
Network Layer Support for Service Discovery in MANETs Ulas Kozat and Leandros Tassiulas University of Maryland, College Park Presented by Wei Gao
Background • Service discovery • Users search for their desired network services • Examples of network services: remote printers, scanners, data sources, and webpages……
Background • Components of a service discovery system: • Service requesters • Service providers • Service repositories • A service discovery process include: • Dissemination of service discovery messages • Matchmaking between the services requested and provided
Background • Service discovery in MANETs • MANETs are highly dynamic • Prevalent node mobility • Adverse channel conditions • Unpredictable node failures • Requirements for service discovery • Scalable to search for matching services quickly and efficiently • Robust against unpredictable network topology changes
Overall approach • Efficient dissemination of service discovery messages among service providers and repositories • Backbone management phase: Formulation of 1-hop clusters and a virtual backbone • Distributed service discovery phase: Distribution of service discovery messages
Backbone Management (BBM) Phase • Some notations regarding the virtual backbone • Service Broker Node (SBN): the clusterheads connecting other non-backbone nodes • Virtual Access Point (VAP): the non-backbone node acting as the access point for another node to the backbone • Directory Agent (DA): software agent that store service registrations
Backbone Management (BBM) Phase • Construct a virtual backbone for dissemination of service discovery messages • Essentially, a clustering process • Clusterheads consists of a dominating set • Criterion of clusterhead selection • Normalized link failure frequency (NLFF) • Properties: • Convergence of cluster formation: a dominating set can always be constructed (Lemma 1) • Cluster completeness: every node has an associated VAP (Lemma 2)
Backbone Management (BBM) Phase • An example
Distributed Service Discovery (DSD) Phase • Service registration • A node registers its service to the DA located on its VAP • For a backbone node, just register at itself • Multicast dissemination to other DAs located on other VAPs • A multicast tree is rooted at every DA consisting all the other VAPs in the network • Construction of multicast tree: broadcast + self-pruning
Distributed Service Discovery (DSD) Phase • Example of multicast tree construction
Distributed Service Discovery (DSD) Phase • An example of resource discovery
Simulation results • Successful ratio • Advantage is shown compared to plain network architecture • Slightly affected by node mobility
Simulation results • Service discovery delay • Improved over anycast-DSR • Slower response than DSDP and AODV
Simulation results • Constructions of backbone and multicast trees cost extra communication overhead
Summary • Pros: • Propose an approach to constructing a clustered network architecture for service discovery • The clustered network architecture is able to achieve high successful ratio of service discovery • Distributed service discovery provides sound robustness against link failures
Summary • Cons: • The proposed approach leads to extra communication overhead, and thus may not be suitable for large-scale networks • This is mainly because the 1-hop clustered architecture. There will be too many backbone nodes and VAPs in the network
Relation to course project • Multi-hop extension based on the approach in this paper • Service discovery on multi-hop clusters • Clusterhead selection considering both efficiency and stability • Solving possible inconsistency • Distributed storage of service registrations • Deployment of local DHTs on multi-hop clusters • Local searching based on Hash results