Coflow transmission optimization for satellite distributed computing: a joint strategy of delay-balanced routing and conflict-avoiding optical circuit switching scheduling

As the pressure on satellite on-orbit data processing and information backhaul increases, the demand for satellite computing continues to grow. Satellite optical networks must efficiently process the massive data flows generated by satellite distributed computing. However, satellite optical networks face unique challenges. Limited inter-satellite links (ISLs) lead to fierce link contention and significant queuing delays. The non-negligible long-distance propagation delay and optical circuit switch (OCS) reconstruction delay in multi-hop topology introduce significant multi-hop node delays. In addition, given the bandwidth exclusivity of the OCS, it is necessary to use intelligent optical circuit timing scheduling to achieve efficient concurrent transmission, thereby maximizing the utilization of idle links in the network while minimizing the overall reconstruction overhead. These complexities make it difficult for existing coflow scheduling schemes to be directly applied to satellite optical networks. To address these issues, this paper proposes an innovative joint routing and optical circuit scheduling for coflow. The scheme consists of two phases. First, a queue-aware Dijkstra coflow routing (QAD-CR) algorithm is proposed, achieving an effective trade-off between contention queuing delay and multi-hop node delay for coflow routing. Second, a bottleneck-aware graph theory is designed to optimize the optical circuit scheduling. By constructing an optical circuit conflict graph to model the scheduling constraints between flows, a greedy coflow matching algorithm based on maximum degree conflict first (MDCF) is proposed to heuristically find the maximum independent sets, thereby effectively resolving optical circuit conflicts. Simulation results demonstrate that the proposed joint optimization algorithm of coflow routing and optical circuit scheduling significantly reduces coflow completion time (CCT).