An application-level network performance measure method for distributed information retrieval

To balance the load for the crawlers in distributed information retrieval system, a new network performance measure method is put forward. Adopting the application-level interaction procedure and theory of the distributed crawling, pipeline number and page size are incorporated into the network performance metric. Experimental results prove that the new method has a high stability under the statistics, and can be a standard of the network distance to build the network coordinates.