Browse Prior Art Database

Method for Efficiently Providing Different Levels of Network Availability in High-Speed Networks

IP.com Disclosure Number: IPCOM000110937D
Original Publication Date: 1994-Jan-01
Included in the Prior Art Database: 2005-Mar-26
Document File: 6 page(s) / 242K

Publishing Venue

IBM

Related People

Hervatic, EA: AUTHOR [+4]

Abstract

A communications network may have to provide services for various traffic classes with different quality of service (QOS) requirements. In a network the importance of a connection (which relates to the connection's QOS requirements) can be expressed by a priority level. To minimize the impact of the nonstationary conditions on high-priority connections, it might be necessary for high-priority connections to be able to preempt connections of lower priorities. In such a case, a preempted connection may have to be rerouted, which in turn can cause other connections of even lower priority to be preempted. When preemption is necessary to allocate bandwidth at a link to set up a connection, a connection preemption selection algorithm needs to choose one or more candidate connections of lower priority for preemption.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 22% of the total text.

Method for Efficiently Providing Different Levels of Network Availability
in High-Speed Networks

      A communications network may have to provide services for
various traffic classes with different quality of service (QOS)
requirements.  In a network the importance of a connection (which
relates to the connection's QOS requirements) can be expressed by a
priority level.  To minimize the impact of the nonstationary
conditions on high-priority connections, it might be necessary for
high-priority connections to be able to preempt connections of lower
priorities.  In such a case, a preempted connection may have to be
rerouted, which in turn can cause other connections of even lower
priority to be preempted.  When preemption is necessary to allocate
bandwidth at a link to set up a connection, a connection preemption
selection algorithm needs to choose one or more candidate connections
of lower priority for preemption.  Described is a mechanism for
preempting connections and for efficiently selecting connections to
be preempted.

      Each connection is assigned two priority attributes:  a
requesting priority and a holding priority.  Established connections
"defend" their reservations with their holding priority.  A request
with requesting priority 'x' will preempt connections with holding
priority 'y' only when 'x' is greater than 'y'.

The following are the characteristics of the requesting and holding
priorities:

o   The Requesting priority must be less than or equal to the holding
    priority of the same connection so that connection requests do
    not "thrash".  A connection's holding priority may be much
    greater than its requesting priority.  This says:  "I don't want
    to disturb any connections in progress, but if I am established,
    then I don't want to be disturbed!"

o   It may be desirable to try alternate routes instead of triggering
    preemption of already established network connections.  Setting
    the Requesting Priority to zero guarantees that no connections
    will be preempted by the request.

o   The requesting and holding priorities assigned for a network
    connection are based on the QOS requirements (e.g., the blocking
    probability).

      Preemption is triggered when a requesting connection is unable
to secure network resources (e.g., bandwidth and labels (For the
Label Swap routing mode, a label corresponds to a table entry.)  at
the link.  Because of processing delay, requests for connection setup
and/or takedown may be queued at the link.  Takedown requests are
given higher processing priority in order to make available as much
network resources as possible to the setup requests, thus minimizing
preemption.  When a setup request is processed and not enough
resources are available, the link selects connections to be preempted
from the set of all connections currently using the link with holding
priority lower than the requesting priority...