In this paper we propose the double auction allocation models for computational grids, and two double auction algorithms for resource allocation: Single Price Double Auction algorithm (SPDA), and Multiple Attribute Double Auction algorithm (MADA). The fairness and comprehensive performance are introduced into the traditional double auction models to evaluate impact of price, speed and storage of grid resource. The results show that MADA algorithm is better from both resource's and user's perspective providing high resource utilization. The results also highlight how a resource user can improve his performance using multi-attribute characteristics.