Abstract: Network routing apparatus employs multi-level tree data structures in a centralized routing table and in distributed forwarding tables. Each level of each structure is associated with a different field of a network address appearing in received packets. Pointers in each structure are used to identify either an address of a next hop network, or a next-level tree to be examined for a next-hop address. An uncompressed tree routing table uses directly addressed trees in order to simplify the storage and retrieval of pointers, and the next-tree pointers directly identify next trees. Compressed tree forwarding tables are generated from the uncompressed routing table by reducing the number of pointers stored at one or more levels to substantially the number of unique next hop addresses associated with network addresses at that level. A single mapping table maps pointer values at one level to the locations of trees at the next level in the compressed trees.