Benchmarking the effects of operating system interference on extreme-scale parallel machines Export

@article{citeulike:3206617, abstract = {Abstract\ \  We investigate operating system noise, which we identify as one of the main reasons for a lack of synchronicity in parallel applications. Using a microbenchmark, we measure the noise on several contemporary platforms and find that, even with a general-purpose operating system, noise can be limited if certain precautions are taken. We then inject artificially generated noise into a massively parallel system and measure its influence on the performance of collective operations. Our experiments indicate that on extreme-scale platforms, the performance is correlated with the largest interruption to the application, even if the probability of such an interruption on a single process is extremely small. We demonstrate that synchronizing the noise can significantly reduce its negative influence.}, author = {Beckman, Pete and Iskra, Kamil and Yoshii, Kazutomo and Coghlan, Susan and Nataraj, Aroon}, citeulike-article-id = {3206617}, citeulike-linkout-0 = {http://dx.doi.org/10.1007/s10586-007-0047-2}, citeulike-linkout-1 = {http://www.springerlink.com/content/x6763683070x6766}, day = {3}, doi = {10.1007/s10586-007-0047-2}, journal = {Cluster Computing}, keywords = {osnoise}, month = {March}, number = {1}, pages = {3--16}, posted-at = {2008-09-08 19:35:40}, priority = {2}, title = {Benchmarking the effects of operating system interference on extreme-scale parallel machines}, url = {http://dx.doi.org/10.1007/s10586-007-0047-2}, volume = {11}, year = {2008} }

TY - JOUR ID - citeulike:3206617 L3 - citeulike-article-id:3206617 N2 - Abstract   We investigate operating system noise, which we identify as one of the main reasons for a lack of synchronicity in parallel applications. Using a microbenchmark, we measure the noise on several contemporary platforms and find that, even with a general-purpose operating system, noise can be limited if certain precautions are taken. We then inject artificially generated noise into a massively parallel system and measure its influence on the performance of collective operations. Our experiments indicate that on extreme-scale platforms, the performance is correlated with the largest interruption to the application, even if the probability of such an interruption on a single process is extremely small. We demonstrate that synchronizing the noise can significantly reduce its negative influence. IS - 1 JF - Cluster Computing EP - 16 TI - Benchmarking the effects of operating system interference on extreme-scale parallel machines VL - 11 SP - 3 KW - osnoise AU - Beckman, Pete AU - Iskra, Kamil AU - Yoshii, Kazutomo AU - Coghlan, Susan AU - Nataraj, Aroon PY - 2008/03/03/ UR - http://dx.doi.org/10.1007/s10586-007-0047-2 ER -