Telecommunication applications require an increasingly high throughput; their task graph often exhibits a high level of coarse grained parallelism. We extend the KPN model by multiple writers and readers and present the SystemC based co-design of a packet classification application on a multi processor system on chip.