Generic Single Edge Fault Tolerant Exact Distance Oracle

Abstract: Given an undirected unweighted graph $G$ and a source set $S$ of $|S| = \sigma $ sources, we want to build a data structure which can process the following query {\sc Q}$(s,t,e):$ find the shortest distance from $s$ to $t$ avoiding an edge $e$, where $s \in S$ and $t \in V$. When $\sigma=n$, Demetrescu, Thorup, Chowdhury and Ramachandran (SIAM Journal of Computing, 2008) designed an algorithm with $\tilde O(n^2)$ space ($\tilde O(\cdot)$ hides poly $\log n$ factor.) and $O(1)$ query time. A natural open question is to generalize this result to any number of sources. Recently, Bil{`o} et. al. (STACS 2018) designed a data-structure of size $\tilde O(\sigma^{1/2}n^{3/2})$ with the query time of $O(\sqrt{n\sigma})$ for the above problem. We improve their result by designing a data-structure of size $\tilde O(\sigma^{1/2} n^{3/2})$ that can answer queries in $\tilde O(1)$ time. In a related problem of finding fault tolerant subgraph, Parter and Peleg (ESA 2013) showed that if detours of the {\em replacement} paths ending at a vertex $t$ are disjoint, then the number of such paths is $O(\sqrt{n\sigma})$. This eventually gives a bound of $O( n \sqrt{n \sigma}) = O(\sigma^{1/2}n^{3/2})$ for their problem. {\em Disjointness of detours} is a very crucial property used in the above result. We show a similar result for a subset of replacement path which \textbf{may not} be disjoint. This result is the crux of our paper and may be of independent interest.?