{"title":"本机ATM与IP over ATM:比较研究","authors":"T. Chahed, S. Ben Fredj, C. Fayet","doi":"10.1109/ATM.1999.786777","DOIUrl":null,"url":null,"abstract":"The main focus of this paper is native ATM versus IP over ATM solutions. We investigate the theoretical and experimental comparative performance of native ATM and CLIP. We explicitly quantify loss, throughput at the receiver (inversely proportional to delay) and CPU utilization. The key results of this work are: first, we observe a relatively important loss rate for small packet sizes which gets smaller for larger packets. UDP shows larger loss rate than native ATM. For larger packet sizes loss rate tends to zero. Second, for both UDP and native ATM, throughput at the destination is unstable for packets of small size. It reaches equilibrium for packet sizes on the order of 10k octets. Native ATM has a higher throughput owing to the smaller overhead it presents. Third, for both protocols, the CPU is used more at the receiver. At the transmitter, UDP uses more CPU than native ATM, owing to the overhead it introduces across the layers. Those results are further discussed.","PeriodicalId":266412,"journal":{"name":"IEEE ATM Workshop '99 Proceedings (Cat. No. 99TH8462)","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"1999-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Native ATM versus IP over ATM: comparative study\",\"authors\":\"T. Chahed, S. Ben Fredj, C. Fayet\",\"doi\":\"10.1109/ATM.1999.786777\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"The main focus of this paper is native ATM versus IP over ATM solutions. We investigate the theoretical and experimental comparative performance of native ATM and CLIP. We explicitly quantify loss, throughput at the receiver (inversely proportional to delay) and CPU utilization. The key results of this work are: first, we observe a relatively important loss rate for small packet sizes which gets smaller for larger packets. UDP shows larger loss rate than native ATM. For larger packet sizes loss rate tends to zero. Second, for both UDP and native ATM, throughput at the destination is unstable for packets of small size. It reaches equilibrium for packet sizes on the order of 10k octets. Native ATM has a higher throughput owing to the smaller overhead it presents. Third, for both protocols, the CPU is used more at the receiver. At the transmitter, UDP uses more CPU than native ATM, owing to the overhead it introduces across the layers. Those results are further discussed.\",\"PeriodicalId\":266412,\"journal\":{\"name\":\"IEEE ATM Workshop '99 Proceedings (Cat. No. 99TH8462)\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"1999-05-24\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"IEEE ATM Workshop '99 Proceedings (Cat. No. 99TH8462)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ATM.1999.786777\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE ATM Workshop '99 Proceedings (Cat. No. 99TH8462)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ATM.1999.786777","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
摘要
本文的主要焦点是本机ATM与IP over ATM解决方案。我们研究了原生ATM和CLIP的理论和实验性能比较。我们明确量化了损失、接收器上的吞吐量(与延迟成反比)和CPU利用率。这项工作的关键结果是:首先,我们观察到小数据包大小的相对重要的损失率,对于较大的数据包会变小。UDP比本机ATM丢包率高。对于较大的包大小,丢包率趋于零。其次,对于UDP和本机ATM,对于小尺寸的数据包,目的地的吞吐量是不稳定的。它在10k字节量级的数据包大小上达到平衡。本机ATM具有更高的吞吐量,因为它的开销更小。第三,对于这两种协议,CPU更多地用于接收端。在发送端,UDP比本机ATM使用更多的CPU,因为它跨层引入了开销。这些结果将进一步讨论。
The main focus of this paper is native ATM versus IP over ATM solutions. We investigate the theoretical and experimental comparative performance of native ATM and CLIP. We explicitly quantify loss, throughput at the receiver (inversely proportional to delay) and CPU utilization. The key results of this work are: first, we observe a relatively important loss rate for small packet sizes which gets smaller for larger packets. UDP shows larger loss rate than native ATM. For larger packet sizes loss rate tends to zero. Second, for both UDP and native ATM, throughput at the destination is unstable for packets of small size. It reaches equilibrium for packet sizes on the order of 10k octets. Native ATM has a higher throughput owing to the smaller overhead it presents. Third, for both protocols, the CPU is used more at the receiver. At the transmitter, UDP uses more CPU than native ATM, owing to the overhead it introduces across the layers. Those results are further discussed.