Bandwidth bandit: Understanding memory contention
作者： David Eklov , Nikos Nikoleris , David Black-Schaffer , Erik Hagersten
DOI: 10.1109/ISPASS.2012.6189214
关键词:
摘要: Applications that are co-scheduled on a multicore compete for shared resources, such as cache capacity and memory bandwidth. The performance degradation resulting from this contention can be substantial, which makes it important to effectively manage these resources. This, however, requires insight into how applications impacted by contention. In paper we present quantitative method measure applications' sensitivities different degrees of off-chip bandwidth real hardware. This is then used demonstrate the varying sensitivity across selection benchmarks, explains why some them experience substantial slowdowns long before overall saturates.
uu.se 
sci-hub.se 参考文章(4)
John D. C. Little, A Proof for the Queuing Formula:L= λW Operations Research. ,vol. 9, pp. 383- 387 ,(1961) , 10.1287/OPRE.9.3.383
Brian M. Rogers, Anil Krishna, Gordon B. Bell, Ken Vu, Xiaowei Jiang, Yan Solihin, Scaling the bandwidth wall Proceedings of the 36th annual international symposium on Computer architecture - ISCA '09. ,vol. 37, pp. 371- 382 ,(2009) , 10.1145/1555754.1555801
Tanima Dey, Wei Wang, Jack W. Davidson, Mary Lou Soffa, Characterizing multi-threaded applications based on shared-resource contention international symposium on performance analysis of systems and software. pp. 76- 86 ,(2011) , 10.1109/ISPASS.2011.5762717
Lingjia Tang, Jason Mars, Neil Vachharajani, Robert Hundt, Mary Lou Soffa, The impact of memory subsystem resource sharing on datacenter applications Proceeding of the 38th annual international symposium on Computer architecture - ISCA '11. ,vol. 39, pp. 283- 294 ,(2011) , 10.1145/2000064.2000099