System information

Chapter 4. ATM LAN backbone interoperability and migration

In this chapter we consider a typical scenario of an installation based on an

IBM 8265 ATM backbone. In all probability, this backbone operates well, there

are no bottlenecks, and it scales well by adding more nodes and/or edge

devices when more capacity is required.

The ATM core of such a network is rarely saturated, and we may expect that

the life of ATM switches will extend several years from today. However, the

market direction is clearly to use Gigabit Ethernet at the edges of the network

to attach servers and workstations.

The IBM-Cisco alliance is an opportunity to position the ATM backbone for a

future migration. We discuss a migration to a Gigabit Ethernet backbone,

which we consider most likely today, but certain requirements may mandate

that an ATM core is maintained even in the future.

In the last part of this chapter, we describe in some detail how we added a

Cisco Catalyst 5500 LANE Module to an Ethernet ELAN as an MPOA client to

our laboratory network and established short-cut routing between the

Catalyst 5500 and an IBM 8270 token-ring switch with an MSS client uplink.

4.1 ATM LAN backbone migration scenario

This migration scenario consists of three steps:

1. First we take a look at what may be characterized as a typical ATM LAN

backbone.

2. Then we develop an interoperability solution that provides greater

flexibility in connecting servers and workstations. We base the solution on

the overlay principle. This way we enable an orderly migration to an

Ethernet backbone.

3. Finally, we discuss when or whether it is desirable to eliminate the ATM

backbone.

4.1.1 A typical ATM LAN backbone

The network depicted in Figure 44 on page 68 may solve many requirements

for an organization. Some of them are listed below. Undoubtedly this

versatility has been the reason for many customers to choose such a solution.

The network has been simplified to better illustrate the basic principles. A

real-life network will often consist of considerably more equipment including

several ATM backbone nodes, and servers are often distributed to the