Priority Sketch: A Priority-aware Measurement Framework

Network measurement plays a crucial role in modern communication systems, including satellite networks and data centers. Sketches, a series of probabilistic data streaming algorithms, have been widely used in network measurement due to their efficiency and scalability. However, ensuring the Quality of Services (QoS) for high-priority traffic requires more precise measurement techniques. Despite their widespread use, existing sketches lack the capability to differentiate between priority levels, which is essential for priority-aware measurement. In this paper, we propose Priority sketch, a framework that can make many existing sketches priority-aware. We devise an attention function to adjust the accuracy for different priority levels, and optimize the accuracy by trimming counters. Extensive experiments show that the Priority sketch is $10 \times$ more accurate than the baseline solution.