Kayhan Erciyes

A dominating set based clustering algorithm for mobile ad hoc networks

We propose a new Connected Dominating Set (CDS) based algorithm for clustering in Mobile Ad hoc Networks (MANETs). Our algorithm is based on Wu and Li’s [14] algorithm, however we provide significant modifications by considering the degrees of the nodes during marking process and also provide further heuristics to determine the color of a node in the initial phase. We describe, analyze and measure performance of this new algorithm by simulation and show that it performs better than Wu and Li’s [14] algorithm especially in the case of dense networks.

A distributed backbone formation algorithm for mobile ad hoc networks

Construction of a backbone architecture is an important issue in mobile ad hoc networks(MANET)s to ease routing and resource management. We propose a new fully distributed algorithm for backbone formation in MANETs that constructs a directed ring architecture. We show the operation of the algorithm, analyze its message complexity and provide results in the simulation environment of ns2. Our results conform that the algorithm is scalable in terms of its running time and round-trip delay against mobility, surface area, number of nodes and number of clusterheads.

A cluster-based dynamic load balancing middleware protocol for grids

We describe a hierarchical dynamic load balancing protocol for Grids. The Grid consists of clusters and each cluster is represented by a coordinator. Each coordinator first attempts to balance the load in its cluster and if this fails, communicates with the other coordinators to perform transfer or reception of load. This process is repetaed periodically. We show the implementation and analyze the performance and scalability of the proposed protocol.

A merging clustering algorithm for mobile ad hoc networks

Clustering is a widely used approach to ease implementation of various problems such as routing and resource management in mobile ad hoc networks (MANET)s. We propose a new fully distributed algorithm for clustering in MANETs that merges clusters to form higher level clusters by increasing their levels. We show the operation of the algorithm and analyze its time and message complexities and provide results in the simulation environment of ns2. Our results conform that the algorithm proposed is scalable and has a lower time and message complexities than the other algorithms.

Merging clustering algorithms in mobile ad hoc networks

Clustering is a widely used approach to ease implementation of various problems such as routing and resource management in mobile ad hoc networks (MANET)s. We first look at minimum spanning tree(MST) based algorithms and then propose a new algorithm for clustering in MANETs. The algorithm we propose merges clusters to form higher level clusters by increasing their levels. We show the operation of the algorithm and analyze its time and message complexities.

A replication-based fault tolerance protocol using group communication for the grid

We describe a replication-based protocol that uses group communication for fault tolerance in the Computational Grid. The Grid is partitioned into a number of clusters and each cluster has a designated coordinator that manages the states of the replicas within its cluster. The coordinators belong to a process group and the proposed protocol ensures the correct sequence of message deliveries to the replicas by the coordinators. Any failing node of the Grid is replaced by an active replica to provide correct continuation of the operation of the application. We show the theoretical framework along with illustrations of the replication protocol and its implementation results and analyze its performance and scalability.

Performance evaluation of group communication architectures in large scale systems using MPI

Group communication is an important paradigm for fault tolerance in large scale systems We describe various group architectures as pipelined, hierarchical, daisy and hypercube groups each consisting of separate clusters, investigate the theoretical performance bounds of these architectures and evaluate their experimental performances using MPI group communication primitives We first derive time bounds for multicast message deliveries in these architectures and then provide tests to measure the times taken for the same operation The multicast message delivery times are tested against the number of clusters within a group and the size of the multicast message We conclude that daisy architecture is favorable both in terms of delivery times and message sizes theoretically and experimentally.

Implementation of a cluster based routing protocol for mobile networks

We show the implementation and the simulation results of a hierarchical, cluster based routing protocol for mobile ad hoc networks using Parallel Virtual Machine (PVM). The network represented by a graph is partitioned into clusters by a graph partitioning algorithm and the shortest routes are first calculated locally in each cluster in the first srep. The simplified network which consists only of the nodes that have connections to other clusters called the neighbor nodes is then formed and the shortest routes are calculated for this simple network as the second step. A complete route between the two nodes of different clusters is formed by the union of intra-cluster and inter-cluster routes. We show the implementation results using PVM where a workstation represents a cluster and each node is a PVM process. The results obtained support the theoretical considerations where the efficiency increases by the number of clusters in use.

A Software Architecture for Shared Resource Management in Mobile Ad Hoc Networks

We propose a three layer software architecture for shared resource management in mobile ad hoc networks(MANETs). At the lowest layer, the Merging Clustering Algorithm(MCA)[11] partitions the MANET into a number of balanced clusters periodically. At the second layer, the Backbone Formation Algorithm(BFA) provides a virtual ring using the clusterheads found by MCA. Finally, an example resource management protocol which is a modified and scaled version of the Ricart-Agrawala algorithm implemented using the virtual ring structure is presented with the performance results.

Multilevel static real-time scheduling algorithms using graph partitioning

We propose static task allocation algorithms for the periodic tasks of a distributed real-time system. The cyclic task consists of task threads which may communicate and share resources. A graph partitioning process and a thread sequencing algorithm are applied to these threads to yield local schedules. The exact analysis is then obtained and further refinements are performed if the worst case response time of a task is greater than its deadline.