Social Networks, The Recovery Of The Performance Impact Of Large-scale Distributed Systems Research

With the development of information technology, large scale distributed system design has been a booming research field. Related researches have been successfully applied to the domains such as military, aeronautics, commercial services and information retrieval. In those applications, distributed systems are always organized as a social network. Moreover, when the network is organized as a social network topology, it is a key factor to influence the system performance and has been intensively studied in the disciplines of physiology, economics, physics, and computer science. Preliminary studies have found network topologies have also been shown as a key factor to influence system performance even under the same coordination algorithm. Although many distributed algorithm designs have been proved to be feasible to make up some functions in the large scale distributed systems as claimed, for example, recovering the network from link or node failures, the efficiency of recovery and the change of the network topologies have never been tested. Therefore, their influences to the overall system performance are unknown. In this paper, we have made an initial effort to find how network topologies influence the system performances after a standard network recovery policy is performed on large scale distributed systems. Those interesting discoveries are helpful to predict their influences on system performance and in turn to be useful for new algorithm design.The first part investigated the network models of large scale distributed systems. This part firstly analyzed the complex network topologies—random networks, grid networks. Then, the small-world networks and scale-free networks which present the characteristics of social networks were further studied in terms of realistic networks. Moreover, the algorithms building the models above were not only described, we also proposed an algorithm, named RX, which can fast locate nodes to simulate a scale-free networks.The second part researched about the influence of the characteristics of different social networks above on network recovery. This part firstly described a standard network recovery policy—an unbiased recovery algorithm based on MPLS. Then, we mainly compare the efficiency of network recovery on different topologies based on the parameters - hops, backup links and transmission paths. By comparing the influence on different networks, while there are some differences in different cases, our major discoveries are: the efficiency of network recovery in a small-world network is superior to the performance in other networks.The third part further studied the influence of social networks above on systems performances. We firstly judged the trend of changes about the system robustness of different network topologies based on the MPLS recovery algorithm and network congestion. By comparing the influence on different networks, while there are some differences in different cases, our major discoveries are: a random network is more likely to be congested if more links are failed, while a small-world network is the least likely. On the other hand, a scale-free network appears to be more vulnerable to node congestions, while a grid network is the most robust. Moreover, we tested how the network topology may be changed from network recover operations based on the characteristics of social networks. We have found that most network topologies are immune to the network recovery when the congestions are not so bad. However, when network congestions are very likely to destroy network connectivity, network topologies may be significantly changed, especially when the network is organized as a scale-free network or a small-world network, which are two most important social network topologies of a distributed system. From our previously experience, those changes in network topologies must result a significant effect on system's overall performances.