Skip to Main Content
The continuous increase of Internet users worldwide, as well as the extensive need to support real-time traffic and bulk data transfers simultaneously, has directed research towards service differentiation schemes. These schemes either propose techniques that provide users with the necessary quality guarantees or follow a "better-than-best-effort" approach to satisfy broadly the varying needs of different applications. We depart from our new service principle called Less Impact Better Service (LIBS) and propose a novel service differentiation method, namely size-oriented dropping policies, which uses packet size to categorize time-sensitive from delay-tolerant flows and prioritize packet dropping probability, accordingly. Unlike existing proposals, the distinction of flows is dynamic and the notion of packet size is abstract and comparative; a packet size is judged as a unit within a dynamic sample space, that is, current queue occupancy. We evaluate size-oriented dropping policies both analytically and experimentally; we observe a significant increase on the perceived quality of real-time applications. Delaysensitive flows increase their bandwidth share, to reach a state of system fairness, regulating the dominant behavior of bulk-data flows.