Central Control over distributed Routing

Central Control over distributed Routing
2.Flexible fibbing
3.Augmenting topology
4.Implementation
5.Reaction to Failures
6.Frequently asked Questions
7.Related work
8.Conclusion

ABSTRACT
The central routing option increases flexibility, but compromises the robustness of distributed protocols. This article presents the fibbing (Lie), an architecture that implements central control over distributed routing. Lies with a potential link-state routing protocol to spoof nodes and links, allowing routers to calculate their forwarding based on an increasing topology. Lies are expressive and support flexible shunt balancing, forwarding engineering, and routing support. Based on advanced forwarding requirements, the Lie controller calculates a compact, growing topology level and injects spoofed elements through standard routing protocol information. The lie can be compatible with any non-modifiable router that supports the OSPF protocol. Our experiments also show that it can increase network size by forwarding requirements, introducing minimum limits, quickly reacting to network conditions, and controller errors.

Introduction
As a network administrator, suspect that a denial-of-service attack (Dos) L has occurred on the network, to check for the network, you need to do:
1. The orphaned target is a stream of these IP addresses
2. Give them directly to the scrubber, if necessary, to clear them.
3. Use a path-balanced approach to reduce congestion by using unused links.

These jobs are very difficult to do in traditional networks.
Because middleware and destinations are far away, many device configurations need to be modified, and correcting an error can change the rules so that errors occur elsewhere.

The software-defined network can easily solve these problems, because it can be centralized and directly control the forwarding behavior. However, removing the distributed routing protocol can cause some problems.

Scalability and reliability are degraded, and because all work is done by the controller, the distributed routing protocol can spread the work. For scalability and reliability, the SDN controller must be geographically distributed repeatedly, but this creates the challenge of managing its state. And the existing network is hard to transition to SDN, so SDN is now only a small-scale deployment.

The fibbing cheats the router, letting the router think that there are some nonexistent nodes or links in the link, based on a link-state protocol like OSPF or Is-is, to let the router calculate the forwarding path.

They did the following work:
1. Abstract: Demonstrates how to express and implement advanced forwarding requirements by manipulating Distributed link state routing protocols.
2. Algorithmic aspects: A new and efficient algorithm is proposed to calculate the compact growth topology.
3. Implementation aspects: Describes a complete implementation on the unmodified Cisco and Juniper routers.
4. Evaluation: Demonstrates that the spoofing controller quickly generates small topologies and paths on the router.

Conclusions
The advent of SDN illustrates the need for higher programmability and centralized management of the network.
In this paper, fibbing shows how to centrally and automatically control forwarding without abandoning the benefits of a distributed routing protocol.
Fibbing are expressive, extensible, and can work with existing routers.
Future work:
Focus on extending the IGP protocol to achieve better control through fibbing.
In an abstract way, fibbing shows how to combine centralized and distributed methods.
We believe that new research can further explore this direction, for example, by selecting a further study of centralized or distributed component research.

Central Control over distributed Routing