Building Distributed Attribute-Value Trees on DHT-based Networks

Today, when a huge number of information resources appear in the Internet, one problem arising is how to search and exploit these information resources efficiently and flexibly with high scalability. We believe that the realization of multiple-attribute searching in a distributed manner is an important solution for this problem. In this paper, we propose a solution of building a distributed attribute-value tree (AV tree) on DHT-based networks to realize multiple-attribute searching. Each resource is named by an AV tree and DHT keys are created from each AV branch of a resource name. Resource information is distributed to a DHT-based network by the use of these DHT keys. Our solution keeps the locality of resource names in the key space of DHT networks such that AV branches that share a number of attribute/value pairs (AV pairs) will be mapped to a continuous portion of the key space. Therefore, multiple-attribute searching can be achieved efficiently by querying nodes responsible for a continuous portion of the key space. Our system can achieve both efficiency and a good degree of load balancing even when the distribution of AV pairs in resource names is skewed. Our simulation result shows the efficiency of our solution in respects of query success ratio and the degree of load balancing, comparing with conventional approaches.