ZigBee vs WiFi: Understanding Issues and Measuring Performances of their Coexistence Zenghua Zhao 1,2 , Xuanxuan Wu 1 , Xin Zhang 1 , Jing Zhao 3 , Xiang-Yang Li 3 1 School of Computer Science and Technology, Tianjin University, China 2 Tianjin Key Laboratory of Cognitive Computing and Application, China 3 Department of Computer Science, Illinois Institute of Technology, USA Abstract—Wireless coexistence is crucial with the explosive development of wireless technologies in recent years. The coex- istence issues of IEEE 802.11 b/g and IEEE 802.15.4 have been well studied, however few work focused on 802.11n new features including MIMO, channel bonding and frame aggregation. In this paper, we conducted extensive experiments to understand how 802.11n impact on 802.15.4 and vice versa in a systematic way. We consider primary features of 802.11n both in symmetric and asymmetric scenarios. The goal of our work is to gain more insights into the coexistence issues of 802.11n and 802.15.4 and thus to help protocol design and co-located network deployments I. I NTRODUCTION The development of diverse wireless technologies has been explosive in the last few decades. Since various network standards such as IEEE 802.11 (WiFi) [1] and IEEE 802.15.4 (ZigBee) [2] share the 2.4 GHz ISM (Industrial Scientific and Medical) band, cross technology interference is inevitable. These two types of networks exploit the same frequency band, and are widely deployed in a number of common applications in which they have to coexist in close proximity. 802.11n, as one of the latest 802.11 standards, supports a maximum 600 Mbps due to several enhancements including MIMO (Multiple Input and Multiple Output), FA (Frame Aggregation) and channel bonding. Its coexistence with other ISM band technologies is crucial to the wide deployment of 802.11n applications. In the past few years, coexistence issues between legacy 802.11b/g and 802.15.4 have been widely investigated through experiments [3]–[9]. It is shown that the throughput of WiFi and Zigbee degrade heavily when they are co-located, therefore various protocols have been proposed for WiFi to survive in the presence of Zigbee and vice versa [10]–[14]. However, all the work are based on legacy 802.11b/g, few of them consider 802.11n [7], [8]. Nowadays, 802.11n networks are being ubiquitous and have new features (MIMO, FA and channel bonding) different from legacy 802.11b/g. One cannot help asking how 802.11n and 802.15.4 impact each other when they are co-located? Shall we design new protocols to make them co-work? or how to deploy them to alleviate the The research is partially supported by NSFC (National Natural Science Foundation of China) under Grant No. 61172063. The research of Li is partially supported by NSF CNS-1035894, NSF ECCS-1247944, NSF CMMI 1436786, and NSFC under Grant No. 61170216, No. 61228202. 978-1-4799-7575-4/14/$31.00 c 2014 IEEE interference from each other? In this paper, we aim to answer these questions via systematic experiments. We establish a testbed composed of one 802.11n network and one 802.15.4 network to carry out the experiments. We choose two popular commercial 802.11n wireless cards: In- tel5300 and UBNT SR71-A, in order to examine the behaviors of different 802.11n products in the presence of 802.15.4 interference. We focus on new features of 802.11n (MIMO, FA and channel bonding) and check their impact on 802.15.4 and vice versa. We further consider symmetric and asymmetric scenarios, which are formed due to the transmit power dis- crepancy between 802.11n and 802.15.4 nodes [15]. In the symmetric scenario, the signal from the 802.15.4 sender is strong enough to trigger the CCA (Clear Channel Assessment) check on the 802.11n sender, therefore both 802.11n and 802.15.4 senders can hear each other. While in the asymmetric scenario 802.15.4 sender can hear 802.11n nodes, but 802.11n nodes are oblivious of 802.15.4 sender. Our primary findings obtained from our experiments are as follows: (1) In symmetric scenarios, the throughput degradation of 802.11n primarily steps from backoff. Accordingly, the packet losses of 802.15.4 are primarily due to ACF (Access Channel Failure) instead of corruption. Different 802.11n wireless cards have different behaviors when they operate at single-stream and double-stream modes. (2) FA and channel bonding have impact on the co- existence. The 802.15.4 network has better performance in terms of PDR (Packet Delivery Ratio) when the 802.11n network operates at 40 MHz or at smaller FA levels. (3) In asymmetric scenarios, 802.15.4 has no impact on 802.11n. However, the PDR of 802.15.4 decreases to almost zero. The packet losses are due to both ACF and corruption. From these observations, we gain some insights into net- work protocol design and co-located network deployments. Some implications are: (1) It is preferred for 802.15.4 protocol to differentiate the different packet loss types (channel-access-failed or corrupted), thus to improve the PDR of 802.15.4 under the interference of 802.11n. (2) The parameters of 802.11n (20 MHz/40 MHz, FA levels) should be selected carefully when 802.11n and 802.15.4 are co-located, to make them co-exist well. 978-1-4799-7575-4/14/$31.00 ©2014 IEEE