Designing Fault-Tolerant TCP/IP WANs
Using the addressing scheme depicted in our diagram, we could define the default gateway for hosts as 193.1.1.6. The process for one of these hosts to deliver a packet to a remote client PC, for example, 200.1.1.1, would be:
· The destination network is not 193.1.1.0, therefore the host will send the packet to the default gateway.
· The ARP table will be referenced to determine the MAC address of the default gateway.
· A packet will be formed with the destination IP address of 200.1.1.1 and destination MAC address as that of the default gateway (in this instance, the phantom router MAC address).
Router 1 and 2 will continually exchange HSRP "hello" packets, and in the event
of the active router becoming unavailable, the standby router will take over routing packets addressed to the MAC address of the phantom router with no change in configuration or ARP tables of the hosts.
While IRDP and HSRP provide a means for hosts to have fault tolerance in the case of a central router failure, you should be aware of the issues involved with these protocols. If you do implement HSRP, the active and backup router should have identical routes to all destinations. This means that the backbone links to the central site must be duplicated (which can be costly). Also, during normal operation, traffic to and from the central site will pass through one router--so those expensive backbone links installed to the standby router are unused.
Designing the Distribution Centers
Distribution centers link the backbone network to remote branch connections. The backbone connections usually are made via dedicated T1 or fractional T1 lines. To connect the remote branches to the distribution c
enter, channelized T1 services may be used.
A channelized T1 can supply 24 separate channels, each of which can be groomed by the carrier to individual 64-Kbps circuits that are terminated in the remote branches. With a channelized T1, a single T1 into a
router can connect 24 remote branches. This type of connectivity provides greater reliability and makes troubleshooting simpler than that required for individual lines, with their associated CSU/DSUs and cabling.
Beyond connecting the remote branches to the backbone, the distribution centers also may be responsible for dial backup.
Designing Dial Backup
There are two major decisions to be made regarding dial backup solutions. You must choose a centralized or distributed solution and you must decide which device at the remote branch will initiate dial backup.
The centralized or distributed question is a complex one. A centralized pool of ports for remote-branch equipment to dial into is easier to manage and more efficient. However, a
centralized configuration has implications for the routing protocols used on the network. If a distance vector routing protocol (such as the Routing Information Protocol or IGRP) is being used, the routing tables are subject to route summarization; this means that to maintain proper routing, there can only be one physical link from devices configured for one network number to devices configured with a different network number.
With a central pool of ports and a distance vector routing protocol, remote branches must use subnet addresses allocated from one network number. A central pool is often preferred to a distributed pool, since it provides a means of bypassing the WAN in the event of a catastrophic WAN failure.
A distributed pool of dial ports is more complex to administer and troubleshoot from one geographic location. However, it does allow route summarization to work for you instead of against you. With pools of dial ports in each distribution center, you can address each distribution center wi
th a different network number, thus reducing the size of routing updates on the network. The result is that only the network number used for each distribution center is advertised on the backbone network, not all subnets in each distribution center.
At the remote
branches, you can have the router, a separate ISDN Terminal Adapter or a CSU/DSU initiate dial backup. The first two options are similar in terms of router configuration. Essentially, you configure the router to initiate an ISDN connection if the carrier signal from the CSU/DSU is down for more than 15 seconds and to remain on the ISDN connection until the carrier signal from the CSU/DSU is up for more than 45 seconds, for example. The third option has the CSU/DSU connected to both the leased line and the ISDN backup connection. The router does not "know" anything about the CSU/DSU initiating ISDN backup.
We prefer having as much equipment built into the router as possible. This type of solution simplifies network management and means that o
perators only need to learn one user interface to obtain device status and alter configurations.
Chris Lewis is vice president of international operations at ILX Systems in New York. He can be reached at chrisl@ilx.com.
|
REPORTS
ANALYTICS
Follow 
