Long-Range Dependence of Traffic across Schedulers with Multiple Service Classes

Long-range dependence (LRD) is a widely verified property of Internet traffic, which severely impacts network performance yielding longer queuing delays. Moreover, LRD has been demonstrated to be almost ubiquitous and hard to remove or control. In this work, we investigated by extensive simulation the effect of schedulers on traffic LRD. We analyzed the output traffic of single schedulers and chains of schedulers merging LRD flows according to various service policies, viz. plain FIFO, strict-priority, earliest-deadlinefirst and general processor sharing (GPS). First, we noticed that traffic LRD is not affected much by crossing schedulers, for any service policy, when the merged flow has no LRD. Then, we showed that LRD may also spread across flows with different service priorities, with any service policy except balanced GPS, which ensures complete separation between classes. Finally, we experienced the same phenomenon also along chains of schedulers. These findings may explain, in part, why LRD is so widespread in Internet traffic