Browse Prior Art Database

System for Improving the Re-Healing of Partitioned Clusters in Hierarchical Networks

IP.com Disclosure Number: IPCOM000015194D
Original Publication Date: 2001-Oct-13
Included in the Prior Art Database: 2003-Jun-20
Document File: 3 page(s) / 66K

Publishing Venue

IBM

Abstract

Problem Statement In link state routing protocols like OSPF and PNNI, the resolution of links state element conflicts during database synchronization is undesirable. It can dramatically increase the amount of routing information exchange at a time when it is already at a peak. This is particularly so in environments where conflicts are most likely occur due to the topology not being stable, namely wireless mobile networks. Wireless networks are bandwidth constrained and are easily overloaded by surges of protocol activity. Such surges can result in extended network convergence times and in extreme cases cause the protocol to thrash, thus preventing convergence completely. Consider the example of a PNNI network that uses configured clustering (peer groups), where each cluster elects a representative from its members (Peer Group Leader) to represent it at higher levels of a hierarchy (Logical Group Node). 1. The cluster is partitioned due to the failure or removal of a link. 2. Each partition of the original cluster initiates a new election and elects a representative 3. The peer group leaders in each partition generate link state elements to summarize the peer groups partitions that they represent 4. The partition is healed when a link between them is reestablished The normal chain of events on the link being reestablished is to initially exchange databases. After databases have been exchanged and a new peer group leader selected, the node which looses the election must flush the link state elements that it created at higher levels.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 3

  System for Improving the Re-Healing of Partitioned Clusters in Hierarchical Networks

Problem Statement

In link state routing protocols like OSPF and PNNI, the
resolution of links state element conflicts during database
synchronization is undesirable. It can dramatically increase the
amount of routing information exchange at a time when it is
already at a peak. This is particularly so in environments where
conflicts are most likely occur due to the topology not being
stable, namely wireless mobile networks. Wireless networks are
bandwidth constrained and are easily overloaded by surges of
protocol activity. Such surges can result in extended network
convergence times and in extreme cases cause the protocol to
thrash, thus preventing convergence completely.

Consider the example of a PNNI network that uses configured
clustering (peer groups), where each cluster elects a
representative from its members (Peer Group Leader) to represent
it at higher levels of a hierarchy (Logical Group Node).
1. The cluster is partitioned due to the failure or removal of a
link.
2. Each partition of the original cluster initiates a new
election and elects a representative
3. The peer group leaders in each partition generate link state
elements to summarize the peer groups partitions that they
represent
4. The partition is healed when a link between them is
reestablished

The normal chain of events on the link being reestablished is to
initially exchange databases. After databases have been exchanged
and a new peer group leader selected, the node which looses the
election must flush the link state elements that it created at
higher levels.

1

Page 2 of 3

The problem occurs between starting the database exchange and the
election of one of the nodes as a peer group leader. The database
exchange introduces link state elements that are conflicting and
which the routing protocol has to resolve. In the case of PNNI, a
proxy flushing process is triggered on the two nodes that are
still operating as peer group leaders. As long as both nodes
(B,F) are operating peer group leaders a stalemate situation
occurs and received link state elements...