NEAR-BOUNDARY DATA SELECTION FOR FAST SUPPORT