The Research On Resource Aggregation In National Key Discipline Grid

By now, there is hardly an effective way for aggregation of academic resources which are distributed on the Internet. Search engines can find tons of resources with their interconnections and qualities in the dark by Web crawlers; traditional information release portals can aggregate slightly part of resources on the Internet by human classification, but they suffer by a low efficiency and high cost.To address the problem presented above, NKDG_RAS (National Key Discipline Grid Resource Aggregation System) supplies a platform for resource aggregation based on Grid Technology and Web 2.0, and meets its end through social contribution and collaborative filtering made by all users. The architecture of NKDG_RAS has 3 layers: Information layer collects resources distributed on the Internet, it includes a RSS crawler to retrieve structural resources, a Web-based literature meta data extraction service to parse semi-structural resources and a user contribution strategy to gather un-structural Web pages; Aggregation layer is constructed above Information layer, it includes a tag service to reveal the relationships between resources by tag interconnection measurement algorithm, an individualized recommendation service to find out resources which have high qualities and probably meet user's interests in recent updates, an import service to transfer resources from other applications or platforms to our system, a duplicate detection service to help users to discover if anybody else has the same resource with themselves; Application layer is the top layer, it includes a internationalization mechanism to support multi-languages, a RSS syndication service to supply resource mashup for other applications.NKDG_RAS is implemented by using PHP and Java language. Analysis is made after the tests. The result of function tests indicates that NKDG_RAS is full functional and works normally. The result of performance tests indicates that the number of concurrent users supported will increase 50% and the response time will reduce 51.08% after separating RSS crawler from center server. Also, when caching policy is added into Web-based literature meta data extraction service and hit ratio equals 0.5, the response time will reduce 42.63%.