| 摘要 |
PROBLEM TO BE SOLVED: To provide a retrieval method of reducing the load of a merging process on a retrieval result of a retrieval server and enhancing expandability of decentralized retrieval.SOLUTION: A retrieval server calculates multiplicity "m" of retrieval from the number of retrieval servers in a retrieval server list. Based upon the calculated multiplicity "m", the retrieval server list is divided into "m" as uniformly as possible, and a retrieval query and a divided retrieval server list are handed over to each representative node. When retrieval servers to which the divided server lists are handed over have the number of retrieval servers in the handed list larger in number than the maximum number "M" of merge nodes, division and transfer of the retrieval server list are performed again. When the number of retrieval servers in the list is equal to or less than the maximum number of merge nodes, retrieval processing is performed for an index storage server, and a merging process is performed for an acquired retrieval result. When a retrieval server at a call source is present for the retrieval server after the merging process, the server at the call source performs a merging process. |
