Introduction

With the expectations of Internet users for content, quality, and performance constantly taxing the resources of even the largest of Web infrastructures, even thinking of downtime isn’t an option. Estimates in the hundreds of millions of dollars in lost value have demonstrated that web site and application downtime can be asdamaging as a successful hacker attack.

In order for Web infrastructure to deliver on its promise of increased productivity and increased profitability, being impervious to failure is a requirement. Accomplishing this feat, however, is a non-trivial task that typically requires a large sum of money. Redundant connectivity, redundant hardware, and a well designed architecture take time, highly qualified (and thus highly paid) IT staff, and of course a lot of hardware.

This paper begins by taking a quick look at the history of providing high availability on the Web. By examining first and second-generation solutions, we learn from past mistakes and proceed into our proposed solution: providing high availability using third-generation Application Front End (AFE) appliances.

A Brief  History

First generation Web infrastructure had a simple goal: Just Make It Work. At the time, elaborate hosting options, Web server choices, and networking options were limited at best. Typical networks carried light enough traffic that a single Unix server could easily handle the load. The problem with many of these networks was that they did not provide a scalable or highly available architecture.

In order to solve the first round of problems associated with availability, a DNS based solution was put into place. Simply put, a single hostname would point to multiple IP addresses. By doing this, clients would cycle

through the choices in a round-robin fashion thereby distributing the load across multiple servers. This technique is aptly called “Round- Robin DNS”.