The paper started by identifying questions that would need to be answered. These are: How do the relationships between ASs affect the routing information they share with each other? What are these relationships? How does this affect the technical research agenda of internet routing?
The internet is composed of ASs (Autonomous Systems) that exchange reachability / routing information. These ASs are operated by a commercial entity (such as ISPs, companies, schools), which define the routing policies that map out the rules for exporting and importing routes. The policies to import, export, and select routes depend on the relationship between ASs, if it is provider-to-customer, customer-to-provider, peer-to-peer, or sibling-to sibling.
Within each AS, different Interior Gateway Protocols (IGPs) operate. These protocols are concerned with optimizing a path metric. The paper did not discuss IGPs, but was focused on inter-AS relationships and features of BGP (Border Gateway Protocol), the current wide-area routing protocol on the internet.
There are two main classifications of AS relationships. These are: transit and peering. Simply put, a transit relationship is one where financial settlement is involved. For example, the provider provides access to all/ most of the destinations in its routing tables. Meanwhile, a peering relationship involves providing mutual access to routing tables. This is also a business deal, but there is no financial settlement.
So why can’t all relationships just be peering?
There are 2 main reasons: First, peering relationships do not generate revenue. Second, peering relationships need to be renegotiated often.
Also, import and export policies are determined based on AS relationships. Export policies are important and should be determined with care because an ISP does not want to be providing transit to packets it will not be earning from. On the other hand, the most important consideration in import policies is knowing who advertised the route. As such, routes are typically imported in the following order of priority: customer > peer > provider.
The next part of the paper discussed in detail the route selection process of BGP. In essence the write-up explained how the BGP routers are configured to implement policies which are dependent on the relationships between ASs.
So what do I think about all this?
The author was thorough enough in explaining the technical details of how interdomain internet routing worked. After reading the article, I had a better appreciation of how flexible yet complicated BGP can be. And the configurations get more complex because of ‘economic relationship’ considerations. Still, it would be interesting if somebody could figure out how to make things simpler, or the business model that would do away with the transit relationship constraints and allow for sharing of the best routes with each and every AS.
Reference:
Balakrishnan,H. Feamster,N. 2001-2005. Interdomain Internet Routing
The internet is composed of ASs (Autonomous Systems) that exchange reachability / routing information. These ASs are operated by a commercial entity (such as ISPs, companies, schools), which define the routing policies that map out the rules for exporting and importing routes. The policies to import, export, and select routes depend on the relationship between ASs, if it is provider-to-customer, customer-to-provider, peer-to-peer, or sibling-to sibling.
Within each AS, different Interior Gateway Protocols (IGPs) operate. These protocols are concerned with optimizing a path metric. The paper did not discuss IGPs, but was focused on inter-AS relationships and features of BGP (Border Gateway Protocol), the current wide-area routing protocol on the internet.
There are two main classifications of AS relationships. These are: transit and peering. Simply put, a transit relationship is one where financial settlement is involved. For example, the provider provides access to all/ most of the destinations in its routing tables. Meanwhile, a peering relationship involves providing mutual access to routing tables. This is also a business deal, but there is no financial settlement.
So why can’t all relationships just be peering?
There are 2 main reasons: First, peering relationships do not generate revenue. Second, peering relationships need to be renegotiated often.
Also, import and export policies are determined based on AS relationships. Export policies are important and should be determined with care because an ISP does not want to be providing transit to packets it will not be earning from. On the other hand, the most important consideration in import policies is knowing who advertised the route. As such, routes are typically imported in the following order of priority: customer > peer > provider.
The next part of the paper discussed in detail the route selection process of BGP. In essence the write-up explained how the BGP routers are configured to implement policies which are dependent on the relationships between ASs.
So what do I think about all this?
The author was thorough enough in explaining the technical details of how interdomain internet routing worked. After reading the article, I had a better appreciation of how flexible yet complicated BGP can be. And the configurations get more complex because of ‘economic relationship’ considerations. Still, it would be interesting if somebody could figure out how to make things simpler, or the business model that would do away with the transit relationship constraints and allow for sharing of the best routes with each and every AS.
Reference:
Balakrishnan,H. Feamster,N. 2001-2005. Interdomain Internet Routing
No comments:
Post a Comment