Structure Construction And Performance Improvement In Distributed Network Information Systems

Nowadays,more and more network information systems are constructed to meet higher demands for information society. The network information system in this dissertation has three main parts: information collection,transmission,and process. In such a system,information is firstly collected by the sensor nodes and then transmitted to the data center. The information is finally processed by the data center. Therefore,it is necessary to improve the network throughput and transmission quality.Most network information systems employ the sensor nodes to collect and transmit the information. Due to the limitation of wireless bandwidth,the network aggregate throughput should be improved to transmit more data. In wireless networks,many transmit parameters (such as carrier sensing threshold,transmission rate etc.) can impact the transmission quality of one link. This dissertation firstly analyzes the roles those parameters play to the single flow throughput and aggregate network throughput. Then an analytical model is proposed to calculate the optimal network throughput. Based on this model,a distributed parameter-tuning algorithm is driven. Each node can adapte its transmission parameters according to the algorithm. With our proposed algorithm,the aggregate network throughput is improved.The information collected and transmitted in the network information system has developed from text messages to multimedia files. In this dissertation,we proposed two admission control schemes to provide QoS for real-time multimedia traffic. One is a node-based admission control scheme based on the two-hop interference model. The other is a cross-layer admission control scheme in multi-rate wireless networks. Both schemes can accurately model the contention and interference one link suffers. The simulation results show that,with the admission control scheme,the network resource can be better utilized,and the quality of transmissions can be guaranteed.The data center is used to process and compute the information collected from the sensors. Since more and more information is processed,the tasks are allocated to multiple servers to improve the process speed. The cooperation among the servers needs reliable communication path with high throughput. In this dissertation,therefore,we adopt the random graph theory and proposed a new server connection strategy. The new structure has good scalability,moreover,it can provide high throughput and reliability. Routing and fault tolerance mechanisms are also provided for the new structure. The routing algorithm can find routing paths as short as possible with low overhead. The fault tolerance mechanism is able to reconnect the network when certain servers and links fail. The simulation results show that,compared with the traditional tree structure,our new structure can provide higher network throughput and better fault resilience. Even when compared with other structures,our structure can still provide better performance-price ratio.