Coflow scheduling with unknown prior knowledge based on traffic characteristics.

Improving data transfer efficiency is critical to reducing application completion time. Existing coflow scheduling schemes fail to take full advantage of the traffic characteristics in the network, especially in coflow scheduling where prior knowledge is unknown. This paper finds that making full use of flow characteristics can further reduce the coflow completion time. We propose a scheduling scheme A-SNCF based on traffic characteristics. Although A-SNCF lacks complete coflow information before transmission, A-SNCF can make reasonable scheduling based on the knowable information during transmission. The scheme proposed in this paper can schedule coflow according to traffic characteristics, which can further reduce the data transmission time. Experimental results on open source datasets show that the average completion time of A-SNCF is 20% less than that of MCS.