Piotr Pecka

Efficient Representation of Transition Matrix in the Markov Process Modeling of Computer Networks

Markov chains are used in analysis in many fields. One of them is performance evaluation of computer systems, especially computer networks. For the analysis we use OLYMP object library that provides features to describe complex systems and find their statistical parameters. We present two compressed data structures for the most space-consuming parts of data processed by OLYMP. Then, we show how these improvements move the barrier of applicability of the utility.

Content localization for non-overlay content-aware networks

Content addressing and localization are the basic issues in design of content-aware networks (CAN). The paper summarizes assumptions for content-aware networks and presents content localization algorithm for PI-CAN, developed within FIE (Future Internet Engineering) project. The algorithm allows to find location of desired content in distributed content database in time O(1).

Influence of Message-Oriented Middleware on Performance of Network Management System: A Modelling Study

Gathering data from Internet of Things and management of IoT devices requires an efficient communication architecture. In this paper we analyse architectures of the scalable, sensor-oriented IoT network management system, as well as the pros and cons of introducing into it a message-oriented middleware server (message broker). We compare two architectures: with distributed buffers and with a centralized message broker. The analysis was conducted on the basis of Markov chains and discrete event simulation.

Global Queue Pruning Method for Efficient Broadcast in Multihop Wireless Networks

The article proposes a novel broadcast algorithm for multihop wireless networks. We compare three reference algorithms: Counter Based, Scalable Broadcast and Dominant Pruning, and propose a novel Global Queue Pruning method, which limits the overhead of the transmission and provides assurance of the delivery of the messages to every node in the network. The developed algorithm creates the logical topology that consists of lower number of forwarders in comparison to the previous methods, the paths are shorter, and the 100 % coverage is guaranteed. This is achieved with the higher cost of propagation of the topology information in the initialisation phase.

Content Registry — The distributed database of content localization for the PI CAN network

Content addressing and localization are basic issues in the structure of content-centric networks. This paper presents the implementation and performance evaluation of the Content Registry (CR), developed within the framework of the FIE (Future Internet Engineering) project to meet the requirements of the PI CAN network. The CR allows to find the desired Content Record (containing basic transmission parameters and localization of a content) in the network-distributed content database. The results confirm the CR is highly efficient in various configurations of the logical topology of connected nodes.

Content Identification in PI CAN Network

Content addressing is one of the basic issues while design of CAN architecture and solution. The paper is devoted to the problem of choice of identifier format, length and characteristic. Basic assumptions of CAN as one of Parallel Internets in IIP (Inzynieria Internetu Przyszlości) project are presented and requirements for content identifier are shown in this context. Discussion of possible solutions is presented and proposal of final identifier format is provided.

Solving Large Markov Models Described with Standard Programming Language

Continuous time Markov chains (CTMC) are one of the formalisms for building models. This paper discusses OLYMP2 - a system for solving big CTMC models (exceeding \(10^9\) states), described with a standard programming language - Java. OLYMP2 is primarily aimed at modelling of computer networks, so its formalism comes from networking concepts, like queueing systems. Using Java as a model description allows for greater flexibility in comparison to model-checker specific languages that often do not employ complete features of an object-oriented programming. Using Java also makes the parsing of models relatively fast, due to optimised Java run-time environment. Introducing dedicated compression of transition matrices allows for keeping memory usage at reasonable level even for large models.

Application of Fault-Tolerant GQP Algorithm in Multihop AMI Networks

The paper presents the evaluation of the GQP algorithm in AMI networks using real map based network topologies (low, medium and high density of buildings). The algorithm was compared to the reference algorithms: CB, SBA and DP. It was shown that the GQP algorithm presents the good scalability, relatively optimal number of selected forwarders and by introducing the redundancy factor, good fault tolerance. An improvement to GQP algorithm to add a redundancy factor is proposed, what allows to execute fault tolerant broadcast transmission.

An Effective Way of Storing and Accessing Very Large Transition Matrices Using Multi-core CPU and GPU Architectures

The topic of probabilistic model checking algorithms of Markov processes makes the problem of storing and processing of very large transition matrices crucial. While the symbolic approach utilises very effective storage and access methods, like Multi–terminal Binary Decision Diagrams, the traditional explicit method of representing and solving the model using sparse matrices has many advantages, especially when we concired the flexibility of computing of different properties. We show and examine a compact representation of a very large transition matrix in the sparse form. The technique employs an effective method of compression, storage, and accessing of the matrix and is suitable for use in mutli-core CPU and GPU environments. Such large transition matrices consume a lot of space and cannot be considered as typical types of data for storage in database systems. Nevertheless, their storage and efficient access to them could be beneficial for solving of Markov models, so constructing specialised compression methods is obviously the way to go.

Simulation and Analytical Evaluation of a Slotted Ring Optical Packet Switching Network

Optical Packet Switching (OPS) is a technology which thanks to its speed of operation may become popular in next-generation networks. Because optical packets are not buffered it is necessary to develop new mechanisms which would allow quality of service management in OPS networks. The paper discusses the architecture of OPS networks of the ring type with nodes enabling optical packet switching, equipped with the mechanism of Quality of Service (QoS) management. The differentiation of the quality of service was achieved by means of the proposed mechanism of priority registers with the timeout mechanism. This enables the network to be used for transporting various classes of data with varying requirements as to the quality of service. Simulation trials have been conducted and an analytical model for this type of network has been proposed.