Where is Your Application StuckVivek Kashyap (firstname.lastname@example.org)
The talk describes per-task delay accounting, a newly introduced kernel mechanism that provides a direct measurement of time spent by Linux tasks waiting for CPU time, completion of submitted I/O and in resolving page faults caused by allocated real memory, all of which delay the forward progress of the application being run within the task. We discuss how such measurements help in resource management and metering, which are increasingly important given virtualization trends. We also describe other related accounting mechanisms introduced in the kernel.
Attendees interested in the design and implementation details should have a reasonable familiarity with CPU and I/O scheduling and page fault handling. The talk will also be of interest to practitioners interested in quickly determining application performance bottlenecks and to developers of resource management applications.