1.4 Interdomain Multicast Routing
For years multicast has enjoyed niche success in many financial and enterprise networks. Financial institutions have applications, such as stock tickers, that require sharing the same data across the network. Using unicast for these applications is inefficient and not cost effective. Likewise, some enterprise networks serve companies with applications ideally suited to multicast deliveryfor example, a central headquarters that must feed hundreds of branch sites with price lists and product information. Transferring these identical files to all sites individually with unicast simply is not efficient.
In the past, enterprise networks have frequently looked much different than the networks managed by Internet service providers (ISPs). This difference existed because these networks had to meet a set of radically different requirements. Enterprise networks connect the offices of a single company, which often involves transporting primarily a single type of data (for example, file transfer). Transporting only a single type of data enables the network to be built in a way that optimizes delivery of that type of traffic. Also, few, if any, of the routers in an enterprise network connect to routers controlled by another entity.
ISP networks couldn't be more different. ISPs can have up to thousands of different customers, each a separate administrative entity. The data can include an unclassifiable mix of voice, video, e-mail, Web, and so on. Providing ubiquitous support for these various traffic types across the interdomain world of the Internet has always set ISPs apart from enterprises in the way they are designed and operated.
Unicast and multicast routing on enterprise and financial networks has often involved deploying protocols and architectures that best meet the needs of the companies they connect. These protocols and architectures often do not address the scalability and interdomain requirements of ISPs. However, recent trends have shown that the networking needs of enterprises have evolved to more closely resemble those of ISPs. Accordingly, many enterprise networks today are beginning to use the same principles and philosophies found in the engineering of ISPs' networks, albeit on a smaller scale.
The focus of this book is to describe the technologies and challenges faced by ISPs when deploying and operating multicast across the Internet. The first reason for this focus is neglect. Most networking books concentrate on enterprise networks rather than the unique demands of service provider networks. Second, ISP networks generally possess the superset of requirements that are found on other types of networks. For example, financial networks typically need to support many-to-many applications. Other enterprise networks may need to support only one-to-many applications. Because ISPs may be delivering service to both types of networks, they must be equipped to handle both types of applications. Additionally, ISP networks have scalability demands that are rarely found on any other types of networks.
While ISPs continue to have unique requirements for scalability and interdomain stability, most of the same multicast technologies found in ISP networks can be applied for use on other networks. By adopting these ISP philosophies, financial and enterprise networks are capable of ubiquitously supporting all types of multicast traffic. This flexibility enables a network to be prepared if traffic types change in the future.
The scope of this book is confined to the protocols and technologies currently used in the production networks of service providers. In order to provide a pragmatic examination of the challenges faced by ISPs today, little to no mention is made of protocols that have not been implemented by routing vendors or deployed by service providers at the time of writing. Accordingly, IPv6 is outside the scope of this book.