{"title":"The Ethernet capture effect: analysis and solution","authors":"K. Ramakrishnan, Henry S. Yang","doi":"10.1109/LCN.1994.386597","DOIUrl":null,"url":null,"abstract":"We analyze the behavior of the Ethernet in networks with a small number of active stations, and describe what is known as the Ethernet capture effect. Where a station transmits consecutive packets exclusively for a prolonged period despite other stations contending for access. The capture effect causes transient unfairness, which results in substantial performance degradation. We report measurements using TCP/IP that show the performance degradation. A solution is proposed that effectively overcomes the capture effect. The proposed algorithm, which we call the capture avoidance binary exponential backoff (CABEB), uses the standard binary exponential backoff (BEB) with enhancements for collision resolution in the special case when a station attempts to capture the channel subsequent to an uninterrupted consecutive transmit. Using a detailed simulation, we show the efficacy of the CABEB algorithm over the standard BEB in overcoming the unfairness resulting from stations capturing the channel. The CABEB improves throughput for protocols like TCP/IP, reduces variability in the channel access latency and eliminates packet discards due to excessive collisions in a 2-node network. The algorithm is a modification that is compliant with the Ethernet/802.3 standards. For networks with a large number of active stations, the CABEB performs as well as the standard BEB algorithm. Our study placer emphasis on the workload and network configuration that is the worst case relative to the Ethernet capture effect to show that the proposed algorithm is a substantial improvement over the existing backoff algorithm.<<ETX>>","PeriodicalId":270137,"journal":{"name":"Proceedings of 19th Conference on Local Computer Networks","volume":"123 11 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"120","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of 19th Conference on Local Computer Networks","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/LCN.1994.386597","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 120
Abstract
We analyze the behavior of the Ethernet in networks with a small number of active stations, and describe what is known as the Ethernet capture effect. Where a station transmits consecutive packets exclusively for a prolonged period despite other stations contending for access. The capture effect causes transient unfairness, which results in substantial performance degradation. We report measurements using TCP/IP that show the performance degradation. A solution is proposed that effectively overcomes the capture effect. The proposed algorithm, which we call the capture avoidance binary exponential backoff (CABEB), uses the standard binary exponential backoff (BEB) with enhancements for collision resolution in the special case when a station attempts to capture the channel subsequent to an uninterrupted consecutive transmit. Using a detailed simulation, we show the efficacy of the CABEB algorithm over the standard BEB in overcoming the unfairness resulting from stations capturing the channel. The CABEB improves throughput for protocols like TCP/IP, reduces variability in the channel access latency and eliminates packet discards due to excessive collisions in a 2-node network. The algorithm is a modification that is compliant with the Ethernet/802.3 standards. For networks with a large number of active stations, the CABEB performs as well as the standard BEB algorithm. Our study placer emphasis on the workload and network configuration that is the worst case relative to the Ethernet capture effect to show that the proposed algorithm is a substantial improvement over the existing backoff algorithm.<>