Bratislav Milic

NPART - node placement algorithm for realistic topologies in wireless multihop network simulation

Despite a considerable number of topology generation algorithms for simulation of wireless multihop networks it is difficult to find one with output similar to real networks. We propose NPART --- a Node Placement Algorithm for Realistic Topologies whose output topologies resemble networks encountered in reality. The algorithm is flexible since it is sufficient to provide it with different input data to obtain different topologies. To demonstrate its quality, we compare topologies generated by NPART algorithm with our measurements from open wireless multihop networks in Berlin and Leipzig. Compared with real topologies, the generated topologies have almost identical node degree distribution, similar number of cut-edges and vertices, and distribution of component sizes after bridge removal. The importance of node placement algorithm is demonstrated by comparing ns-2 simulation results for grid and uniform node placement with NPART generated topologies. Simulation results show that node placement model plays as important role in simulation outcome as the accuracy of wireless signal propagation model.

Analyzing Large Scale Real-World Wireless Multihop Network

We present the results of analysis of Berlins free multihop wireless network and provide topological, link reliability and traffic statistics we have captured from it. The network is relatively large compared with known wireless multihop networks and testbeds, and it had 316 participating nodes on the average. The observed properties are different from frequently used models: the network is connected but with low average node density, it has large number of bridges (some of them with very low quality of links) and the traffic distribution is highly asymmetrical.

Analytical availability assessment of IT services

The often neglected problem in the service availability analysis is mapping between ICT-infrastructure and service-level availability. We present an approach which allows to map ICT-infrastructure elements to services, and to analytically assess steady-state, interval and user-perceived service availability, based on failure distributions of ICT-elements that implement a composite service. In case that full topology or all failure distributions of ICT-infrastructure elements are unknown, we provide means to estimate upper and lower availability bounds.

Automatic Generation of Service Availability Models

In the world where on-demand and trustworthy service delivery is one of the main preconditions for successful business, service and business process availability is of the paramount importance and cannot be compromised. For that reason service availability is coming into central focus of the IT operations and management research and practice. Still, our understanding of service and process availability is mostly empirical and at best, sketchy. Services are assessed using a mixture of qualitative, quantitative, and analytical methods, with results of varying quality. We introduce a systematic model-based methodology and a tool for service and business process availability assessment. The main advantage of the proposed method is the ability to automatically generate availability models, based on the service/process description and technical infrastructure it is executing on. Therefore, service level agreements can be tested/simulated or return on investment calculation can be performed, without the need for costly experiments and/or actual investments.

Prediction of Partitioning in Location-Aware Mobile Ad Hoc Networks

We propose an algorithm for detection of partitioning in location-aware mobile ad hoc networks. The partitioning occurs when movement pattern of nodes is such that they separate into groups that cannot communicate with each other. We use information about node position and speed in order to build a model that is able to predict when partitioning will occur and which link is critical using properties of planar graphs that represent the network. Our algorithm is distributed and uses only local topology knowledge where nodes keep track of position and speed of their one-hop neighbors.

Modeling Business Process Availability

In the world where on-demand and trustworthy service delivery is one of the main preconditions for successful business, availability of the services and business processes is of the paramount importance and cannot be compromised. We present a framework for modeling business process availability that takes into account services, the underlying ICT-infrastructure and people. Based on a fault model, we develop the methodology to map dependencies between ICT-components, services and business processes. The mapping enables us to model and analytically assess steady-state, interval and user perceived availability at all levels, up to the level of the business process.

Adaptation of the breadth first search algorithm for cut-edge detection in wireless multihop networks

We extend the Breadth First Search (BFS) algorithm to use it for cut-edge(bridge) detection in graphs. The changes in the algorithm are tailored such that the algorithm can be applied in wireless multihop networks: e.g., it fully utilizes the broadcasting nature of the wireless medium. The distributed BFS algorithm (flooding) is widely used for route discovery and information dissemination in wireless multihop networks (WMNs) so the overhead introduced by our bridge detection algorithm is limited - the network is already performing the distributed BFS and we reuse the information from it to detect the bridges. We verify our detection algorithm on the data sampled from Berlins free multi-hop wireless network. Detection precision varies depending on the algorithm parameters but for the representative algorithm configurations it stabilizes around 75%. Analysis of the data samples indicated that due to unreliability of wireless links and frequent occurrence of bridges the route discovery mechanism cannot find the route between two nodes although a valid route exists. We use our bridge detection algorithm to improve the route discovery success ratio from about 47% to approximately 90% by utilizing unicast of route discovery messages over the bridges. We verified by using fault injection the robustness of our approach as precision and route discovery remained high even for frequent node failures in the network.

Dropped Edges and Faces' Size in Gabriel and Relative Neighborhood Graphs

We have developed a mathematical model for calculation of expected value of share of edges being dropped while constructing a Gabriel or a relative neighborhood graph starting from a random geometrical graph created by a homogeneous Poisson point process on a plane. The expressions for the expected values of number of faces and face size are derived as well. All expressions are given as functions of deployment configuration. The results are verified through detailed comparison with extensive simulation results. Since Gabriel and relative neighborhood graphs are used for various applications in ad hoc networking we believe that our model simplifies theoretical analysis of functionality and performance of such networks

Properties of Wireless Multihop Networks in Theory and Practice

Simulation and testbeds are frequently used for the validation of wireless networking protocols, but several assumptions regarding node placement, wireless signal propagation, and traffic type must be made. We compare common models with the measurements made in Berlin’s and Leipzig’s free multihop wireless networks. It is shown that the properties observed in reality are different than in commonly used models: network is connected but with low average node density; it has large number of bridges and articulation points that can compromise its connectivity; and the traffic distribution over nodes is highly asymmetrical. As an illustration of the discrepancy between reality and synthetic models, we present issues of reactive route discovery process that cannot be observed in simulation that use common placement and propagation models. This chapter focuses on the understanding of limitations of simulation methodologies. It also provides general guidelines on ways of reducing the gap between simulation theory and practice of wireless multihop networks.

Distributed biconnectivity testing in Wireless multi-hop networks

Wireless multi-hop network (WMN) is a distributed communication system composed of autonomous processing nodes that is known for its ability to automatically adjust to rapidly changing conditions in the surrounding environment. Connectivity is one of the basic properties of a network. Removal of a bridge or an articulation point partitions a network. Biconnectivity testing identifies bridges and articulation points in a network, and once they are known corrective actions can be performed in order to improve network’s reliability. Numerous biconnectivity testing algorithms are successfully applied in graphs, wired networks and multiprocessor systems. However, they are inadequate for application in wireless networks since the frequent packet losses introduce uncertainty in the system which these algorithms cannot handle. The stochastic analysis shows that errors in decision-making in WMNs are considerable even for seemingly simple tasks such as the detection of links. The main contribution of this work is to provide means for accurate binary decision-making under uncertainty within the context of biconnectivity testing in WMNs. A distributed algorithm is developed that successfully handles the faults caused by message losses and simultaneously utilizes benefits of wireless communication to reduce message complexity from O(e) to O(n). Based on stochastic analysis of WMN topologies and a comprehensive analysis of impact of communication faults on algorithm’s behavior, the algorithm is extended by voting theory to reduce probability of erroneous decisions. The WMNs in Berlin and Leipzig are used as the case study. Topological data collected in them demonstrates that real networks are composed of dense and sparse parts, and confirms existence of numerous bridge and articulation points. The known node-placement algorithms are unable to recreate these properties so a new node placement algorithm for realistic topologies in WMN simulation was developed. The algorithm and the voting rules are evaluated in experiments in Motelab testbed and in the event-based simulator Jist/SWANS. The algorithm is accurate under various conditions which demonstrates its applicability in reality and capability of successful operation in presence of packet losses.