Sensors networks instrument the physical space using motes that run\nnetwork embedded programs thus acquiring, processing, storing and\ntransmitting sensor data. The motes commercially available today are\nlarge, costly and trade performance for flexibility and ease of\nprogramming. A new generation of motes, designed as System-on-a-Chip,\nis promising to deliver a significant improvement in terms of power\nconsumption and price, possibly at the cost of reduced CPU efficiency.\nHow to compare mote performance? How to find out which mote is best\nsuited for a given application. In this paper, we propose a\nvector-based methodology for benchmarking mote performance. Our\nmethod is based on the hypothesis that mote performance can be\nexpressed as the scalar product of two vectors, one representing the\nmote characteristics, and the other representing the application\ncharacteristics. We implemented our approach in TinyOS 2.0 and we\npresent the result of experiments obtained on commercial motes from\nSensinode. We give a quantitative comparison of these motes, and\npredict the performance of a data acquisition application.