Output details

A scalable architecture for maintaining packet latency measurements

<06> Originality: First paper to enhance latency measurement granularity from per-flow to per-packet.

Significance: Our core data structure to store packet latency outperforms other solutions; it achieves two times higher storage efficiency, 7.7 times higher estimation accuracy, 1.8 times faster insert operation, and 1.4 times faster lookup operation. The data structure can be widely used for other applications in different areas that require approximated key-value storage (e.g., security applications that store inter-packet times for covert channel detection).

Rigour: Theoretical analysis was conducted to evaluate the efficacy of the proposed core data structure, followed by extensive experiments with real packet-level traces.