Abstract. In network communication systems, frequently messages are routed along a minimum diameter spanning tree (MDST) of the network, to minimize the maximum travel time of messages. When a transient failure disables an edge of the MDST, the network is disconnected, and a temporary replacement edge must be chosen, which should ideally minimize the diameter of the new spanning tree. Preparing for the failure of any edge of the MDST, the all-best-swaps (ABS) problem asks for finding the best swap for every edge of the MDST. Given a 2-edgeconnected weighted graph G = (V, E), where |V | = n and |E| = m, we solve the ABS problem in O (m log n) time and O(m) space, thus considerably improving upon the decade-old previously best solution, which requires O(n √ m) time and O(m) space, for m = o n 2 / log 2 n .
IntroductionFor communication in computer networks, often only a subset of the available connections is used to communicate at any given time. If all nodes are connected using the smallest possible number of links, the subset forms a spanning tree of the network. When an edge in a communication tree fails, routing information becomes wrong and message transmission is interrupted. For transient failures that are expected to be repaired quickly, the idea of online point-of-failure rerouting has gained popularity recently [2,4, 6,8]: Instead of changing a lot of routing information, only one alternative (so-called swap) edge is used to reconnect the disconnected parts of the tree. For the corresponding change in routing information to be fast, a swap edge for each failing edge needs to be readily available, as the result of an earlier computation. Among all possible swap links for a failing edge, one should choose a best swap link, that is, a swap edge which reconnects the two disconnected parts of the tree in such a way that the resulting swap tree is best w.r.t. some objective. We show in the following that the common computation of all best swaps (ABS) has the further advantage of gaining efficiency (against computing swap edges individually), because dependencies between the computations for different failing edges can be exploited.In this paper, we are interested in using a Minimum Diameter Spanning Tree (MDST) as the communication tree, i.e., a tree that minimizes the largest distance between any pair of nodes, thus minimizing the worst case length of any transmission path, even if edge lengths are not uniform. Consequentially, a best swap edge in our case minimizes the diameter of the resulting swap tree. Interestingly, this choice of swapping against adjusting the entire tree even comes at a moderate loss in diameter: The swap tree diameter is at most a factor of 5/2 larger than the diameter of an entirely adjusted tree [9].Related Work. During the last decade, the ABS problem has been investigated for spanning trees with various objectives [1,8,9,11,13,15].Computing all best swaps of a MDST was one of the first swap problems that were studied. In [9], an algorithm for this problem is given which requires O...