-1

I am currently working with RTOS, their I have Time Tick Value = 10 mSec and Number of concurrent running tasks = 10.

Now my question is, What should be my Single Thread Time (quantum)?

I think, Single Thread Time = Time Tick Value / Number of concurrent running tasks but I am not sure. Please correct me if I am making any mistake here.

Also what are the factors that were responsible for change-in Single Thread Time?

TERMINOLOGIES:

Time Tick Value: The available CPU time is divided into time slices and the RTX kernel assigns a time slice to each task (default time slice is set to 10 ms).

Single Thread Time is the time slice taken by one thread in round robbin scheduling if their are total 10 tasks (threads).

Number of concurrent running tasks always remain constant. It will not cover thread waiting time.

What I wanna know is, whether 10 mSec time slice given to single thread or it gets divide again into 10 parts for 10 concurrent running threads (Consider I am using round robbin scheduling).

appleBoy21
  • 632
  • 8
  • 23
  • Can you define your terms? E.g. is `Time Tick Value` the amount of time a thread will run before (voluntarily) blocking or is it the maximum time a crusty old round robin scheduler might give a thread before preempting it (if it doesn't block sooner)? Is `Single Thread Time` the time one thread would get under various unknown conditions (e.g. that all threads want and get equal time) or is it the amount of time it'd take if all the work was done by one thread instead of 10? Does `Number of concurrent running tasks` remain constant (!) and does it include threads waiting to get CPU time? – Brendan Jan 07 '19 at 09:38
  • @Brendan - Please check updated terminologies. – appleBoy21 Jan 07 '19 at 09:53
  • The question assumes all tasks are running at the same priority and are always in the ready state (i.e. they never yield or block). Whilst you can implement such scheduling in most RTOS schedulers, it is by no means usual to do so and is not a good design from a _real-time_ point of view and not an effective use of an RTOS. Moreover some RTOS that allow round-robing scheduling allow a scheduling slice to be a _multiple_ of the tick period, and may even allow each task to have a different multiple. If a thread blocks, the scheduler will run and another thread will get the rest of its slice. – Clifford Jan 13 '19 at 18:08

1 Answers1

2

For a round robin scheduler using 10 ms time slices, if there is one CPU and always 10 tasks; then there will always be one task actually using the CPU and 9 tasks waiting to use the CPU. In this case the scheduler gives a task 10 ms of CPU time, then preempts and does a task switch to the next task and gives it 10 ms of time, and so on; so each task would get 10 ms of CPU time every 100 ms.

What I wanna know is, whether 10 mSec time slice given to single thread or it gets divide again into 10 parts for 10 concurrent running threads (Consider I am using round robbin scheduling).

For round robin, the time slice length is the time any task will be given - the 10 ms slices won't be sliced up into smaller slices (otherwise it'd be a round robin with 1 ms time slices, or it wouldn't be round robin - e.g. would become "variable time slice").

Brendan
  • 35,656
  • 2
  • 39
  • 66
  • 1
    This is assuming one of the tasks does not relinquish control. If it does the next ready task will start running for the remaining time left before the next tick at which point it will switch task again. – Realtime Rik Jan 07 '19 at 12:41