A Linear-Order Based Access Method for Efficient Network Computations

摘要

It is well known that the disk-I/O costs of network computations, for example, finding a shortest path between two vertices, can be reduced by clustering the connected vertices in the network to the same disk page. This paper studies a network access method which is based from a linear order for the vertices in a given network. In order to compute network queries efficiently, a linear order should preserve a "good" clustering property of a network. For this requirement, we present a hierarchical vertex ordering method which obtains a linear order of the vertices in a simple but effective way. In experiments, various access methods are evaluated by comparing the I/O costs of the network computations. The access method constructed from the proposed ordering method outperforms other methods in most cases.