{"title":"当保证最小流出速率时,基于信用的流量控制的缓冲要求","authors":"M. Katevenis","doi":"10.1109/HPCS.1997.864039","DOIUrl":null,"url":null,"abstract":"Credit-based flow control (FQ offers significant advantages for data communication over networks such as ATM. One concern is the buffer memory requirement of credit FC, which is high when static, per-connection buffer allocation is used. The memory size can be reduced by dynamically sharing the buffer among the connections. We formulate and prove a new upper bound for the dynamically-shared buffer size, which applies when the downstream network guarantees that it will never throttle a connection down below a given minimum \"draining\" rate. This upper bound is inversely proportional to the minimum draining rate, and is independent of the number of connections. The minimum draining rate can be guaranteed by giving to the traffic arriving at a local area through long distance links priority over locally generated traffic in the use of a given fraction of the outgoing network capacity. In practical situations, the new upper bound can be made small, thus providing a new, additional method to reduce the buffer memory cost to a very small percentage of the link cost. Unlike other such methods, no link underutilization may occur and there is no rate ramp-up delay.","PeriodicalId":178651,"journal":{"name":"The Fourth IEEE Workshop on High-Performance Communication Systems","volume":"8 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"1997-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"5","resultStr":"{\"title\":\"Buffer requirements of credit-based flow control when a minimum draining rate is guaranteed\",\"authors\":\"M. Katevenis\",\"doi\":\"10.1109/HPCS.1997.864039\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Credit-based flow control (FQ offers significant advantages for data communication over networks such as ATM. One concern is the buffer memory requirement of credit FC, which is high when static, per-connection buffer allocation is used. The memory size can be reduced by dynamically sharing the buffer among the connections. We formulate and prove a new upper bound for the dynamically-shared buffer size, which applies when the downstream network guarantees that it will never throttle a connection down below a given minimum \\\"draining\\\" rate. This upper bound is inversely proportional to the minimum draining rate, and is independent of the number of connections. The minimum draining rate can be guaranteed by giving to the traffic arriving at a local area through long distance links priority over locally generated traffic in the use of a given fraction of the outgoing network capacity. In practical situations, the new upper bound can be made small, thus providing a new, additional method to reduce the buffer memory cost to a very small percentage of the link cost. Unlike other such methods, no link underutilization may occur and there is no rate ramp-up delay.\",\"PeriodicalId\":178651,\"journal\":{\"name\":\"The Fourth IEEE Workshop on High-Performance Communication Systems\",\"volume\":\"8 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"1997-06-23\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"5\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"The Fourth IEEE Workshop on High-Performance Communication Systems\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/HPCS.1997.864039\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"The Fourth IEEE Workshop on High-Performance Communication Systems","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/HPCS.1997.864039","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Buffer requirements of credit-based flow control when a minimum draining rate is guaranteed
Credit-based flow control (FQ offers significant advantages for data communication over networks such as ATM. One concern is the buffer memory requirement of credit FC, which is high when static, per-connection buffer allocation is used. The memory size can be reduced by dynamically sharing the buffer among the connections. We formulate and prove a new upper bound for the dynamically-shared buffer size, which applies when the downstream network guarantees that it will never throttle a connection down below a given minimum "draining" rate. This upper bound is inversely proportional to the minimum draining rate, and is independent of the number of connections. The minimum draining rate can be guaranteed by giving to the traffic arriving at a local area through long distance links priority over locally generated traffic in the use of a given fraction of the outgoing network capacity. In practical situations, the new upper bound can be made small, thus providing a new, additional method to reduce the buffer memory cost to a very small percentage of the link cost. Unlike other such methods, no link underutilization may occur and there is no rate ramp-up delay.