Multiplex communications – Fault recovery – Bypass an inoperative switch or inoperative element of a...
Reexamination Certificate
1998-06-16
2002-04-09
Olms, Douglas (Department: 2661)
Multiplex communications
Fault recovery
Bypass an inoperative switch or inoperative element of a...
C370S352000, C370S373000, C370S410000, C370S418000
Reexamination Certificate
active
06370112
ABSTRACT:
TECHNICAL FIELD
The present invention relates generally to packet communication systems and telecommunication networks. In particular, the present invention relates to rerouting or reorganizing the path of a connection within a connection-oriented packet network such as an asynchronous transfer mode (ATM) network or a frame relay network.
DESCRIPTION OF THE PRIOR ART
Modern digital communications often use packet switching connection oriented networks of which Asynchronous Transfer Mode (ATM) networks are an important example. In packet switching connection oriented networks, a user digital data stream is broken into discrete units (for example of 48 bytes each). Each such unit is transmitted in a packet (or cell in the case of ATM). Attached to the front of the user data portion of the packet is certain header information. This header information includes fields used by the switches of the packet switching network to identify the correct rule for switching the packet; that is, to move the packet from the particular links or transmission facility from which the packet enters the switch to the particular transmission facility upon which the packet will be transmitted out of the switch. In the context of an ATM network, the VPI (Virtual Path Identifier) and VCI (Virtual Channel Identifier) fields (sometimes only the VPI field is used) in the ATM cell header (see
FIG. 1
) provide this functionality. Before transmitting the packets out of the switch, the switch generally modifies these header field values to align with the control data that will be used by the adjacent downstream switch to further switch the packet or to align with the expectations of the recipient of the data. Since this invention relates to connection oriented networks, all packets belonging to the same data stream, for example, the same telephone voice call, same fax transmission, or same data file download, will all traverse the same path through the network. That is, all packets belonging to the same data steam or call will traverse the same set of switches, use the same links between adjacent pairs of those switches, and upon each such links, use the same header index values. In support of the connection over the path, each participating switch will have stored rules that tell it how to map a packet arriving on a particular facility to a particular out-going facility and how to translate the relevant header routing indices. These rules are determined and come to be stored in the switches along the path of the connection by means of either signaling messages or via provisioning management, or some combination of the two.
To make the explanation more specific, we will refer to ATM connections. Hence packets are referred to as cells and the header routing indices are referred to as the VPI/VCI values. The applicability of the invention to packet switched connection oriented networks other than ATM networks is straightforward.
In a connection oriented packet network such as an Asynchronous Transfer Mode (ATM), or frame relay network, a connection (e.g., an ATM Permanent Virtual Circuit) (PVC) may need be moved to a replacement path despite the fact that the existing path is operational. This invention provides a means for executing that move to a replacement path without the loss, duplication, or mis-ordering of packets (cells). We first explain how the need to move a connection from a working path to a replacement path might arise. Next we explain how differences in cell delay along the original versus replacement path make it difficult to guarantee that the change in path does not cause lost, duplicate, or out-of-order cells.
It may happen that a network administrator wants to move a working connection from an existing path to a some other path, herein termed the replacement path. There could be many motivations for doing this. For example, the existing path may be undesirable because it involves excessively many links or uses congested links. (An example of how such a situation could arise is as follows: Assume the connection was initially created on path A. Due to link failure on path A, the connection needed to be moved, on an emergency basis, to a replacement path B. Given the emergency situation, the network could not expend the time to find an optimum path, and accepted a path B that may be excessively long, congested, or some other way inconsistent with the connection's quality of service expectations or proper allocation of network resources among connections. By and by, the failure that precipitated the emergency change in path from A to B is corrected. Then, the network administrator may wish to return the connection from the use of path B to A or to some third path that is superior to B. This change from the working path B to a replacement path generally need not be done on an emergency basis, but should be done “sooner or later” to improve the performance of the connection or reduce the network resources (e.g. number of links) allocated to the connection.) On the other hand, the motivation for moving a working connection to a new path may be simply to better balance the load on the network.
Solution
Applicant will first analyze the problem in order to derive the proposed solution.
In the context of moving a connection from one path (the existing path) to another path, (the replacement path) we refer to the switch at which the two paths diverge as the originating switch, and we refer to the switch at which the two paths again converge as the destination switch. This terminology assumes a focus upon one direction of cell flow. The methods we discuss for changing a connection's path are described in terms of only one direction of flow. However, the actual implementation would apply these methods to each direction of a bidirectional flow with the role of the “originating switch” and “destination switch” reversed. Note that the replacement and original path may share some common links (i.e. facilities) , but over those links different header indexes (VPI/VCI for ATM) would be used to distinguish cells assigned the respective paths.
To understand why moving a connection from one path to another may result in lost, duplicated, or out-of-order cells, first note that the time it takes a cell to leave an ATM switch and travel along a path involving many ATM switches until it arrives at the destination ATM switch will depend upon the path taken between the originating and destination ATM switch. If we change the path taken (perhaps changing the set of intermediate ATM switches, perhaps only the links used between adjacent ATM switches, perhaps some combination of the two kinds of changes), the delay experienced by cells will generally change. Two elements contribute to this difference in delay. The first is propagation delay, the time cells spend traversing links, since the paths may differ in the number of links traversed and the length (and/or propagation speed) of those links. The second is buffering delay, the total time cells spend stored within (not between) the switches along the path. The paths may differ in buffering delay, which is a function of the cell flow of many connections besides the one of interest to us, encountered within the ATM switches involved in the connection. Even if the two paths use the same set of switches, different queues may be used in those switches for the different links used in the two paths. To further complicate the situation, we may not be able to definitively label one of the two paths as the faster of the two, since the delay in particular queues can change moment to moment. The delay experienced in a queue depends upon the characteristics of the several connections using that queue, and different queues serve different sets of connections with different traffic patterns. So the identity of the “faster path” may change from moment to moment if the variation in buffering cell delay is large relative to the difference in transmission delay between the two paths. Note that the invention described herein does not rely on any particular relationship between pat
Hom Shick
Ulrich Werner
LandOfFree
Seamless path switchover in a connection-oriented packet... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Seamless path switchover in a connection-oriented packet..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Seamless path switchover in a connection-oriented packet... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2899485