Instant Queue Occupancy Used For Automatic Traffic Scheduling In Data Center Networks

Datacenter applications desire low latency for short messages to provide a better user experience. Therefore, one of the goals of datacenter networks is to minimize flow completion time (FCT), especially for short flows. Multiple scheduling disciplines have been proposed to achieve this goal. In this paper, we develop Dynamic Longer Stay Less Priority (D-LSLP) which looks at the current queue occupancy to adjust the demotion threshold for the packets of the flows arriving in the strict priority queues. Initially, D-LSLP considers every flow to be a short flow, and with the passage of time, the flow is demoted to the next priority queue, similar to MLFQ. We leverage the programmable nature of the P4 switches and measure the instant queue occupancy in the P4 to automatically adjust the demotion threshold only for the highest priority queue. This enables DLSLP to increase or decrease its highest priority demotion threshold based on the instant queue status. It enables short flows to complete in a couple of higher-priority queues. In contrast, large flows are demoted to lower priority queues after remaining active in the queues for a certain amount of time. D-LSLP allows multiple traffic patterns to coexist without the need to manually tune the demotion thresholds. Furthermore, it helps reduce tail drop in the highest priority queue when it is overwhelmed by many short flows. The performance evaluation shows that it works well with different traffic patterns without operator intervention.