Chapter 6. Architectures for Reliability

Table of Contents
Introduction
How Will Address Space be Assigned to Networks?
How Many Sites Should Have Internet Connections?
How Many Internet Connections are Required?
How Many ISPs are Required?
How Many Border Routers are Required?
What Route Subset Will You Request on Each Connection?
Private Networks
Border LAN Architecture
How Do Firewalls and NAT Interact with Reliable Internet Connectivity?
Budgeting: What Will Alternative Architectures Cost?
Architectural Transitions

Architectural decisions establish the foundation of your network. Such decisions will place limits on the level of reliability you'll be able to achieve and the amount of money you'll have to spend. They'll also have the largest bearing on the amount of effort and calendar time required to achieve reliable Internet connectivity.

   Purpose
      Readers won't be shown . . . 
      Illustrations will be used heavily to show alternative
        architectures and compare their costs and benefits
      Contrast to following chapter:
        This one covers architectural decisions that may
        take a lot of time, money, or both to revisit later
      

Introduction

Remember: improving reliability very often means increasing complexity. That complexity will cost money. It may even reduce your reliability at first.

How much reliability is enough? That often depends on what it costs. The mission of this chapter is to help you understand the incremental steps of improving reliability and their approximate cost.

One common technique for improving reliability is to get two of everything and configure all equipment to operate in "hot spare" mode. This is often called "duplexing." While there are no failures, load is shared across all components. If a component fails, the entire load is carried by the working equipment.

     Weakest link in the chain
      Need to allow for Cold spare environments
      Accommodate primary and backup terminology
      So why not have Internet connections from two different sites?
        In short, it may not be worth it
        Depends on
          Ease of replicating your service
          Vulnerability to "Backhoe Joe"
          Others below
    

Copyright © 1999-2000 by Robert A. Van Valzah