Saturday 14 April 2012

Link control


The link control protocol is similar to STR. The designers attempted to protect against simple transmission errors. The EBCDIC CRC-16 used to protect data frames is reasonably strong; the Transcode CRC-12 is somewhat weaker. The protocol requires that every message be acknowledged (ACK) or negatively acknowledged (NAK), so transmittal of small packets has high transmission overhead. The protocol can recover from a corrupted data frame, a lost data frame, and a lost acknowledgment.
Error recovery is by retransmission of the corrupted frame. Since bisync data packets are not serial-numbered, it's considered possible for a data frame to go missing without the receiver realizing it. Therefore, alternating ACK0s and ACK1s are deployed; if the transmitter receives the wrong ACK, it can assume a data packet (or an ACK) went missing. A potential flaw is that corruption of ACK0 into ACK1 could result in duplication of a data frame.
The protocol is half-duplex (2-wire). In this environment, packets or frames of transmission are strictly unidirectional, necessitating 'turn-around' for even the simplest purposes, such as acknowledgments. Turn-around involves
the reversal of transmission direction,
quiescing of line echo,
and resyncing.


In a 2-wire environment, this causes a noticeable round-trip delay and reduces performance.
Some datasets support full-duplex operation, and full-duplex (4-wire) can be used in many circumstances to improve performance by eliminating the turn-around time, at the added expense of 4-wire installation and support. In typical full-duplex , data packets are transmitted along one wire pair while the acknowledgements are returned along the other.

No comments: