NOTE the special time ----------------------------------------------------------------------- B O S T O N U N I V E R S I T Y Computer Science Department C O L L O Q U I U M Monday, September 9, 10:00 AM (Coffee served at 9:45 AM) Seminar Room / MCS 135 Routing and Path Coloring in All-Optical Networks Stathis Zachos Brooklyn College of CUNY and NTUA Abstract The rapid development of optical fiber technology eventually will lead to networks consisting solely of optical connections, the so called all-optical networks. Problems arising regarding routing and bandwidth allocation in such networks can be viewed as routing and path coloring problems in graphs, the majority of which are NP-hard. Two of the most widely studied relevant problems are the path coloring problem (PC) and the maximum path coloring problem (MaxPC). In PC a graph G and a set of paths P are given and the goal is to color all paths in P with the least possible colors, so that overlapping paths are not assigned the same color. The problem is NP-hard and cannot be approximated within a constant factor for general topologies. However, it is optimally solvable in chains, and as we show also in bounded degree trees and generalized trees of bounded degree. On the other hand, there exists a 1.5 approximation algorithm for PC in rings, as well as one that achieves a 1.36- approximation with high probability. In trees a 4/3 approximation algorithm exists, which is also the best possible unless P=NP. Studying PC in directed topologies is also interesting. In some cases the problem in a directed topology can be reduced to an undirected one. This is not possible for directed trees, and the best approximation factor found so far is 5/3. A variation of PC, which is also widely studied, is RPC, in which instead of a set of paths a set of connection requests is given. These requests must be routed via paths which are to be colored. In acyclic topologies PC and RPC coincide. In rings a 2-approximation algorithm exists, as well as one that achieves a 1.68 approximation with high probability. Furthermore, in trees of rings a combination of the algorithms for trees and rings can yield a 3-approximation. We also investigate RPC in directed topologies. In MaxPC, in addition to the graph G and the set of paths P, a number w of available colors is given. The goal is to color as many paths in P as possible, using at most w colors, under the restriction that overlapping paths are not assigned the same color. This problem is also NP-hard and non approximable within a constant factor for general graphs. In chains it is optimally solvable, while in rings and trees we obtain 1.5 and 1.58 approximation factors respectively. In directed trees the problem is solved within a 2.22 approximation factor. As far as the routing variation of the problem (MaxRPC) is concerned, we present a 3/2 approximation algorithm for rings, while in directed rings the best factor we have achieved so far is 11/7. Better approximations may be found in the future for many of the topologies mentioned above. Furthermore, there are several variants of these problems related to all optical networks. Thus, research around this subject is active and expanding. Host: George Kollios ------------------------------------------------------------------------- For colloquium info, including directions, see http://cs-www.bu.edu/colloquium -------------------------------------------------------------------------