A Segment-Based SACK Scheme for TCP Over the Error-Prone Links

Abstract

The TCP Selective ACK (SACK) is used to notify information on the sequence space of the receiving buffer to the sender. By definition of RFC 2018, however, each SACK block has to be described by the two 32-bit unsigned integers, which may introduce significant extra load in the reverse ACK path when TCP connection is error-prone. Moreover, due to the 40-byte limitation of TCP option field, the available option space may not be sufficient to report all of the sequence blocks when the errors occur in a sudden burst. This may result in the unnecessary spurious retransmissions while the reverse SACK path is severely congested. To overcome such shortcomings, we propose a segment-based TCP SACK scheme in this paper. Both the theoretical analysis and simulation results show that the proposed scheme not only can significantly reduce the SACK overhead but also can avoid the unwanted spurious retransmissions in some specific error-prone scenarios.

TCP Acknowledgment Optimization in Low Power and Embedded Devices

“…The latter concern is described in [12], and it happens because the CPU load of the network device depends on the number of TCP packets to process but not on packet size. Such condition is more frequently observed on high-speed networks [11], since the increase of data throughput increases the number of the TCP ACK packets on the network [13]. Therefore, handling the heavy traffic of short TCP packets requires more powerful network devices.…”