Load Management Policies Of The Distributed Stream Processing System ARTs-SH

The data stream system is always overloaded at the running time because the data stream is unpredictable,brusty,high ,continuous , and the available system resources are limited .It leads to the system latency deteriorated and the system can not provide the ti- mely response.Under these conditons the given answers will have no value.So it is nece- ssary to reasearch on the load management policies including the dynamicly moving load policy and the load shedding polify in the distributed data stream system.An improved police named bidding load moving policy fits our developed ARTs-SH system was proposed based on analysing the present moving load policies at the run-time of the distributed data stream management system.It can implement moving the load across the nodes ,reducing the latency of the system .It fits for widely range. Different queries have different importance in our system,however, previous load shedding policies of the data stream system are almost assume that all queries are equally important . So they do not fits for our system.A new load shedding policy based-on unequal priority on different queryis is proposed .Some main factors of the queries are considered and selected to determine the priorities of the query .Based on them the priorities of the input streams is determined.Then different sample coefficient was setted on different streams, so more important queries can make more convincing outputs,.The load shedding policy can provide a good QoS.The load management policy of the data stream management system was researched. The bidding load moving policy and the load shedding policy based-on unequal priorities'queries were proposed .The benefits of them were confirmed by theory analysis and experiments .