Browse Prior Art Database

Medium Access Control Protocol for Ring and Bus Networks Based on Reservation and Deferment Providing Throughput Fairness Control and Tight Access Delay Bounds

IP.com Disclosure Number: IPCOM000107878D
Original Publication Date: 1992-Mar-01
Included in the Prior Art Database: 2005-Mar-22
Document File: 6 page(s) / 282K

Publishing Venue

IBM

Related People

Lemppenau, WW: AUTHOR [+2]

Abstract

This article describes a combined reservation and deferment mechanism for the Medium Access Control (MAC) of ring and bus networks with a slotted transmission data structure operated with or without insertion buffer of the type described in (*). Reservation guarantees tight access-delay bounds for all priorities. Its combination with access deferment ensures throughput fairness. Low access-delay bounds are crucial for prioritized access and low-traffic users during heavy load periods. Throughput-fairness control might be required among heavy traffic users such that all users of the same throughput class obtain equal shares. A scheduler determines cyclically a throughput-fairness reference denoted as threshold T from which class-dependent thresholds T'= kT. are derived.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 31% of the total text.

Medium Access Control Protocol for Ring and Bus Networks Based on Reservation and Deferment Providing Throughput Fairness Control and Tight Access Delay Bounds

       This article describes a combined reservation and
deferment mechanism for the Medium Access Control (MAC) of ring and
bus networks with a slotted transmission data structure operated with
or without insertion buffer of the type described in (*).
Reservation guarantees tight access-delay bounds for all priorities.
Its combination with access deferment ensures throughput fairness.
Low access-delay bounds are crucial for prioritized access and
low-traffic users during heavy load periods. Throughput-fairness
control might be required among heavy traffic users such that all
users of the same throughput class obtain equal shares. A scheduler
determines cyclically a throughput-fairness reference denoted as
threshold T from which class-dependent thresholds T'= kT. are
derived. Nodes currently having a lower throughput obtain reserved
bandwidth, whereas nodes exceeding their threshold T' are temporarily
forced to refrain from accessing the medium.

      Consider a ring or bus network with a slotted transmission data
structure where slots are accessed through two distinct mechanisms:
(1) immediate access of so-called gratis slots and (2) access of
previously reserved slots. A gratis/reserved flag is used to
distinguish between the two states of a slot. In either case, the
busy/free flag needs to be free for the slot to be accessible.
Reservation is done by a Reserve/Confirm command exchange where
cyclically transmission requests are collected, requests are
processed by a scheduling node, and confirmations are returned to the
requesting nodes. Marking of slots as reserved is done by the
scheduler whereby a reservation is for only one transmission access.
When a slot is freed by the destination, subsequent access to that
slot is made via the gratis access mechanism, until the next marking
takes place. In (*), gratis access is unrestricted and does not
completely constrain the added throughput capacity obtained by
spatial reuse. Here a mechanism to enforce fair allocation of both,
the network's bit rate capacity and the extra bandwidth originating
from spatial reuse, while maintaining the tight access-delay-bound
property, is disclosed.

      The basic idea is to use a throughput-fairness reference
(denoted as threshold T) so that all nodes currently having a lower
throughput obtain confirmations for reserved slots, whereas all other
nodes are temporarily forced to refrain from gratis slot access. In
general, class-dependent thresholds T' may be used which are derived
from threshold T by multiplication with the corresponding throughput
class factors k. Related processing can be done completely in the
scheduler, or may be split between the scheduling node and all other
nodes. The description is for a distributed approach which reduces
scheduling processing delay (owing to para...