Matching Integer Intervals by Minimal Sets of Binary Words with $don’t cares

• Published in: Combinatorial Pattern Matching Vol. 5029; pp. 217 - 229
• Main Authors: Fraczak, Wojciech, Rytter, Wojciech, Yazdani, Mohammadreza
• Format: Book Chapter
• Language: English
• Published: Germany Springer Berlin / Heidelberg 2008; Springer Berlin Heidelberg
• Series: Lecture Notes in Computer Science
• ISBN: 9783540690665; 3540690662
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-540-69068-9_21

An interval [p,q], where 0 ≤ p ≤ q < 2n, can be considered as the set X of n-bit binary strings corresponding to encodings of all integers in [p,q]. A word w with don’t care symbols is matching the set L(w) of all words of the length |w| which can differ only on positions containing a don’t care. A set Y of words with don’t cares is matchingX iff X  = ∪ w ∈ YL(w). For a set X of codes of integers in [p,q] we ask for a minimal size set Y of words with don’t cares matching X. Such a problem appears in the context of network processing engines using Ternary Content Addressable Memory (TCAM) as a lookup table for IP packet header fields. The set Y is called a template in this paper, and it corresponds to a TCAM representation of an interval. It has been traditionally calculated by a heuristic called “prefix match”, which can produce a result of the size approximately twice larger than the minimal one. In this paper we present two fast (linear time in the size of the input and the output) algorithms for finding minimal solutions for two natural encodings of integers: the usual binary representation (lexicographic encoding) and the reflected Gray code.