QoS Routing Solutions in MPLS Networks

QoS Routing Solutions in MPLS Networks

The main challenges the Internet currently faces are all related to its increasing size on different levels. First there is the exponentially increasing number of hosts which evolves at surprising rates (i.e. in July 1993 it included around 1.7 million hosts, in July 1998 it already comprised more than 36 million hosts, while the latest statistics show that in January 2010 it had more than 732 million hosts). Second there is its increased complexity and importance. That is because the Internet serves as the common ground for the convergence of all communication types (i.e. voice, video and data). Numerous solutions and technologies were proposed as a response to the challenge of Internet’s rapid growth in number of users and increasing requirements for service quality, reliability, and efficiency. The research described in this book encompasses five of such technologies which are major areas of current research in network communications: network routing, Quality-of-Service (QoS), Multi-Protocol Label Switching (MPLS), active networks and mobile software agents. The focus of the present book is to introduce optimal solutions for implementing efficient QoS routing strategies for large networks. This work has two aims:

– to minimise the overhead introduced by QoS in the routing process;

– to identify improved approaches to the multi-constrained routing problem.

Deploying QoS routing schemes in the legacy Internet can be very difficult. MPLS facilitates this by its separation between the control and forwarding planes. Moreover, traditional QoS routing schemes which use source routing generate considerable overhead, i.e.:

– communication overhead, as routing state information has to be disseminated more frequently and the state information is more consistent (e.g., more metrics can be considered) than in legacy/nonadaptive routing;

– computational overhead, as the entire path is centrally computed at the source node using the “collected” global state;

– storage overhead, as all the state in the network has to be stored on every node that participates in the routing process.

The approaches presented in this book use distribute routing implemented by using modern techniques such as active networks and mobile agents. This will alleviate the storage and the computational overhead requirements. Also, such distributed routing algorithms use only local states, thereby eliminating the communication overhead generated by dispatching routing information. However, by using mobile agents, the mobile agent population generated to search multiple feasible paths can itself generate a significant communication overhead. The active network approach, although it does not generate as great a communication overhead, can find only a single path, which might not be the optimal one. The preferred choice for implementing QoS routing protocols is, in this author’s view, to use mobile agents. However, there are situations when active networks prove to be a very useful tool. To exploit the potential of active networks the author proposes the integration of MPLS and active networks in order to overcome the inability of MPLS to perform switching above layer two in access areas, where such operations are needed. QoS routing for large networks is still a research area. The author proposes a new hierarchical QoS routing protocol, called Macro-routing, which is able to find multiple end-to-end paths. The use of local states allows the Full-Mesh aggregation method to be used. This is recognised by the research community as being the most accurate method available but is considered to be impractical because it generates too much overhead in the form of state advertisements. Using mobile agents, such advertisements are not needed as the routing information is consulted “in-situ”. The author also proposes to use mobile agents not only for the routing purposes, but also for resource reservation and path setup. Finding paths based on multiple constraint is an NP-complete problem. Many approximations and heuristical solutions have been proposed by the research community. For hierarchical multi-constraint routing the problem is even more complex as it is required to select a single path between any pair of border nodes for building the aggregate representation. Therefore, the author proposes a new aggregation method, called Extended Full Mesh (EFM), which is a compromise between Full-Mesh aggregation and no aggregation at all. This technique stores the cost of more than one path between the two border nodes when performing aggregation, which considerable increases the chances of finding a final multi-constraint path.

The book is organised as follows.

Chapter 1 introduces the five main topics that underpin the research presented in this book, i.e. routing, Quality of Service (QoS), MultiProtocol Label Switching (MPLS), mobile agents and active networks.

Chapter 2 presents the main QoS routing challenges and existing solutions. Topology aggregation and efficient distribute routing mechanisms are the preferred approaches to overcome the main QoS routing problems by improving the scalability and reducing the computational overhead respectively.

Chapter 3 compares active networks and mobile agents as techniques to implement QoS routing and presents some application examples for which active networks prove suitable. This chapter concludes by arguing that mobile agents are the preferred choice for implementing efficient QoS routing mechanisms.

Chapter 4 proposes a new hierarchical routing protocol, called Macro-routing, which reduces the computational and storage overhead introduced by QoS routing, while also finding multiple feasible paths. This comes at the price of a potentially large communication overhead. Thus, techniques for limiting this communication overhead are developed.

Chapter 5 proposes a new aggregation method, called Extended Full Mesh (EFM), which is able to obtain better results while searching for multi-constrained paths by using Macro-routing. Path selection mechanisms are also presented and compared through simulations.

Lecturer, PhD Sanda-Maria Dragoş
Cluj-Napoca
12 October 2010