Premium
Mitigating tail latency in IEEE 802.11–based networks
Author(s) -
Kim Dohyung,
Yeom Ikjun,
Lee TaeJin
Publication year - 2017
Publication title -
international journal of communication systems
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.344
H-Index - 49
eISSN - 1099-1131
pISSN - 1074-5351
DOI - 10.1002/dac.3404
Subject(s) - computer science , computer network , latency (audio) , network packet , ieee 802.11 , random access , distributed coordination function , node (physics) , network allocation vector , hidden node problem , throughput , real time computing , wireless network , wireless , telecommunications , structural engineering , engineering , wi fi array
Summary Delay sensitive applications are being actively introduced with the advent of 5G and vehicular communications, and such applications are very sensitive to tail latency. However, tail latency has not been seriously considered so far, especially in IEEE 802.11–based networks. Channel access is scheduled by random Contention Window (CW) values in IEEE 802.11–based networks, and the node with the larger CW waits longer, and it may even observe multiple transmissions from a single contending node, which results in a long latency tail. In this paper, we propose a new decentralized MAC called SynchMAC to mitigate this latency tail. In SynchMAC, every competing node transmits exactly one packet within a virtual time slot without a centralized controller. Using the proposed approach, the maximum channel access latency is bounded by T ×2 N , where T is the time required for transmitting a single packet (including Inter‐Frame Space and CW) and N is the number of competing nodes. To maximize the system throughput, the proposed scheme optimizes the value of T by considering the probability of successful transmission. Our simulation study shows that SynchMAC reduces the maximum access latency by up to 94% and 53% compared with the conventional IEEE 802.11 MAC and the comparative scheme, respectively, without degrading throughput performance. We also show that SynchMAC is easily extended to support weighted access.