This paper discusses the post-commercialization routing between different networks occurs by describing the possible relationships between domains and how BGP is used to exchange/propagate information from one to another. In addition, the paper describes how policies for inter-domain routing are implemented by modifying BGP entries when importing or exporting them.
The entire process is quite complex, and there are no rules per se for how providers import/export routes. I think the paper does a good job explaining the complexities by starting with an explanation of different kinds of inter-AS agreements; this helps explain why distributing routing information is not simple and intuitive. BGP is explained well, and although is a simple point-to-point algorithm, has complexities that arise mostly in import and export rules.
Clearly BGP is one of the aspects of the internet that has room for improvement. Because of all the odd complexities of the internet architecture, I'm not convinced that a point-to-point protocol is the only way. Even within a domain, having some kinds of collective decision-making would possibly be more efficient in the end; some kind of distributed data structure (with sufficient replication) such as a dist hash could make routing more robust against failures.
The paper itself was excellent for introducing the issues with interdomain routing and BGP; I knew very little of the details of the post-commercialization realities of routing.
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment