I wouldn't give up yet. You need to see an implementation and how it benchmarks.

Implementing Suffix Trees in pure perl is messy, memory expensive (hugely so in my attempts), and rarely works out more efficient. Don't forget that it's O(n) to build as well as O(m) to use. If you are only using it once, that snips into the benefits.

For short strings a linear search (at C speed) and no build time wins easily. For longer strings, the indexing a suffix tree provides is a winner, but it gets set back awfully by the build time and memory management. If you only use the tree the once (as for this challenge), there little or no win.