This paper focuses on the long-term preservation of the web resource. The authors try to explore web archives strategy from two viewpoints which are metadata format and the structure of contents. In the analysis of the metadata format, review several web archives programs in the world and analyze their rule of organizing information. Secondly, study the archival principle of provenance and its application in controlling levels of archives description. In addition, analyze multilevel structure of web resources to explore the possibility of developing a practical model for multilevel description of web resources. Finally, make use of Topic Maps to deconstruct web pages and build up the relational model of topics to achieve the goal of preservation and providing continues access to web resources.