The scalability of a parallel simulation of self-propelled microorganisms in a fluid in the regime of low Reynolds numbers is investigated on current HPC clusters. For this, the lattice Boltzmann solver from the WALBERLA framework is coupled with the rigid body physics engine pe. The latter is enhanced with spring-like pair-wise interactions that can not only span over neighboring process domains but as well act as long-range pair-wise interactions between distant process domains. The communication is designed in such a way that only neighboring processes and the processes hosting the objects attached to the springs communicate. The suitability and efficiency of this approach is analyzed by a weak scaling of a representative setup for a swarm simulation on two supercomputers currently in the top ten of the TOP500 list, namely JUQUEEN and SUPERMUC.