Efficient fair queueing using deficit round-robin
IEEE/ACM Transactions on Networking (TON)
Latency-rate servers: a general model for analysis of traffic scheduling algorithms
IEEE/ACM Transactions on Networking (TON)
Preserving quality of service guarantees in spite of flow aggregation
IEEE/ACM Transactions on Networking (TON)
Determining End-to-End Delay Bounds in Heterogeneous Networks
NOSSDAV '95 Proceedings of the 5th International Workshop on Network and Operating System Support for Digital Audio and Video
Traffic scheduling in packet-switched networks: analysis, design, and implementation
Traffic scheduling in packet-switched networks: analysis, design, and implementation
End-to-end delay bounds for traffic aggregates under guaranteed-rate scheduling algorithms
IEEE/ACM Transactions on Networking (TON)
Effect of flow aggregation on the maximum end-to-end delay
HPCC'06 Proceedings of the Second international conference on High Performance Computing and Communications
Instant service policy and its application to deficit round robin
AAIM'06 Proceedings of the Second international conference on Algorithmic Aspects in Information and Management
Hi-index | 0.00 |
Because of the scalability problem, the aggregation of flows and the queueing/scheduling based on those flow-aggregates is unavoidable in Quality of Service (QoS) architectures for large scale networks. We investigate the effect of flow aggregation on the end-to-end delay bounds. It has observed that with traditional work-conserving schedulers, the maximum burst size of each flowincreases linearly as it traverses the network. The increased maximum burst size does not affect the delay bound of a flow in cases where the schedulers are flow-based. In cases where deaggregation and aggregation take places in the middle of the network, however, the increased maximum burst size affects severely in terms of delay bound. This is in fact the case for the most of real network deployments since at the edge of a subnetwork the flows have to be deaggregated and then handed over to another subnetwork. We suggest a simple alternative to the existing work-conserving scheduler, the Smoothing-DRR (S-DRR) server, which is based on the Deficit Round Robin (DRR) server. S-DRR has a non-work conserving characteristic. S-DRR is proved to suppress the maximum burst size of the aggregated flow to a constant throughout the path, so that the delay bound is only linearly proportional to the hop counts.