Dynamic packet classification is the basis of emerging network services, but the update performance of existing packet classification algorithms is unsatisfactory. Based on the Recursive Space Decomposition and Interpreter approach, this paper designs and implements a two-stage multi-dimensional algorithm TICS with fast incremental update support. It allows incremental update of rule set by reconstructing and replacing the local data structure, and allows parallel synchronous execution of search and update through appropriate memory management. The experimental results show that TICS is at least an order of magnitude faster than the current fastest algorithm BRPS, with less memory consumption and good parallel scalability.