double in a business-as-usual scenario by 2020. A remarkable part of the large energy volume consumed in the Internet today is due to the over-provisioning of network resources such as routers, switches and links to meet the stringent requirements on reliability. Therefore, performance and energy issues are important factors in designing gigabit routers for future networks. However, the design and prototyping of energy-efficient routers is challenging because of multiple reasons, such as the lack of power measurements from live networks and a good understanding of how the energy consumption varies under different traffic loads and switch/router configuration settings. Moreover, the exact energy saving level gained by adopting different energy-efficient techniques in different hardware prototypes is often poorly known. In this article, we first propose a measurement framework that is able to quantify and profile the detailed energy consumption of sub-components in the NetFPGA OpenFlow switch. We then propose a new power-scaling algorithm that can adapt the operational clock frequencies as well as the corresponding energy consumption of the FPGA core and the Ethernet ports to the actual traffic load. We propose a new energy profiling method, which allows studying the detailed power performance of network devices. Results show that our energy efficient solution obtains higher level of energy efficiency compared to some existing approaches as the upper and lower bounds of power consumption of the NetFPGA Openflow switch are proved to be 30% lower than ones of the commercial HP Enterprise switch. Moreover, the new switch architecture can save up to 97% of dynamic power consumption of the FPGA chip at lowest frequency mode.

Introduction

Nowadays networked applications such as cloud computing, social networks as well as multimedia services have become ubiquitous. The deployment of broadband networking infrastructure in the core and data centers at the edge of the networks has become very common in recent years. Consequently, the ICT industry accounts for 2–4% of the worldwide carbon emissions that are estimated to double in a business-as-usual scenario by 2020 [1]. However, until recently, ICT has not sufficiently applied energy-efficient concepts, even in fast growing sectors like telecommunications and the Internet. Backbone networks nowadays are specifically designed to be extremely over dimensioned in terms of switching capacity and number of deployed links and nodes in order to guarantee zero-loss and minimum latency packet forwarding [2], [3]. When taking a closer look at the network components, the energy consumption of current network devices is substantially flat and almost independent of their actual workloads. Thus, the large volume of energy consumed in the network infrastructure results in high operational expenses and environmental pollution.

For these reasons the concept of green networking has become increasingly important and received considerable attention from research and industrial communities. It focuses on different research issues, including designing more energy efficient technologies inside physical network components, deploying energy-efficient methods in network architecture and control, energy-aware traffic engineering and routing and so forth. Among these methods, one research direction is to re-design the physical hardware architecture of routers with energy-efficient concepts, as switches and routers consume most energy in the Internet [4], [5]. However, making such kinds of network devices more energy-efficient is challenging because of several difficulties:

• Firstly, the detailed energy consumption in different sub-components of a router under different traffic load and network configuration settings is usually not well understood. This is because of the lack of analytical tools and power measurement methods in live networks.

• Secondly, it is difficult to deploy different energy-aware concept in a real network device, as commercial routers do not offer sufficient flexibility and mechanisms for prototyping and implementing novel energy-optimized network algorithms.

• Although there are several proposed strategies for energy-aware switches and routers recently, such as dynamic adaptation of the link rates [6], [7] or smart sleeping [2], [8], the exact energy saving by adopting these techniques is poorly known. The reason is that the energy-saving level depends very much on specific hardware implementation and the type of routers (number of ports, the type of CPU used, etc.).

In this research we try to answer one important issue in the design of energy-aware router architectures, which is how to optimize the energy volume consumed by a router proportionally to its actual traffic load. As illustrated in Fig. 1, in the ideal case, the energy consumption of a network device should be proportional to its traffic load. That is, energy consumption in a low utilization scenario should be much lower than in case of high traffic utilization. However, in the Internet today, redundant or lightly utilized links and devices usually consume nearly as much energy as devices that are switching large volumes of traffic do [9].

In this article, we first analyze the power performance of sub-components in the NetFPGA-based OpenFlow gigabit router documented in [10]. We then propose a new power-scaling method for the FPGA that remarkably reduces the energy consumption, and propose a new method for detailed power profiling of network devices. The main contributions of our work are the followings:

• A measurement framework for quantifying and profiling the detailed energy¹ consumption of sub-components in the NetFPGA OpenFlow switch. Based on the detailed energy profile, the behavior of the switch in terms of energy consumption under various traffic utilization scenarios and network configuration settings has been studied and well understood.

• A new power scaling mechanism fully embedded on the FPGA chip which is able to adapt the operations of the FPGA core and the link capacities of each Ethernet port individually to the real-time traffic load. Although the deployment of the new mechanism is not done on a commercial router, our purpose is to prove that the energy performance of a network device can greatly be improved by re-optimizing the existing hardware resources.

• An analytic model based on the detailed energy profile, which calculates the bounds on minimum, maximum and average energy consumption of network devices under different traffic utilization. The model can be used as a framework to evaluate the energy performance of a router.

• Based on the proposed power profiling method, we study the power profile of the new developed energy-aware OpenFlow switch and compare its energy efficiency with some commercial energy aware switches. Analytical and measurement results show that our concept obtains higher level of energy efficiency compared to some existing approaches [11], [12].

The rest of the article is organized as follows. In Section 2, current work related to energy-efficient networking as well as current approaches for energy-aware gigabit routers is discussed. We focus on the measurement framework for profiling the energy consumption of the NetFPGA in Section 3. The detailed energy profile of the NetFPGA gigabit router is also given in this section. Section 4 proposes a new power scaling mechanism for the NetFPGA router and presents the corresponding energy consumption of the router when operating in different modes. Section 5 shows the performance evaluation of the new developed energy-aware functionalities. Conclusions are drawn in Section 6.