保证数据流融合中处理一致性的事务机制

在分布式实时传感器网络应用中,主要目标之一是保证协作交互的数据流融合任务花费较少的能源代价,能够正确、实时、原子地外化监测区域信息.因此,本文重点研究了分布式传感器网络环境下对多数据流交互处理过程的事务控制机制.首先,定义了传感器网络数据融合过程的协作执行模型.其次,用运控制区域的概念,提出了一种基于系统时间戳的自适应的数据融合交互处理的事务模型及其ACID规则.最后,设计了一种适宜实时环境的数据融合事务的两阶段提交、回滚协议.测试结果分析表明,它能够显著提高采集数据进行事务处理的外化速率,降低了事务的失败率及网络能耗,特别适用于处理间歇、高速的传感器网络外部采集数据.

A Transaction Mechanism for Providing Efficient Data Services in Streaming Data Fusion

With the continuing advances in sensor networks and application design,appropriate middleware is needed to provide efficient QoS for coordinate concurrent applications on the sensor networks.Providing efficient data service is one of the fundamental goals for sensor networks.The data service paradigm requires that coordinate concurrent streaming data fusion applications can report information to observers as accurately and quickly as possible while making use of lower sensor energy cost.This paper focuses mainly on the transaction control techniques for multi interconnection streaming data fusions on distributed real-time sensor networks.In the paper,we first propose an executing model upon data fusion procedures on sensor networks.By using the concept of sphere of control(SoC),an adaptive transaction model based on timestamp is then presented so as to providing efficient data services for streaming data fusions while guaranteeing time synchronization and transaction control.Finally,we educe the ACID regulations upon the model,and design a two-phrase submission and back-rolling protocols for the real-time applications.Our performance evaluations show that by applying the two-phrase submission protocol,the outer transformation ratio of collection data can be significantly enhanced and the amounts of failed real-time transactions and energy cost used to fuse streaming data can be reduced correspondingly.The technique is suitable to process intermittent and high speed acquisition data in sensor networks.