TCP and UDP performance for internet over optical packet-switched networks
Computer Networks: The International Journal of Computer and Telecommunications Networking
JumboGen: dynamic jumbo frame generation for network performance scalability
ACM SIGCOMM Computer Communication Review
| Hi-index | 0.00 |
Abstract: With the emergence of the new multimedia applications and the large standardization effort of packet telephony products, a large volume of small packet traffic flows in the network. And it is expected to get worse with the popularity of VoIP applications. To mitigate the impact of small packets we propose the aggregation of packets present in output queue and which have the same forwarding equivalence class, such as those which will leave a routing domain from the same node (an egress node). The set of aggregated packets will be subsequently treated as a single packet, thus requiring smaller processing overhead at intermediate routers. We report the results of several simulations that evaluate the performance benefits of aggregation. We note that aggregation keeps the same behavior as pure FIFO. Thus it doesn't disturb the aggregator router behavior. It saves CPU processing time and buffer space at intermediate routers, thus improving the overall network performance.