EABF: Energy Efficient Self-Adaptive Bloom Filter for Network Packet Processing Yachao Zhou School of Electronic Engineering Dublin City University Dublin, Ireland Tian Song * School of Computer Science Beijing Institute of Technology Beijing, China Xiaojun Wang School of Electronic Engineering Dublin City University Dublin, Ireland e-mail: yachao@eeng.dcu.ie * Corresponding Author: e-mail: songtian@bit.edu.cn e-mail: wangx@eeng.dcu.ie Abstract—Future Internet requires re-thinking of network infrastructure towards the balance between computing capacities and energy sustainable techniques. As one of computing intensive components, Bloom filters are widely used for network packet processing. In this paper, an energy efficient self-adaptive Bloom filter, EABF, is devoted to a balance of power and performance especially for high performance networks. The basic idea is to give the Bloom Filter the capability to adjust the number of active hash functions according to the current workload automatically. This adaption depends on its control policies. Three policies are presented and compared. We also give the method to implement EABF in hardware for higher performance. It is presented in a two-stage platform based on FPGA where Stage 1 is always active and Stage 2, a secondary stage, is only active when necessary. The platform can also be extended to multi-stages. A control circuit is designed for flexibly changing working stage and reducing both dynamic and static power consumption. Analysis and experiments show that our dynamic two-stage EABF can achieve almost the best power savings as that of the fixed schemes; unlike the fixed schemes that might have much longer latency, EABF maintains nearly 1 clock cycle latency as that of a regular Bloom filter. Keywords - power-efficiency; Bloom filter; self-adaptive I. INTRODUCTION In recent years, the statistics of network energy consumption reported by Internet Service Providers (ISP) show an alarming and growing trend. The Internet power is estimated to be over 4% of the whole electricity consumption as the access rate increases [1]. To support the fast growth of customer population and the expanding offer of new services, ISPs need more high speed network devices to process the larger volume of traffic data. The drive for low power comes from environmental and economical motivations. Therefore for the emerging future Internet, besides other basic concepts and key aspects, it is recognized that energy efficiency should also become part of the network design criteria [2]. Network applications primarily rely on hardware platforms to keep up with network speed. However, the energy efficiency of silicon technologies improves at a slower pace, by a factor of 1.65 every 18 months, compared with router capacity which increases by a factor of 2.5 every 18 months [3]. Thus it is difficult to reduce power consumption while guaranteeing network processing capacity based on current technology. In fact, power efficient Internet is still at an exploratory stage, in which the key components in future routers should be made more energy efficient. Bloom filters have been widely used in various network applications like network security, storage network or traffic engineering etc. For example, Bloom filters can be used to share web cache information, or to summarize contents for nodes cooperation for peer-to-peer networks [4]. Efficient Bloom filter design will reduce power consumption for various network applications. However, most existing Bloom filter based low power solutions are achieved at the cost of performance degradation [6,7,8]. Compared with regular Bloom filter approach, these solutions show significant power savings for average traffic conditions, but they cannot handle peak traffic and seriously affect processing speed in the worst case. Our objective is to reduce overall energy consumption of Bloom filter based network applications without adversely affecting processing latency. We propose an energy efficient self-adaptive Bloom filter, EABF, as an optimal balance of power and latency. EABF is capable of automatically adjusting the number of active hash functions according to the current workload. Hash functions in EABF can be switched among three states: working in the first stage, sleep or working in other stages. The first stage is always active, whereas the second and subsequent stages are activated by a match output from their respective previous stages. We make the following contributions: 1) a self-adaptive two-stage platform for EABF; 2) three control strategies for the movement of hash functions between two stages according to workload; 3) a key control circuit aimed at flexibly moving hash functions between working stages in one clock cycle and reducing both dynamic and static power consumption. Experiments show that EABF can significantly reduce power consumption to a level similar to that of fixed stage low power Bloom filters [6,7,8]; moreover, EABF has low latency close to one clock cycle as a regular Bloom filter, much shorter than that of fixed stage low power Bloom filters. The rest of the paper is organized as follows. We review the related work in Section II and describe the overall method of EABF in Section III. The detailed control policies follow in Section IV. Section V presents the implementation and improvements to further reduce power consumption. The theoretical analysis and simulations are presented in Section VI. Finally, we draw our conclusion in Section VII.