Browse Prior Art Database

Deadline Monotonic Server

IP.com Disclosure Number: IPCOM000116538D
Original Publication Date: 1995-Sep-01
Included in the Prior Art Database: 2005-Mar-30
Document File: 4 page(s) / 118K

Publishing Venue

IBM

Related People

Arida, P: AUTHOR [+3]

Abstract

Described is a Deadline Monotonic Server (DMS) for the IBM* Microkernel, providing dynamic support for the "Deadline Monotonic" scheduling discipline, which deals with periodic scheduling where the deadlines of real-time tasks may be shorter than the period. In this discipline, tasks with shorter deadlines are given higher priority. "Rate Monotonic" scheduling, in which the deadline is equal to the period, is considered to be a subset of the "Deadline Monotonic" discipline.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Deadline Monotonic Server

      Described is a Deadline Monotonic Server (DMS) for the IBM*
Microkernel, providing dynamic support for the "Deadline Monotonic"
scheduling discipline, which deals with periodic scheduling where the
deadlines of real-time tasks may be shorter than the period.  In this
discipline, tasks with shorter deadlines are given higher priority.
"Rate Monotonic" scheduling, in which the deadline is equal to the
period, is considered to be a subset of the "Deadline Monotonic"
discipline.

      Details of the performance characteristics of any new thread to
be scheduled are passed to the DMS server.  If the new thread can be
scheduled without causing any of the existing threads to miss their
deadlines, the new thread is accepted; otherwise an error is
returned, making it necessary for the caller to perform a recovery
action.  Thus, the DMS server is not responsible for the run-time
scheduling of any threads, assigning only a fixed priority within a
reserved range.  Once a thread has been assigned this priority, it is
scheduled by the standard Fixed Priority scheduling systems of the
Microkernel.  In this way, an effective division of labor is achieved
between the server running at a user level, as the long and complex
calculations of the server are not performed in the scarce "wired"
memory of the Microkernel.  These calculations are performed
infrequently, without urgent time constraints.

      The DMS server is appropriately used only when system overload
can be avoided.  Because the priorities assigned to different threads
bear no relationship to their importance, the behavior of a system
controlled by the server under overload conditions may become
undesirable.  A user may ensure that certain threads receive
preferential treatment by means of the well-known technique of period
transformation.

      The Mach Microkernel provides interfaces to facilitate putting
some of the functionality which would conventionally be part of the
kernel into a suer space module.  The DMS server has both an ability
to set the priority of a thread and an ability to keep track of all
the threads placed under its control.  By using the dead-name
notification facility of the Microkernel, the DMS server arranges to
receive a message when any such thread is terminated.  An existing
mechanism for accessing privileged scheduling operations, the
processor set control port, is preferably used to provide such access
for the server.  The DMS server has a well known na...