Where’s all the time going?

In a UNIX system, time command reports real, system and user time used bycommand. For various reasons, the reported times are likely to be at variance with the actual resources used by the program. We added high resolution timing of the actual time spent in various states in the kernel to Linux 2.5. 1 The Problems When you time a process using time(1), you get back three numbers: real time, user time, and system time. These numbers are generated statistically; at regular intervals, an interrupt occurs and the current process gets an extra tick in either its system time counter or its user time counter. These numbers should be viewed with extreme scepticism. • Multi-threaded processes are not accounted for correctly. Only the ‘main ’ thread is counted; time used by threads that it starts are accounted for separately. As LinuxThreads has an extra manager thread which starts all the other threads, created threads are grandchildren not children of the main thread, so their times are never accumulated into the main thread’s timers. • There are also processes that wake at regular intervals, do a small amount of processing (less than a tick) then sleep (e.g., Apache) — their wakeup interval tends to become synchronised with the timer interrupt so their ∗ This work was funded by HP through the Gelato programm