Answers are grounded in this paper and cite their sources.
In any System on a Chip (SoC) or networking communication device, residence time is calculated by subtracting the time when the message was received from the time when the message was sent. If an adjustment is made to the actual time between the time that the message arrived to before the message is sent, then this adjustment will create an error in the measured residence time. In order to prevent this error, two things can be done: make no adjustments while one or more messages that require residence time calculations are present in the local device, and recognize that an adjustment was made and provide a means to compensate for that adjustment. Not making the adjustment delays the time at which the two-clock variation is reduced, thus resulting in less accuracy, and also require that a device control and keep track of how many outstanding residence time calculations there are at any given time. Part one of this solution uses the second technique in order to maintain higher accuracy and to reduce the burden of having to keep track of the outstanding number of calculations.
No takes yet. Share an insight, caveat, or question.
Dharane et al. (2017) studied this question.