Until now, most results reported for parallelism in production systems have been simulation results - very few parallel implementations exists. In this paper, we present the inicial implementation of our parallel OPS5 compiler on the IBM SP2. It is evident that the Rete match algorithm is suitable for parallel processing on distributed memory systems, as supported by the fact that the best performance of our parallel OPS5 compiler is archieved on the benchmark program make-teams with 100 employees using 16 processors and runs 7.76 faster then parallel OPS5 using a single processor. From our observation and experimental results, we believe that enhancements can make our parallel OPS5 compiler more effective. We also analyze the approach to parallelization taken in our prototype compiler and discuss various improvements.