Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Control Path Redundancy with No-Load Sharing across Routers

IP.com Disclosure Number: IPCOM000004588D
Original Publication Date: 2001-Feb-23
Included in the Prior Art Database: 2001-Feb-23
Document File: 2 page(s) / 41K

Publishing Venue

Motorola

Related People

Adam Lewis: AUTHOR [+2]

Abstract

Control Path Redundancy with No-Load Sharing across Routers

This text was extracted from a WORD97 document.
This is the abbreviated version, containing approximately 49% of the total text.

Control Path Redundancy with No-Load Sharing across Routers

by

Adam Lewis, Dave Helm

Background

In wide area, packet switched communications networks (such as a Motorola X-Zone system) customers often wish to achieve router and link redundancy in the MSO's core network. To achieve this redundancy, multiple routers capable of reaching remote sites are connected to a WAN switch within the MSO. For the MSO to communicate with the remote sites, customers lease T1/E1 connections from local carriers, and pay a recurring charge for bandwidth. Because these recurring costs can be high, a site will often times sacrifice redundancy and provide only a single physical link, with multiple PVCs traversing that link to reach the redundant routers in the MSO core.

Problem

Because physically diverse application control paths are required within the MSO core, load sharing occurs across the two PVCs. Because the two PVCs begin on two physically diverse links and terminate on one physical link, over subscription of the end-link can occur. This causes the WAN switch to back-up and not maintain TOS priority; this leads to the degradation of audio quality. To illustrate this, consider the following: assume the two redundant routers in the MSO core each forward packets reaching their full bandwidth capacity. Say for example each link in the MSO core is allocated for 64K of bandwidth. Assume the same for the site link. If each Core router were to forward packets to the site at a rate of 64K per second, then that would be a combined 128K of packets being pushed over a 64K link.

Solution

The solution to the above problem consists of a combination of multicast and unicast routing protocols, static routes, and strategic placement within the network of multicast Rendezvous Points. The solution effectively accomplishes the following: all packets flowing to and from the site are forced down a single PVC, with the exception of application level Control packets application level Control packets are made to flow down both links simultaneously; one acting as the primary control path, the other as backup.

Therefore, it is useful to explain the solution from two viewpoints: that of the application control plane, and that of everything else.

The application control plane:

In order to accomplish the use of both paths for the application control plane, control packets are unicast upstream (from the site to the MSO) and multicast downstream (from the MSO to the site). Unicasting across both paths upstream is accomplished by configuring two static routes within the site router, one per destination subnet (note that the application on the upstream end for example the Zone Control in X-Zone must be physically attached to two separate subnets). Each of the static routes is configured with a physically separate router as a next hop. Once the next hop is reached, there is only one viable n...