{"title":"Performance Evaluation of DCB's Priority-Based Flow Control","authors":"Mikkel Hagen, Ryan Zarick","doi":"10.1109/NCA.2011.58","DOIUrl":null,"url":null,"abstract":"Data Center Bridging (DCB) is a group of technologies defined within the IEEE to transform Ethernet into a lossless fabric. This paper uses data center network cards and switches to analyze the performance of Priority-based Flow Control, a DCB technology. The network hardware required to support this new technology is still new and not widely available. This paper is the first to experimentally show how Priority-based Flow Control performs well in simple systems but as the system becomes more complicated traffic streams within a priority show fate sharing characteristics. Notably, the throughput of different traffic streams converges to the lowest throughput among the streams. One of the most common applications used in data centers for interprocessor communication, MPI (Message Passing Interface), is analyzed. Finally, an iSCSI initiator and target are used to observe the throughput of iSCSI over a network with PFC. This paper shows the performance benefits of using PFC with TCP-based applications.","PeriodicalId":258309,"journal":{"name":"2011 IEEE 10th International Symposium on Network Computing and Applications","volume":"1 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2011-08-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"5","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2011 IEEE 10th International Symposium on Network Computing and Applications","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/NCA.2011.58","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 5
Abstract
Data Center Bridging (DCB) is a group of technologies defined within the IEEE to transform Ethernet into a lossless fabric. This paper uses data center network cards and switches to analyze the performance of Priority-based Flow Control, a DCB technology. The network hardware required to support this new technology is still new and not widely available. This paper is the first to experimentally show how Priority-based Flow Control performs well in simple systems but as the system becomes more complicated traffic streams within a priority show fate sharing characteristics. Notably, the throughput of different traffic streams converges to the lowest throughput among the streams. One of the most common applications used in data centers for interprocessor communication, MPI (Message Passing Interface), is analyzed. Finally, an iSCSI initiator and target are used to observe the throughput of iSCSI over a network with PFC. This paper shows the performance benefits of using PFC with TCP-based applications.