Providing Backup and Non-Disruptive Recovery of Local Area Network Traffic Across a Connection Oriented Wide Area Network
Original Publication Date: 1997-Jun-01
Included in the Prior Art Database: 2005-Apr-01
Ogle, DM: AUTHOR [+3]
Disclosed is an algorithm for providing backup and non-disruptive recovery between multiple gateways connecting Local Area Networks (LAN) via a Wide Area Network (WAN).
Providing Backup and Non-Disruptive Recovery of Local
Traffic Across a Connection Oriented Wide Area Network
an algorithm for providing backup and
non-disruptive recovery between multiple gateways connecting Local
Area Networks (LAN) via a Wide Area Network (WAN).
disclosure addresses an effective way to allow customers
to install multiple gateways to connect LANs via a connection
oriented WAN, while minimizing the amount of administration required.
The algorithm provides for backup and non-disruptive recovery. The
algorithm is predicated on the fact that the WAN is connection
oriented and that once a connection is established between two
points, it is easier, faster, and more cost effective to use the same
path for all traffic between those two points.
The problems addressed by this disclosure are:
o Backup - How to ensure that when there are multiple gateways
serving a site and one of those gateways goes down, that the
other gateway(s) picks up the duties of the down gateway.
o Non-Disruptive Recovery - How to ensure that when a gateway
comes back online, it can start picking up its load without
disrupting any existing connections.
defines some terms. X, Y, and Z are
gateways since they all connect the same LAN to the WAN. A parallel
gateway 'set' is a set containing all the parallel gateways
connecting a particular LAN to a particular WAN. For instance, X, Y,
and Z make up a parallel gateway set. Gateways A and B are
non-parallel gateways. A parallel gateway "serves" a set of
destinations across the WAN. Which set of destinations a particular
parallel gateway serves is not important for the backup and recovery
discussion. When a parallel gateway, say X, serves a particular
destination, say M1, it means that all traffic coming from LAN 'W'
that is destined for M1, will come through X. All traffic coming
from M1, destined to any address on LAN 'W', will come through X. It
is important to note that the definition of 'destination address' is
intentionally vague since it could be the machine's address, the
network address, the port number of the application, etc.
problem addressed by this disclosure is backup.
method is necessary for one parallel gateway to determine that
another parallel gateway is down and to dynamically determine which
gateway should take over for the down gateway. The algorithm used to
do this is explained below.
It is worth
noting now that the parallel gateways may or may
not be exchanging routing information both with each other and with
the other machines on the LAN using a native routing protocol (like
Routing Information Protocol (RIP)). Use of a native routing
protocol makes it easier for machines on the LAN to determine which
gateway they should send packets to, and it helps the machines on the