Stream Traffic Communication in Packet Switched Networks: Destination Buffering Considerations

In this paper we consider the problem of sending a stream of data (speech, for example) through a packet-switched network which introduces variable source-to-destination delays for different packets of the stream. Ideally, this delay difference should be smoothed so as to preserve the continuity of the stream. We investigate an adaptive destination buffering scheme which may be used to achieve the smoothing of the output stream. The scheme uses delay information, measured for previous streams, in order to compute destination buffering information. Specifically, of the last
m
packet delays, one discards the largest
k
and then the range of this partial sample is used for the destination wait time
D
. We obtain a rule of thumb for choosing
m
and
k
, and demonstrate its applicability on some empirical delay distributions from ARPANET measurements. It is, in general, necessary to deal with discontinuities which occur even after smoothing. To this end, we consider two possible playback schemes: method
E
(time expanded in order to preserve information) and method
I
(late data ignored in order to preserve timing). The two methods are at opposite ends of a continuum of possible playback schemes. We study the implication of methods
E
and
I
on the choice of smoothing parameters and establish a foundation for evaluating all schemes in this continuum.