ORDER: Optimal Routing with Path Indexing in Exchange Graph [Data-Intensive & Data Science Application]

We study the problem of optimal routing on financial exchange graphs. Given a directed graph G = ( V , E ) where vertices represent assets and edges encode tradable pairs with effective exchange rates and capacities, the goal is to find a sequence of routes from a source asset to a target asset that maximizes the delivered output. This problem is data-intensive and time-critical, requiring real-time decisions under fragmented liquidity, size-dependent prices, heterogeneous fees, and tight capacity limits. To address the challenges, we present ORDER: O ptimal R outing with path in D exing in E xchange g R aphs, a high-efficiency routing solution for fast, iterative execution in financial exchange graphs. ORDER introduces a hierarchical bucket path index that localizes maintenance to impacted candidates and eliminates expensive global rescans. It further incorporates lazy computation and an adaptive controller for active-bucket sizing to stabilize update cost under discrete tier shifts and heterogeneous market depths. On real-world DeFi datasets, ORDER achieves up to 114× speedup over state-of-the-art routers while preserving execution quality.We also demonstrate robustness across market regimes and token pairs, enabling timely, high-quality routing.