Multiplex communications – Network configuration determination – Using a particular learning algorithm or technique
Reexamination Certificate
1999-10-12
2004-08-03
Pham, Chi (Department: 2663)
Multiplex communications
Network configuration determination
Using a particular learning algorithm or technique
C370S254000, C370S401000
Reexamination Certificate
active
06771610
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to network protocols and network intermediate devices executing such protocols; and more particularly to algorithms for managing the tree of network devices for a data network according to a spanning tree protocol.
2. Description of Related Art
Local area networks (LANs) specified according to Institute of Electrical and Electronic Engineers (IEEE) Standards for Local and Metropolitan Area Networks under section 802.x of all types may be connected together with media access control (MAC) bridges. Bridges interconnect LAN segments so that stations connected to the LANs operate as if they were attached to a single LAN for many purposes. Thus a bridged local area network provides for interconnection of stations attached to LAN segments of different MAC types, for an increase in the physical extent, the number of permissible attachments and the total performance of a LAN, and for the partitioning of physical LAN support for administrative or maintenance reasons: The MAC bridge is specified according to the IEEE standard 802.1D (IEEE Std 802.1D-1990, IEEE Standards for Local and Metropolitan Area Networks: Media Access Control (MAC) Bridges.) The protocol has application for establishing interconnection of devices on network segments (whether the segments are characterized as LANs or as other network constructs) in any type of network.
When a bridged network is established, it is possible to create loops in the network by providing more than one path through bridges and LAN segments between two points. Thus, according to the 802.1D standard, an active topology for the bridged network is maintained according to the spanning tree protocol which is described in the standard. The spanning tree protocol automatically establishes a fully connected (spanning) and loop-free (tree) bridged network topology. It uses a distributed algorithm that selects a root bridge and the shortest path to that root from each LAN. Tie breakers are used to ensure that there is a unique shortest path to the root, while uniqueness of the root is guaranteed by using one of its MAC addresses as part of a priority identifier.
Every LAN in the network has one and only one “designated port” providing that LAN's shortest path to the root, through the bridge of which the designated port is a part. The bridge is known as the designated bridge for that LAN.
Thus, bridges other than the root bridge at the root of the network can be termed a branch bridge. Every branch bridge has a “root port” which is the port providing that bridge's shortest path to the root. Ports other than the root port are designated ports, or alternate ports according to the standard. An alternate port is connected to a LAN for which another bridge is the designated bridge, and is placed in a blocking state so that frames are not forwarded through that port.
The frame forwarding path through any bridge is thus between its root port and designated ports. When spanning tree information has been completely distributed and is stable, this connectivity will connect all of the LANs in a loop-free tree.
When a bridge first receives spanning tree information that dictates new connectivity through that bridge, it does not establish the new connectivity immediately. Ports that were connected previously as either the root port or a designated port, but are no longer connected, are immediately made blocking. However, the transition to a forwarding state of ports that were previously not connected in a forwarding role is delayed. The delay is needed because:
(a) Frames forwarded on the previous topology may still be buffered by bridges in the network. Thus an instantaneous change to the new topology can cause these to be forwarded back to their LAN of origin, causing duplication of the frame once.
(b) New spanning tree information in the network may not have been fully distributed yet. Thus an immediate change to a new topology may cause temporary loops. These loops could generate high traffic volumes, disrupting end stations, causing frame loss in bridges, and possibly delaying the propagation of spanning tree information further.
The first of these two reasons is far less important than it once was, because the protocols prevalent on LANs today deal with immediately duplicated frames. Some old implementations of LLC type 2 will reset connection under these circumstances, but they are no longer in widespread deployment. Thus the problem presented by reason (a) is of less significance than reason (b).
Reason (b) continues to be a fundamental problem to the spanning tree configuration.
According to the spanning tree protocol of the standard, each port on a bridge can assume a blocking state in which frames are not forwarded through the port, or a forwarding state in which frames are forwarded through the port. For a transition from the blocking state to the forwarding state, the protocol requires the port to proceed through transitional states referred to as the listening state and the learning state. In the listening state, the port is preparing to participate in frame relay, however frame relay is temporarily disabled to prevent temporary loops. In the listening state, the port monitors bridge protocol data unit (BPDU) frames or other information related to the topology in the network for an interval referred to as the forward delay timer. If no information is received which causes a change in state of the port before expiry of the forward delay timer, then the port transitions to the learning state.
In the learning state, the port continues to prepare for participation in frame relay. The relay is temporarily disabled to prevent loops. In this state, in addition to monitoring BPDU frames and other information related to operation of the spanning tree algorithm, the port learns information about end stations that are accessible through the port for use in the forwarding of frames once the frame enters the forwarding state. Upon expiration of the forward delay timer in the learning state, if no better information about the protocol is received, then the port assumes the forwarding state. Thus, the transition from a blocking state to the forwarding state takes two times as long as the forward delay timer interval. From the moment of detection of a change in topology which causes a transition from the blocking to the forwarding state, until the moment that the forwarding state is assumed, can be a significant amount of time, as much as 20 to 50 seconds in some cases. Thus, when a link or switch fails, reconfiguration takes place at unacceptably slow rates for mission critical networks. Significantly reducing this recovery time remains a problem.
Three approaches to managing reconfiguration times include the following:
(1) spanning tree timer values can be manually configured for optimal values;
(2) a scheme known as “backbone fast” detects the changing topology and allows a bridge to determine whether or not connectivity in the network has been lost, by sending a test packet called a root link query PDU to the bridge in the network that is the root in the spanning tree protocol; and
(3) network topologies can be specially designed to provide fail over in some cases without requiring expiration of the relevant timer, as has been described in the above cross referenced application Ser. No. 09/141,803.
Managing timers according to the first approach listed above is error prone, and negates the low administration benefits of the standard spanning tree. Further, the timers must be set to values that are a worst-case to some very high probability, so the first approach listed above provides limited improvement.
The “backbone fast” scheme of the second approach depends on a particular bridge being the root bridge, so the scheme cannot be introduced into a network by upgrading arbitrary pairs of bridges. If the network topology is changing, the root link query message to the root may not reach the actual root and cause the wrong initial steps to be taken in the effort to speed
3Com Corporation
George Keith M.
Haynes Mark A.
Haynes Beffel & Wolfeld LLP
Pham Chi
LandOfFree
Spanning tree with protocol for bypassing port state... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Spanning tree with protocol for bypassing port state..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Spanning tree with protocol for bypassing port state... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3360092