| In traditional von Neumann architecture, computing resources and storage re-sources are statically bound together and form a local self-government, independence decision-making calculation entity. On the one hand, the entity'efficiency decreases at cooperation of computing task for their relatively independent state. On the other hand, decentralized management and maintenance of separate computing entities are a huge workload, and software fees are paid according to the number of entities. As the concept of transparence computing mode or pervasive computing is proposed, the computing tasks and resource are submited to the various distributed computing re-sources on Transparence Computing Mode, and data are storied focused on an on-demand deployment storage equipment or system, the system data saved in the storage resources is no longer belong to a particular computing entity which can be shared by all network resources. If a system volume shared by some computer, the total fee will be lessen and also be propitious to managemet. But the consistency will be a challenge.there are some study at present,but also exist som disfigurement .such as data disappear when power down and has a low performance in the process of sys-tem rebuilt. This paper is studied how to configure a system from the shared data and each entity private property and also some optimize aboat it.As each calculated entity has its own inherent attributes which is distinctive from other entities. The distributed system data with service ability should reconfigured from the shared data and each entity private property.so how to make it is the key. This paper is involved in below subjects:(1) An ant colony algorithm was proposed to analyse the rule of Multi-Host in the process of start from the shared system valums. At last, we come to some of its laws of it.(2) Based on the above conclusions,A system similar model was proposed and designed a volume reconstruction algorithm .this algorithm can apperceive fleetly and position rewritten blocks to maintain the consistency of data volumes.at the same time, the data structure use both arry and index to find data quickly and easily and also have a higher utilization of space.(3) Based on the mult-host reconstruction model, we designed a prefetching al-gorithm involve both the sequence and double jumping prefetching, while the algo- rithm which based on the sample shoot straight to decide whether prefetch which further improve the accuracy of the shooting. |