Abstract
The data transport structure of the modern networks is moving toward a model of high-speed packet/frame/timeslot capable routers/switches/cross-connects interconnected by automatically configured lambda switch capable optical core networks. One of the main challenges today is the engineering of scalable and resilient automated control planes to encompass these diverse technologies in order to dynamically perform the task of bandwidth provisioning, under quality of service (QoS) constraints, for large and multi-layered networks. In this context, a considerable number of parameters concerning resource state (i.e., not only link state) must be frequently updated across the network for distributed decision taking concerning traffic routing; and flooding may no longer be the best strategy for disseminating this information in order to synchronize these distributed databases used in scalable approaches. This chapter presents a proposal for an overlay service-oriented information plane based on distributed hash tables (DHT) for resource discovery and sharing using content-addressable networks (CAN). It also investigates deterministic reconfigurations over CAN topology for taking advantage of the small-world effect in order to reduce the number of hops needed per control plane functionalities. Analytical results indicate significant reduction in routing traffic information over the physical layer of large networks but the latency for information retrieval and updating should be a concern for future investigations.
TopIntroduction
The widespread use of the Internet in recent years has led to a multitude of new services emerging from the Internet protocol (IP) networks. For example, voice-over-IP, electronic commerce, peer-to-peer (P2P) file sharing, and IP-TV are becoming very popular among domestic users. However, the best-effort policy of today’s large networks (e.g., public Internet) does not provide the appropriate platform for forthcoming highly interactive services. Moreover, specific demands from the business community for service level agreement (SLA) contracts to support reliable mission-critical services, storage area network (SAN), and applications under real-time communications constraints will certainly pose more challenging problems to next generation networks. Although the issue of providing quality of service (QoS) beyond the current best-effort strategy has been met in limited corporative networks, QoS can no longer be separated from underlining scalability bottlenecks that have to be faced in order to meet tomorrow's network requirements, namely, i) to make possible agile, flexible, and reliable bandwidth provisioning at the data transport infrastructure across non-independent network layers; and ii) to ease the cross-layer integration of control and management planes with facilities for performing operation administration and management (OAM) tasks. These network requirements are also presented in the literature under a single flag: operation, administration and provisioning (OAM&P)
It is well known that the resilience and scalability of present-day Internet is only achieved thanks to its distributed architecture. The link state advertisements (LSA), i.e., messages that are sent by individual routers to their neighbors, are used for supporting global information dissemination in order to perform the optimal routing calculations in a completely distributed way. Reliable flooding, i.e., where an acknowledgement is sent back, is the mechanism used by the widespread open-shortest path first (OSPF) protocol (Moy, 1998). However, any change in link state must be communicated to peer nodes, and link-state database must be periodically disseminated to keep state consistency among nodes. The problem future networks will face starts with the incorporation of dynamic parameters to LSAs. The routing under QoS targets and following traffic engineering (TE) constraints requires frequent dissemination of information that is used in the decision-making process. And this dissemination of information is further complicated in multi-layer networks. Unless preventive measures are taken, this may lead to endless storms of resource state information, from different network layers, being exchanged by the flooding mechanism.
Another problem is the lack of centralization concerning resource state information. Sub-optimal routing, and even route loops, may result from the use of stale resource state while up-to-dated information is still being propagated across the network. Note that the overall routing convergence may be severely jeopardized; similarly to the route-flapping problems experienced by OSPF (Ohara et al. 2003). Finding out scalable ways of disseminating resource state and simultaneously improving information consistency in multi-layer distributed routing is one of the biggest challenges for the designers of next generation networks. Especially when a wide variety of dynamic traffic demands have to be served by diverse network stacks; each of which with large number of resources to be managed.
This chapter addresses these issues by introducing a P2P overlay service-oriented information plane to replace the link state database (LSDB) and traffic engineering database (TED) in OSPF-TE. The proposed approach is not focused on particular data plane technologies despite the fact that most of the introductory discussion will be on optical networking as it is the underlying technology of large and high-capacity future networks. The organization of the remainder of this chapter goes as follows. Next Section presents the future network context in more details. Our proposal is then put forward in Section 3, presenting the information plane built through a P2P overlay network. A deterministic small-world reconfiguration for the resulting overlay topology is outlined in Section 4. An analytical approach for performance assessment is discussed in Section 5. Results are presented in Section 6 contrasting advantages e disadvantages of our approach in comparison with a lower bound for flooding mechanism. Future research directions are suggested in Section 7 before the concluding remarks are presented.
Key Terms in this Chapter
Quality of Service (QoS): ITU-T Rec. E.800 defines QoS as “the collective effect of service performance which determine the degree of satisfaction of a user of the service.” ITU-T Rec. G. 1000 goes further in the definition of QoS: “from different perspectives: Customer’s QoS requirements; Service provider’s offerings of QoS (or planned/targeted QoS); QoS achieved or delivered; Customer survey ratings of QoS.
Constrained Shortest Path First (CSPF): The path computed using CSPF is a shortest path meeting a set of constraints. Thus it runs the shortest path algorithm after discarding those links that do not meet a given set of constraints. Examples for constraints: minimum bandwidth required per link, end-to-end delay, maximum number of hops, and minimum OSNR limits.
Lightpath: A point to point channel which the network interface cards experience as a dedicated line.
Grooming Routing and Wavelength Assignment (GRWA): In a wavelength division multiplexing (WDM) all-optical network, a traffic stream may be less than the maximum capacity of a network interface card served by a lightpath. In order to avoid assigning an entire lightpath to a small request, it is possible to decide whether combine or not traffic streams across the network while finding a solution to the routing and wavelength assignment (RWA) problem. An optimal solution is expected for a given GRWA problem, such as minimizing the cost of equipment by reducing the number or network interface cards.
Optical Signal-To-Noise Ratio (OSNR): It can be seen as the QoS at the physical layer of optical networks. OSNR is directly related to bit-error rate, which will lead to packet losses seen by higher layers.
Small-World Networks: Although most nodes are not neighbors of one another, most nodes can be reached from every other by a small number of hops because of few long reach connections.
Control Plane (CP): Physical infrastructure and distributed intelligence that controls the establishment and maintenance of connections in the network. This includes protocols and mechanisms to disseminate information as well as algorithms for engineering an optimal path between end points.
Traffic Engineering Database (TED): For a given traffic engineering domain, the traffic engineering databases must be kept updated and consistent among all routers. Its contents are used by CSPF when pruning those links not meeting TE constraints from the shortest path trees.