BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20121115T180000Z
DTEND:20121115T183000Z
LOCATION:355-EF
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: When multiple threads or processes run on a multicore CPU they compete =0Afor shared resources, such as caches and memory controllers, and can =0Asuffer performance degradation as high as 200%. We design and evaluate a =0Anew machine learning model that estimates this degradation online, on =0Apreviously unseen workloads, and without perturbing the execution.=0A=0AOur motivation is to help data center and HPC cluster operators =0Aeffectively use workload consolidation.  Consolidation places many =0Arunnable entities on the same server to maximize hardware utilization, =0Abut may sacrifice performance as threads compete for resources. Our =0Amodel helps determine when consolidation is overly harmful to =0Aperformance. Our work is the first to apply machine learning to this =0Aproblem domain, and we report on our experience reaping the advantages =0Aof machine learning while navigating around its limitations. We =0Ademonstrate how the model can be used to improve performance fidelity =0Aand save power for HPC workloads.
SUMMARY:A Practical Method for Estimating Performance Degradation on Multicore Processors and its Application to HPC Workloads
PRIORITY:3
END:VEVENT
END:VCALENDAR
BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20121115T180000Z
DTEND:20121115T183000Z
LOCATION:355-EF
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: When multiple threads or processes run on a multicore CPU they compete =0Afor shared resources, such as caches and memory controllers, and can =0Asuffer performance degradation as high as 200%. We design and evaluate a =0Anew machine learning model that estimates this degradation online, on =0Apreviously unseen workloads, and without perturbing the execution.=0A=0AOur motivation is to help data center and HPC cluster operators =0Aeffectively use workload consolidation.  Consolidation places many =0Arunnable entities on the same server to maximize hardware utilization, =0Abut may sacrifice performance as threads compete for resources. Our =0Amodel helps determine when consolidation is overly harmful to =0Aperformance. Our work is the first to apply machine learning to this =0Aproblem domain, and we report on our experience reaping the advantages =0Aof machine learning while navigating around its limitations. We =0Ademonstrate how the model can be used to improve performance fidelity =0Aand save power for HPC workloads.
SUMMARY:A Practical Method for Estimating Performance Degradation on Multicore Processors and its Application to HPC Workloads
PRIORITY:3
END:VEVENT
END:VCALENDAR