{"title":"二维网格的容错自适应路由","authors":"C. Cunningham, D. Avresky","doi":"10.1109/HPCA.1995.386549","DOIUrl":null,"url":null,"abstract":"Many massively parallel computers in use today utilize simple deterministic XY wormhole routing to transmit messages between nodes. Because XY routing does not provide any routing adaptability, it lacks the ability to avoid congested links, as well as faults. Therefore, the focus of this paper will be two-fold: improving the performance of wormhole routing and providing fault tolerance for up to N-1 faults in an N/spl times/N two-dimensional mesh. A simulation model based on the Intel Paragon is presented that compares several known routing strategies with the proposed strategy to illustrate how local state information can be used to provide a potential network throughput improvement of up to 20%, while achieving fault tolerance.<<ETX>>","PeriodicalId":330315,"journal":{"name":"Proceedings of 1995 1st IEEE Symposium on High Performance Computer Architecture","volume":"93 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"1995-01-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"75","resultStr":"{\"title\":\"Fault-tolerant adaptive routing for two-dimensional meshes\",\"authors\":\"C. Cunningham, D. Avresky\",\"doi\":\"10.1109/HPCA.1995.386549\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Many massively parallel computers in use today utilize simple deterministic XY wormhole routing to transmit messages between nodes. Because XY routing does not provide any routing adaptability, it lacks the ability to avoid congested links, as well as faults. Therefore, the focus of this paper will be two-fold: improving the performance of wormhole routing and providing fault tolerance for up to N-1 faults in an N/spl times/N two-dimensional mesh. A simulation model based on the Intel Paragon is presented that compares several known routing strategies with the proposed strategy to illustrate how local state information can be used to provide a potential network throughput improvement of up to 20%, while achieving fault tolerance.<<ETX>>\",\"PeriodicalId\":330315,\"journal\":{\"name\":\"Proceedings of 1995 1st IEEE Symposium on High Performance Computer Architecture\",\"volume\":\"93 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"1995-01-22\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"75\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Proceedings of 1995 1st IEEE Symposium on High Performance Computer Architecture\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/HPCA.1995.386549\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of 1995 1st IEEE Symposium on High Performance Computer Architecture","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/HPCA.1995.386549","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Fault-tolerant adaptive routing for two-dimensional meshes
Many massively parallel computers in use today utilize simple deterministic XY wormhole routing to transmit messages between nodes. Because XY routing does not provide any routing adaptability, it lacks the ability to avoid congested links, as well as faults. Therefore, the focus of this paper will be two-fold: improving the performance of wormhole routing and providing fault tolerance for up to N-1 faults in an N/spl times/N two-dimensional mesh. A simulation model based on the Intel Paragon is presented that compares several known routing strategies with the proposed strategy to illustrate how local state information can be used to provide a potential network throughput improvement of up to 20%, while achieving fault tolerance.<>