Fibbing: IT Routes

Today's overview covers routing innovations. Evaluation of this process in IP networks allows for a better understanding of the architectural shift proposed by the SDN (See earlier ).

There are internal (in relation to the network) and external, or inter-network routing.

The two systems use different protocols, for example, OSPF (Dynamic Routing Protocol, based on link state tracking technology, and using Dijkstra's algorithm for finding the shortest path) is widely used for internal routing, and BGP (Internet Dynamic Routing Protocol) is the standard for Internet routing. ). A key element of the network infrastructure is the router, which performs two functions: routing and forwarding packets from one interface to another. Although sometimes they understand both functions as routing, in fact, they differ significantly.

Algorithmically, packet forwarding is fairly simple, and performance is the main concern. The senders and recipients of the packets are the interfaces of the router, and the definition of the destination is carried out using the so-called transfer tables - FIB .
')
Let's take a closer look at the modern Fibbing technology, which provides direct control over the routing table of the routers using the distributed routing protocol input.

Fibbing relies on traditional protocols such as OSPF, where routers compute the shortest paths. Fibbing manages routers, eliminating the need to configure them.
This technique ensures that routers work on the calculation of master forwarding records, presenting them with a carefully constructed augmented topology, including fake nodes (providing fake end addresses) and fake links (with fake values). In essence, Fibbing inverts the routing function: taking into account the redirection records (desired result) and the routing protocol (functions), Fibbing automatically converts the routing messages into a computational form to send to the router.

Fibbing is expressive and can direct flows in any direction. In other words, she can exercise complete control; for this reason, Fibbing willingly supports advanced forwarding applications, such as:

• traffic engineering,
• load balancing
• fast switching in case of failures
• traffic management.

Based on routing protocols, Fibbing does not support fine-grained routing and forwarding policies, such as port mapping. At the same time, such a policy can be easily maintained using Middleboxes .

Fibbing scales are crash resistant. Providing invalid data to routers is a powerful but complex tool. Indeed, Fibbing must be fast in transforming the computational form of augmented topologies to avoid network failures. At the same time, Fibbing technology needs to calculate small, padded topologies, since routers have limited resources. Finally, Fibbing must be reliable and correctly handle controller errors.

Fibbing is different from previous approaches that rely on routing protocols for software routers. Previous approaches, such as the routing control platform, rely on BGP as the SDN protocol to establish forwarding rules for each destination prefix of each router. Fibbing, by contrast, uses the routing protocol implementation on routers. Thus, Fibbing can adapt the forwarding scheme of several routers at the same time, allowing the calculation of the forwarding table entries and issuing them for their own. That is, while the controller calculates routing entries centrally, the forwarding scheme is still computed in distributed mode.

Fibbing runs on existing routers. The real Fibbing prototype is already used to program real routers (Cisco and Juniper). These routers can set up hundreds of thousands of forwarding records with an average installation time of less than 1 ms per write. This provides a much larger functional scale than traditional SDN routers, without requiring the deployment of new equipment and controller actions for each device. It also means that Fibbing can implement the latest SDN versions, such as Google B4 and Microsoft SWAN - on top of existing networks. Without closing in on specific cases (for example, traffic engineering), we describe the use of Fibbing for various high-level approaches.

Fibbing provides the following benefits:

• Abstraction : Fibbing shows how to express and implement high-level forwarding requirements using a distributed protocol,
• Algorithms : Fibbing offers new efficient algorithms for computing compact topologies,
• Implementation : Fibbing can now be implemented with Cisco and Juniper products,
• Assessment : The Fibbing controller quickly generates small, enlarged topologies that cause minimal load for routers.

In the following articles, we will present a more detailed study of the structure of SDN, as well as describe the interesting observations of Cloud4Y regarding the impact of SDN on the future of wireless networks.