Scalable Signaling Underlay for Overlay Networks Yangcheng Huang, Saleem N. Bhatti Y.Huang@cs.ucl.ac.uk, S.Bhatti@cs.ucl.ac.uk Networks Research Group, Department of Computer Science University College London, Gower Street, London WC1E 6BT, UK Abstract - This paper presents the design of a scalable decentralized signaling underlay infrastructure, which features a DHT based management information storage and query-based state lookup mechanism. The signaling underlay is aimed to apply a decentralized “peer-to-peer” style searching and discovering engine into the management and control plane of the overlay network, including grid networks and p2p applications, to facilitate deployment of QoS service. 1. INTRODUCTION Over the years, there has been an increasing desire that the Internet be used to carry data flows with specific QoS constraints. However, this requires all network elements on a path to cooperate to provide such a “better-than best-effort” service. The networks would require a tremendous amount of state information to provision, maintain, validate, and bill for these new services, across heterogeneous networks and devices with different management capabilities. Such a requirement demands more research effort on network management and control issues, especially on signaling between elements, to control and distribute network state. This paper presents the design of a scalable decentralized signaling underlay infrastructure, which features distributed hash table (DHT) based management information storage and query-based state information lookup mechanism. This proposed signaling infrastructure applies a decentralized peer-to-peer (p2p) style searching and discovering engine into the management and control plane of the overlay network, including grid networks and p2p applications, to facilitate deployment of QoS services. The paper is arranged as follows. In section 2, we discuss the design principles of the signaling underlay. Then in section 3 we present details of the proposed infrastructure and related mechanisms. In section 4, we give some implementation details, and in section 5, we show some preliminary experimental results. We introduce related work in section 6, and draw conclusions and list future work in section 7. 2. SCALABLE SIGNALING UNDERLAY In this section, we present motivating factors for scalable signaling functionality, and from them extract requirements and design principles for a scalable signaling underlay. 2.1 Problems with overlay networks In recent years, peer-to-peer overlay networks have received much attention, especially in decentralized data-sharing and discovery ([2] [4] [8] [9]). The network overlay abstraction provides flexible and extensible application-level management techniques that can be easily and incrementally deployed despite the underlying network. However, there are several issues to be addressed. Despite performing suitably in scalability and deployment, overlay solutions make data-sharing and state management more complex, resulting in additional complications in developing such systems. The complexity of managing a network increases dramatically as the number of services and the number and complexity of devices in the network increases. Also, although an overlay network may facilitate end-to-end management and control, it still needs efficient skills to manage underlying network resources. 2.2 Design Principles Generally, the signaling underlay should be applicable in a very wide range of scenarios and at the same time be simple in implementation and lightweight in resource consumption. This section will analyze such requirements in detail. The design principles include: • Decentralized architecture. Traditionally, the common framework based on a centralized approach is straightforward but not proved to scale well; although a distributed approach with a hierarchical architecture appears flexible, conceptually, it scales only in number of nodes, not at the network level. That is, inside network domains, a distributed framework may work well but with the cost of increased complexity in measurement and control. However, a distributed approach is not satisfactory at the domain level, lacking efficient cross-domain network measurement and management support. For a cross-domain signaling underlay, a decentralized architecture seems the sensible choice. • Modularity and isolation. To work in multiple scenarios, the system must be designed modularly. With such a layered, modular design (Figure 1), the system can adapt itself to different scenarios, with the least modification of kernel signaling function design, and be more flexible in resource consumption, self