next up previous
Next: The Content Routing Problem Up: An Architecture for Content Previous: Abstract

Introduction

With the emergence of the World Wide Web, the primary use of the Internet is content distribution, primarily in the form of web pages, but increasingly audio and video streams as well. Some measurements indicate that 70 to 80 percent of wide-area Internet traffic is HTTP traffic. Much of the remainder consists of RealAudio streams and DNS[4, 12]. That is, almost all of the traffic in the wide area is delivery of content, and ancillary traffic to locate it. Today, millions of clients are accessing thousands of web sites on a daily basis, with relatively few popular sites supplying a large proportion of the traffic. Moreover, new popular web sites and temporarily attractive web sites can prompt the sudden arrival of a ``flash crowd'' of clients, often overwhelming the resources of the associated servers.

To scale content delivery to support these demands, more and more content providers and content hosting sites are replicated at multiple geographically dispersed sites around the world, either on-demand (i.e. caching) or by explicit replication. The problem is then to route client requests to a nearby replica of the content, the content routing problem.

In this paper, we argue that current content routing designs are unsuitable due to their closed nature and scalability limits. Section 3 describes a content routing system that forms part of an explicit Internet content layer, and we claim that this system provides better latency and scalability than current approaches. We support these arguments by an analysis of the scalability of name-based routing in section 4, and close with a description of related work, future goals, and conclusions.



Mark Geoffrey Gritter
Fri Jan 19 09:19:43 PST 2001