Techniques for the Optimization of Communication Flows in Distributed Systems

Mugurel Ionut Andreica (1)

1 Parallel and Distributed Systems Laboratory [Bucarest]

Many types of distributed systems have been developed and deployed throughout the world recently, in order to solve various problems or in order to provide novel types of services. Communication is an essential property of every distributed system, because the components of the system must communicate with each other in order to meet the system's goals (e.g. solve a problem, or provide a service). As the attributes of distributed systems may vary considerably, so can their communication requirements. Some types of requirements can be easily met, but most of them pose challenges which have been insufficiently addressed so far. This book focuses on the broad topic of communication flow optimization in distributed systems and proposes two types of solutions: algorithmic (online and offline methods and techniques) and architectural (centralized, decentralized and hybrid system architectures). The book considers both point-to-point and multicast data transfers, as well as a wide range of communication parameters (e.g. bandwidth, latency, cost, reliability). The addressed problems are modeled based on the communication requirements of a subset of the most important types of distributed systems. These requirements and the main communication parameters are identified and categorized in Chapter 2, where a thorough analysis of the current state-of-the-art in the field is also presented. Chapter 3 presents several new decentralized peer-to-peer architectures for optimizing point-to-point communication flows, as well as flows generated by content search and retrieval. Several new types of techniques applicable in these architectures are also presented. Chapter 4 proposes a centralized real-time point-to-point data transfer scheduling architecture, together with many algorithmic techniques, for solving the problem of providing end-to-end Quality-of-Service (QoS) guarantees to the communication flows occurring in a distributed system. Chapter 5 presents new algorithms for several offline point-to-point communication flow scheduling problems. In Chapter 6 new solutions for optimizing multicast data transfers are presented, like a peer-to-peer multicast tree architecture with bounded node degrees and small diameter, as well as offline algorithms for new models of multicast and broadcast in tree networks. Chapter 7 considers several replica placement problems in tree-like networks and focuses on solving the associated communication optimization problems. New algorithms for optimally placing replicas such that the communication cost (e.g. access time) associated to accessing the replicas is minimized are presented. Chapter 8 concludes the book and presents an overview of the original contributions of this book. The contents of this book are based upon the author's Ph.D. thesis, which was written under the guidance of Prof. Dr. Eng. Nicolae Tapus, while the author was a Ph.D. student at the Politehnica University of Bucharest, in the Computer Science & Engineering Department. In fact, the book is a slightly extended version of the thesis.

Mots clés: communication flow optimization; peer-to-peer architectures; online and offline data transfer scheduling; multicast strategies; optimal replica placement

Domaines: Calcul parallèle, distribué et partagé [cs.DC]; Algorithme et structure de données [cs.DS]

Mugurel Ionut Andreica. Techniques for the Optimization of Communication Flows in Distributed Systems. CIBERNETICA MC Publishing House, pp.198, 2010, 978-973-88451-4-5. ⟨hal-00449878v2⟩