Building and Searching New Packet Classification and Organization Rule

Abstract

Several efforts were made in the existing solutions to identify a successful packet classification solution. A variety of decision-tree-based packet classification algorithms were analyzed in our works. Earlier decision tree algorithms chooses field as well as number of cuts based on a nearby optimized decision, which compromises search speed as well as memory requirement. Algorithms of decision tree make possible maximum priority match and multi-match packet classification. New applications of networks have demanded a multi-match packet classification in which the entire matching results along with highest-priority matching rule have to be returned and it is essential to discover competent algorithms to resolve classification problems. A new resourceful packet classification algorithm was introduced in our work named as boundary cutting. Mainly conventional applications of packet classification necessitate the maximum priority matching. The algorithm that was introduced has two principal advantages. Boundary cutting of projected algorithm is more effectual than that of earlier algorithms as it is based on rule boundaries to a certain extent than permanent intervals thus, amount of necessary memory is considerably reduced. Although boundary cutting loses indexing capability at internal nodes binary search recommend better-quality search performance.

Keywords: Decision tree algorithms; Packet classification; Boundary cutting; Priority matching; Binary search.