Random early detection gateways for congestion avoidance
IEEE/ACM Transactions on Networking (TON)
High performance TCP in ANSNET
ACM SIGCOMM Computer Communication Review
IEEE/ACM Transactions on Networking (TON)
The BLUE active queue management algorithms
IEEE/ACM Transactions on Networking (TON)
Provisioning internet backbone networks to support latency sensitive applications
Provisioning internet backbone networks to support latency sensitive applications
Proceedings of the 2004 conference on Applications, technologies, architectures, and protocols for computer communications
Characterizing residential broadband networks
Proceedings of the 7th ACM SIGCOMM conference on Internet measurement
Medium access control protocols performance in satellite communications
IEEE Communications Magazine
| Hi-index | 0.00 |
A critical component of subscriber management in a DOCSIS-based cable access network is the buffer management strategy that is in operation at the upstream service flow queue located in the cable modem. The strategy must contend with conflicting goals: large buffers might be required to ensure TCP flows can utilize available bandwidth, however large buffers can impact application flows that are latency sensitive. In this study, we have explored the relationship between application performance and upstream queue management. We define the optimal queue capacity as the buffer size that maximizes throughput and minimizes packet delay. Our results agree with previous work in wired Internet router contexts that suggest that the queue capacity should not exceed a bandwidth*delay product (BDP) amount of data. However, the upstream data rate available to a cable modem varies with the number of competing cable modems. Upstream service flow queues that are provisioned to hold a BDP of data might suffer queue delays exceeding several seconds during periods of congestion. Active queue management such as the Random Early Discard algorithm is able to reduce average queue levels but is not able to provide a consistent balance between the needs of both high throughput and latency sensitive applications. The conclusion is that an adaptive queue management algorithm is required to maintain a consistent balance between throughput and delay.