AI BriefingKakaoMap Improves Bicycle Routing TPS by 70x with CCH Algorithm
Key point
Reduced the number of bicycle routing servers to 1/15 and stabilized nationwide search response times.
Details
KakaoMap has completely overhauled the search engine for its walking and bicycle routing services, migrating from the A algorithm* to Customizable Contraction Hierarchies (CCH). The existing A* algorithm had limitations where search time increased sharply with distance even when applying heuristics, causing response delays of tens of seconds in bicycle routing, which supports nationwide queries.
CCH is an algorithm that separates the graph's connectivity structure from costs, allowing it to modify only specific parts without recalculating the entire shortcut network when costs change. Through this, KakaoMap improved bicycle routing TPS by 70x and walking routing TPS by 3x. The actual number of servers was reduced to 1/15, and performance improved to a level where the 30km distance limit for walking routing could be lifted.
Technical Optimization
During the CCH implementation, KakaoMap utilized the Inertial Flow Cutter (IFC) to optimize graph partitioning and applied Line Graph transformation to handle turn costs. By removing unnecessary edges through IFC Ordering and reordering, the number of arcs in the CCH graph was reduced by approximately 46%, improving memory efficiency.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.