Toward Sustainable Networking

Figure 1: Estimation of expected total annual energy  
consumption per IT industry in the period 2010–2030.

The plethora of data generated by scientific applications, the Internet of Things, social media, and e-commerce fuel large-scale data analytics systems. As a result, data transfer over the Internet has been increasing each year exponentially and has already exceeded the zettabyte scale. With the increased data generation rate, the data movement’s carbon footprint is becoming an overwhelmingly critical problem, especially for data centers and wired access networks. It is estimated that information and communication technologies will use between 8% - 21% of the world’s electricity by 2030. The estimation of expected total annual energy consumption per different IT industries in the period 2010–2030 is shown in Figure 1. The share of data centers and communication networks in the total IT power consumption is 69%. Among this share, the data transfers alone consume over a hundred terawatt-hours of energy with a price tag of 20 Billion US dollars annually. Moreover, the environmental side-effect is monumental that information and communication technologies will be responsible for a staggering 14% carbon emission by 2040. This trend has motivated a considerable amount of work in reducing the energy consumption at the hardware, software, and networking infrastructure.

Figure 2: Estimated network energy 
consumption per communication sector.

Despite the advances in networking technologies, sending data over networks is still very costly in terms of energy consumption. The researchers found that sending hard drives between collaborating institutions would be many orders of magnitude less carbon-emitting than transferring the data over communication networks. And, due to increased energy costs pushed by constantly increasing traffic volumes, current network energy costs of telecommunication service operators in developing countries already span between 40% and 50% of provider operational expenditures. In Figure 2, network energy consumption is broken down into six main sectors, and it is shown that the data centers are responsible for 47.3% of network energy power consumption. 

State of the Art

The predominant focus in current research on power-aware networking revolves around mitigating power consumption in networking infrastructure (i.e., routers, switches, and hubs). Various strategies have been proposed to address this, such as putting idle sub-components (i.e., line cards, etc.) to sleep, adapting the rate at which switches forward packets depending on the traffic, putting the Ethernet cards to low power mode when there is no network traffic, development of architectures with programmable switches, switching layers that can incorporate different policies, and power-aware network protocols for energy efficiency in network routing.

The current approaches are hindered by several limitations: (1) Cost-effectiveness is a concern, particularly when considering solutions that involve replacing all switches with energy-efficient alternatives. (2) Short-term practicality poses an obstacle, especially when contemplating solutions like replacing TCP with a more energy-efficient version, which may not be feasible in the immediate future. (3) Balancing performance with energy efficiency is challenging; for instance, implementing measures such as putting certain components to sleep during periods of inactivity can lead to performance penalties despite some gains in energy efficiency.

Our Efforts in this Area

Our group has several ongoing projects in this area: GreenDataFlow, GreenNFV, and GreenABR. These are novel application-layer solutions, which are low cost, very easy and practical to deploy, and do not penalize the performance while increasing energy efficiency.

GreenDataFlow project reduces energy consumption during end-to-end data transfers. It achieves this through a novel two-phase application-layer solution based on offline analysis and real-time dynamic tuning of the transfer parameters. During the offline analysis phase, it analyzes historical transfer logs to perform knowledge discovery about the characteristics of past transfers with similar requirements. During the online phase, it uses the discovered knowledge from the offline analysis along with real-time investigation of the network condition to optimize the protocol parameters for both minimal energy consumption and maximum transfer throughput. GreenDataFlow also provides several service level agreements (SLAs) which give the service providers and the consumers the ability to fine tune their goals and priorities in this optimization process. 

GreenNFV project optimizes resource usage for network function (NF) chains using deep reinforcement learning. It focuses on the concern that Network Function Virtualization (NFV) platforms consume significant energy, introducing high operational costs in edge and data centers. GreenNFV translates the resource scheduling problem into a deep deterministic policy gradient (DDPG) algorithm, a value-based actor-critic reinforcement learning algorithm, which is very effective for continuous and high-dimensional action space. It harmonically controls hardware components and identifies the packet arrival rates and traffic patterns for the network function chains, minimizing the energy consumed.

GreenABR project proposes a new deep reinforcement learning-based adaptive bitrate (ABR) scheme that optimizes the energy consumption during on-demand video streaming without sacrificing the user quality of experience. It employs a standard perceived quality metric and real power measurements collected through a streaming application. GreenABR’s deep reinforcement learning model makes no assumptions about the streaming environment and learns how to adapt to dynamically changing conditions in a wide range of real network scenarios.

These three solutions can adaptively tune several application-layer and kernel-layer data transfer and streaming parameters, ensuring efficient utilization of the end system and networking resources to decrease the data transfer energy consumption without sacrificing the end-to-end performance or quality of experience for the users. This is a radically different approach for energy-efficient data movement and sustainable networking, which can potentially save millions of dollars for the US economy and reduce the carbon footprint of moving data over the networks.