Preserving quality of service guarantees in spite of flow aggregation

Abstract

We investigate the preservation of quality of service guarantees to a flow of packets in the presence of flow aggregation. In flow aggregation, multiple flows, known as the constituent flows, are merged together resulting in a single aggregate flow. Packet schedulers located after the network point where the aggregation occurred are aware of the aggregate flow, but are unaware of its constituent flows. In spite of this, we show that quality of service may still be guaranteed to the constituent flows if the aggregation is performed fairly. Furthermore, contrary to intuition, the quality of service guaranteed to a flow may be greater under flow aggregation than in the case where no aggregation is performed.

Work-Conserving Fair-Aggregation Across Multiple Core Networks

“…We thus have that the end-to-end delay has a per-hop increase proportional to L R g , as in the case of regular flow aggregation (see (4)). However, we have the additional WFI term W t g .…”

“…To prevent the above, in addition to being a GR scheduler, aggregators must restrict their output rate, and thus be nonwork-conserving [4]. In this case, the per-hop delay of a flow f as it traverses a scheduler t is…”

“…To reduce the amount of state managed by each router, multiple flows can be combined together to form a single aggregate flow [4], [5], [9], [18].…”

“…Fair aggregators provide better performance guarantees to individual flows [4]. In flow aggregation, all flows that are going into and coming out of the core network through the same ingress and egress routers are aggregated together.…”

“…This leads to the design of DiffServ scheduling algorithms. Dynamic Packet State (DPS) [22], [14] and Flow Aggregation [4] are among approaches to make GRS schedulers more scalable.…”