The traffic demands in optical networks are often irregular associated with various physical constraints such as SRLG disjoint, taboo link, taboo node and so on. The virtual topology design methods in the literature could hardly to solve traffic grooming problems with multiple kinds of constraints. We propose a packing grooming with max revenue algorithm (MRA). The traffic demands are placed directly in the physical topology, which are treated as low-granularity single-hop-pipes. To save the low-granularity switch capacity in the intermediate nodes, MRA checks which demands could be packed into high-granularity multi-hop potential pipes. Among the high-granularity potential pipes, we search the maximum set of potential pipes which do not collide with each other by a max-revenue approach and instantiate them. Numerical results show that MRA algorithm has better performance than existing algorithms in capacity and cost of network, and it could be used in large networks with more than one hundred nodes.