Browse Prior Art Database

MAC Layer Handling of Preemptive and Nonpreemptive Priorities in Buffer-Insertion LANs

IP.com Disclosure Number: IPCOM000111097D
Original Publication Date: 1994-Feb-01
Included in the Prior Art Database: 2005-Mar-26
Document File: 4 page(s) / 150K

Publishing Venue

IBM

Related People

Heddes, M: AUTHOR [+2]

Abstract

Consider an insertion-buffer LAN where a frame is segmented into slot entities but the transmission itself takes place in contiguous slots [*]. Insertion buffers in the data path of the transmission medium may cause significant variations in the transfer time of a frame propagating from a source to a destination. It is therefore essential that time-sensitive data is able to bypass and even to preempt low-priority data that comes either from a node's insertion buffer or its transmit buffer. This makes it necessary to use insertion buffers with priority management (Figure).

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 41% of the total text.

MAC Layer Handling of Preemptive and Nonpreemptive Priorities in
Buffer-Insertion LANs

      Consider an insertion-buffer LAN where a frame is segmented
into slot entities but the transmission itself takes place in
contiguous slots [*].  Insertion buffers in the data path of the
transmission medium may cause significant variations in the transfer
time of a frame propagating from a source to a destination.  It is
therefore essential that time-sensitive data is able to bypass and
even to preempt low-priority data that comes either from a node's
insertion buffer or its transmit buffer.  This makes it necessary to
use insertion buffers with priority management (Figure).

      A hierarchical insertion-buffer structure as well as frame
preemptions require a medium access mechanism that meets the
following objectives:  (1) lossless operation by avoiding
buffer-insertion overflows, (2) repeated preemptions within the
priority hierarchy, and (3) avoidance that frames of the same
priority become interleaved.

      A node transmits frames of P+1 priorities, whereby the priority
ranking is by increasing priority number k (with k = 0, ...  P).  The
frames are enqueued by priority into one of the k transmit queues TX
sub k. The pending frame of priority k consists of r sub %k slot
entities.  The node has an insertion buffer with buffer units, each
able to contain a slot.  The number of free buffer units is denoted
by f. To control the frame transmission of different priorities, each
priority k maintains its own insertion-buffer queue IB sub k.  Its
size consists of d sub %k dedicated buffer units plus a common buffer
part that depends on the priority level.  A higher priority is
allowed to use the buffer part belonging to all the lower priorities.
The access mechanism makes further use of two flags per priority: an
insertion-buffer occupancy flag b sub %k (indicating whether the
insertion-buffer queue IB sub k is empty or not) and a transmit ready
flag x sub %k (indicating whether the frame waiting in the transmit
queue TX sub k meets the access conditions as discussed below or
not).

      In order to ensure that no slots are lost on the medium because
of insertion buffer overflows, it must be guaranteed that during the
transmission of a frame all busy slots coming from upstream can be
held up in the insertion buffer.  This slot quantity may be as large
as the number of slots r sub %k necessary to transmit the current
frame of priority k. Since priority k is allowed to use the buffer
units d sub <k-1> ' to ' d sub 0, the number of free buffer units
that can be used by each priority class decrease with the priority
level.  Thus, for four priority classes (P = 3), one obtains the
basic access conditions as given in the left side of the table below.
These conditions can be simplified by replacing the sum of the buffer
units d sub %k by a single variable m sub %k = sum from <n=k+1> to P
of d sub %n for priorities k = 0 to P, whe...