Purdue Federal Credit Union
In this paper, the authors analyze the impact of scheduling decisions on dynamic task performance. Performance behavior is analyzed utilizing support workloads from SPECWeb 2005 on a multicore hardware platform with an Apache web server. Hardware performance counter data is collected via extending the Linux scheduler and analysis is then performed by core, by task, and by various metrics. The results show that considering a single per-core metric is not sufficient to categorize application behavior, since different thread types often have highly varying characteristics.