Few days ago we were pleased to learn that another paper on the Recursive InterNetwork Architecture (RINA) was accepted. The paper titled “Bounding the router table size in an ISP network using RINA” and co-authored by J. Day, E. Grasa, M. Ponce de Leon, P. Phelan, S. Bunch, L. Chitkushev, I. Matta, L. Pouzin and myself has been accepted as an invited paper at the Second International Conference on the Network of the Future (NoF 2011).
Our goal in this paper was to highlight the some of the advantages of RINA for ISP networks and in particular focus on how the architecture improves routing efficiency and manages to bound the router table size in a provider network.
Let’s start by giving an example of a typical configuration of a provider network in RINA (Fig. 1).
As we can see there is a top level DIF formed, named T-DIF, which can provide end-to-end service to hosts. T-DIF acts like the traditional legacy architecture until the first border router. The border router determines the next hop and multiplexes the traffic to the lower layer DIF which encapsulates traffic in a new routing domain in a lower DIF (L-DIF), where traffic is routed in the normal way. To the T-DIF, the L-DIF is a single hop.
An idealized top view of the provider network we saw in Fig. 1 would be similar to the image illustrated in Fig. 2. There is a necklace of subnets around the central hole. Each subnet is not a single DIF but a subset of the top-level DIF (T-DIF).
Keeping in mind the typical configuration of a provider network, let us see how RINA achieves significant improvements in routing efficiency. First of all, addresses in RINA are private, meaning that the addresses are internal to the DIFs and not global. This results in reduced storage requirements for routes as the length of an address and thus the length of a route is reduced.
If a hierarchical addressing strategy is adopted so that the address hierarchy reflects the adjacency of the subnets, then we can significantly reduce the number of routes we need to store. Specifically, the routing in the T-DIF would only have to store routes to the border routers of either adjacent subnets or the hole. A border router at the edge of a hole can determine where to forward a Protocol Data Unit (PDU) based on the address only. So, no route calculation is necessary. Moreover, routes only need to be computed within the subnet. This drastically reduces the number of routes and the length of the routes that need to be stored.
RINA’s breakthrough here is that under the wise choice of the addressing schema, it allows the network designer to effectively bound the number of routes that need to be stored at each level. This number is determined by the number of elements within a subnet and not the number of elements in the whole network.
This work will be presented during the first day of the 2nd International Conference on the Network of the Future that will be held in Paris, France on Monday, 28th of November (check the conference program). We are always interested in meeting people interested in RINA! 🙂 So, in case you can make it, we would be very happy to meet you, discuss and respond to questions you might have. If not, keep in mind you can always contact us via e-mail, me (eleni.trouva AT i2cat.net) or Eduard (eduard.grasa AT i2cat.net).
At last, this paper was for us a good start!… 😉 Currently, we are in the process of exploring the benefits that RINA brings to ISP networks concerning security, performance and manageability issues.