Experiments with packet switching of
`voice traffic
`
`P.N. Clarke. B.Sc., Ph.D., and Prof. L.F. Turner. B.Sc.. Ph.D.
`
`Indexing terms:
`
`Telephone exchanges and networks. Voice trafiic
`
`Abstract: There has been much interest recently in integrated services digital networks carrying both voice and
`data traffic. Packet switching is being used to carry data in an attempt to make better use of trunk capacity
`than with circuit switching. In a telephone conversation, for most olthe time only one person is talking. and it
`has been suggested that packet switching can lead to economies in carrying voice traflic also. In view or the
`variable delays associated with store and forward switching. buffering is usually required at the receiver to
`enable received speech to be reconstituted at the proper rate. Simulation experiments of packet switching of
`voice traffic with fixed packet routing have been carried out. The results of these simulation experiments. which
`are described in this paper, show that, for a single link between two exchanges. 22 conversations can be carried
`by packet switching with reasonable delay. For the same inter-exchange-link capacity. only 15 conversations
`can be carried by circuit switching. For a larger network with more exchanges and links per path. a similar
`advantage is also found with packet switching. The results show that
`the standard deviation of intcrpacket
`delay for successive packets of the same talkspurt is an order of magnitude less than the standard deviation or
`packet transit time for all packets. This suggests correlation cl’ flows of packets within the same talkspurt. The
`wider variation of transit delay applies to each lalkspurt as a whole and all packets within the talkspurt have
`correlated transit times, and hence interarrival times. The fact that the standard deviation ofinterpacltet delay is
`small as compared with the standard deviation of packet
`transit
`time suggests that
`the receiver buffering
`requirement is less than that indicated by the standard deviation ofthe packet transit time.
`
`1
`
`Introduction
`
`As a result of the recent increases in data traffic, various
`suggestions have been put forward relating to the use of
`separate data networks. The existing analogue circuit-
`switched telephone network has transmission and noise
`characteristics which
`vary significantly
`through
`the
`network, and call set-up times of the order of seconds are
`involved. Although this situation is acceptable in so far as
`voice traffic is concerned, it is unacceptable for many data
`applications. On account of the burst-like nature of the
`data,
`in many applications store and forward switching
`methods, such as packet switching, have been proposed
`and implemented [I-6]. Packet switching makes better use
`of expensive high-capacity irtterexchange trunks by trans-
`mitting small biocks of data. or packets, only when there
`are data to be sent. If there are, for short periods, more
`packets for transmission than can be dealt with, some are
`stored for forwarding in later less busy periods. Packet
`switching makes eliicient use of trunk capacity at
`the
`expense of variable delay.
`Rather than have two separate networks, one for data
`and one for voice trallic, a single network for both types of
`traliic may be more economical. As digital
`transmission
`and switching methods are being used increasingly for
`speech, and as data are best handled in digital form, inte-
`grated services digital networks (ISDN) are being pro-
`posed. These might
`be
`implemented using the new
`electronic digital circuit-switching exchanges, such as in
`System X [7]. Alternatively, depending on the relative
`costs of switching and transmission, packet
`switching
`might be used to make use of trunk capacity during silent
`periods. Systems such as TASI [8] have been used in the
`past on both transocean cable and satellite circuits in
`order to make use of silent periods.
`Packet switching with its variable delays might be con-
`sidered unsuitable for real-lime application such as conver-
`sational speech. If, however, a buffer is used at the receiver
`
`Paper 25865. first received 3rd June 1932 and in revised fonn l'l'th February I983
`Professor Turner is, and Dr. Clarke was fonnerly. with the Department of Electrical
`Engineering. Imperial College of Science & Technology. South Kensington. London
`SW? 251’. England. Dr. Clarke is now with British Telecommunications, Gower
`Street. London WCI F.6BA. England
`
`LEE PROCEEDINGS, Vol. J38. Pt. G. No. 4. AUGUST I983
`
`can be
`times
`arrival
`in packet
`then the variations
`smoothed out and the received speech reconstituted at the
`correct rate. This does, of course, add to the total speech
`delay. The total delay resuiting from packet creation,
`network transit time. and receiver buffering and decoding
`must not be too long (cf. 270 ms l-way delay through a
`satellite link). It has been observed [9] that delay in excess
`of 900 ms can give rise to considerable difiiculties. Replies
`and nonverbal
`responses,
`together with their
`relative
`timings, provide the speaker with clues as to the listener's
`understanding and thus aid the conversation process.
`Minoli [10, ll] considered theoretically talker behav-
`iour and end-to-end, that is, packet transit delay for a link
`packet-switched voice system. He also considered delay
`dependencies on packet size and the efiects of the number
`of queue buffers at
`the link output. Coviello [12] also
`considered end-to-end delay for a variety of network par-
`ameters and a variety of alternative network protocols to
`facilitate packet switching of voice traffic. Gruber [13]
`reviews a variety of switching techniques for voice traffic
`and is again concerned with cnd-to-end delays. A variety
`of speech coding techniques are reviewed and the results of
`some ARPA network voice experiments are described by
`Gold [14].
`These works [l0—l4] have been concerned very largely
`with the end-to-end, or packet transit, delay and its varia-
`tion, and the workers invoived have considered this varia-
`tion to be the principal factor determining the buliering
`requirement at the receiver; with the buffer being necessary
`to even out irregular packet arrivals. Although the packet
`transit time, if large, and its variation may have a signifi-
`cant effect on conversational behaviour (see Reference 9), it
`is, however, the variation of interpacket delay. rather than
`packet transit time, which determines the receiver buffering
`requirement. In the experiments carried out and described
`in this paper
`the interpacket delay (that
`is,
`the delay
`between arrivals of successive packets within the same
`tallcspurtj and its standard deviation were measured in
`order to investigate the correlation of packet flows. The
`results of simulation experiments carried out with a fixed
`packet routing system Show that the standard deviation of
`interpackcl delay for successive packets of the same talk-
`spurt is an order of magnitude less than the standard devi-
`IDS
`
`Apple 1017
`U.S. Pat. 7,535,890
`
`

`
`ation of the packet transit time. This thus suggests that the
`receiver buffering requirements are significantly less than
`suggested by packet transit-time statistics.
`This paper describes an investigation into the delays
`involved in the use of packet switching for voice traffic. In
`the course of the investigation, a computer simulation
`model was devised and this is described in Section 2 ofthe
`paper. The experiments carried out
`and the results
`obtained are described in Section 3, and some conclusions
`to be drawn from the work are presented in Section 4.
`
`2
`
`Packet-switched voice network simulation
`model
`
`The simulation model developed will be described in two
`parts:
`(a} the talker activity model (Section 2.1)
`(b} the packet-switched network model (Section 2.2}.
`
`Part (a) deals with the nature of the interaction between
`the talkers, and part (b) with the packet-switched network
`itself, which transports the speech in packet form.
`
`2.1 Talker activity mode!
`In most conversational speech between two people, one is
`silent at any given time (listening while the other
`is
`talking). There are, however, occasions when both are
`silent and or when both are talking simultaneously (e.g.
`when one person interrupts the other). Talker activity can
`be thought of in terms of active periods (talking) or silence
`periods. These periods can be the main active periods of
`significant utterances, such as sentences, and the silence of
`a listener while another person is talking. Alternatively, the
`fine structure of the significant utterances can be taken
`into account. This fine structure refers to the actual time
`
`during which a sound is being made by a talker and the
`pauses between sentences, words and syllables.
`The principal object of a packet-switched network is to
`make efficient use of network transmission capacity. It is
`thus clear
`that packets should only be carried by the
`network for any conversation, while either of the parties of
`that conversation is actually speaking.
`in this way,
`the
`silence periods of conversation can be filled in on the high-
`capacity trunks which are shared by many talkers. A larger
`number of talkers can thus use a given trunk capacity than
`with circuit switching. Speech detection equipment should
`produce an output to be put into packets according to the
`coarse or the fine structure of talker activity, depending on
`the speech-detector sensitivity and switching speed.
`Studies have been carried out of the talker activity
`during telephone calls. Norwine and Murphy [IS] con-
`sider principally the coarse structure of the interactions
`between talkers. Brady describes an experimental arrange-
`ment for measuring fine structure of talker activity [16],
`the analysis of data gathered using this apparatus [17] and
`the fitting of such data to a theoretical model for gener-
`ating probabilities of transition between states of talking,
`silence, interruption etc. [18].
`the simulation
`The talker activity model used for
`experiments, and reported on in this paper, was based on
`the results given in Figs. 3 and 5 of the paper by Norwine
`and Murphy [15], and thus does not
`take account of
`pauses within talkspurts.
`It would have been possible,
`using a Markov chain model,
`to obtain finer details of
`talkspurt activity, but as this approach is considerably
`more difiicult to implement than the probability density
`function approach, it was not adopted in the simulations
`leading to the results presented in this paper. However, an
`approach involving the consideration of the finer details of
`I06
`
`
`
`talkspurt activity may well be of value, and could form the
`basis of a more extensive further consideration of packet-
`switched systems used for the transmission of voice traffic.
`In Reference 15, graphs are given of talkspurt length and
`response time distributions, with response time being
`defined to be the length of time between the end of one
`talker‘s talkspurt and the beginning of the next talker’s
`talkspurt. The distribution of response time includes nega-
`tive values,
`that
`is,
`interruptions. A positive value of
`response time corresponds to the more normal period of
`mutual silence between talkspurts before the next talker
`begins. Using the talkspurt duration statistics given in Ref-
`erence IS, the talkspurt duration was approximated in the
`work reported on in this paper, using a lognormal dis-
`tribution [19] having the same mean and modal values.
`The lognormal distribution has a PDF,f(x}, given by
`
`_
`
`f(x) _ x. /21:62 exp i
`
`I
`
`-0089* - #12
`
`252
`
`i
`
`where _l.t and tr‘ are the parameters of the distribution.
`With the mean and mode of the distribution, as given in
`Reference
`15,
`,u = 0.435
`and 01 = l.8'r'l. As
`regards
`response time, this was approximated using a normal dis-
`tribution with mean 0.32 and standard deviation 0.584 (all
`times in seconds].
`With the model used,
`defined to be:
`.
`.
`mker acnmy :
`
`1
`I
`mean ta kspurt ength
`2{mean talkspurt length + response time]
`' can be seen to be:
`
`the talker activity, which is
`
`talker activity =
`
`4.14
`
`2(4.l4 + 0.4!)
`
`= 0.45
`
`(or 45%}
`
`In the model used in the simulation, a talker was allowed
`to talk for a talkspurt length, with the length being drawn
`from the lognormal distribution. The response time for the
`second talker was drawn from the normal distribution.
`After the talkspurt length, the first talker stops, and the
`second talker is allowed to begin at a time equal to the
`sum of the talkspurt length and the response time after the
`start of the first talker’s talkspurt. The length of the talk-
`spurt for the second talker was determined from the log-
`normal distribution. In this way the times for the second
`talker to stop and for the first talker to begin again were
`determined. It as a result ofa combination of interruptions
`and long talkspurts a talker was scheduled to start a new
`talkspurt during the course of an existing talkspurt, it was
`arranged for the current talkspurt to be completed before
`the start of the next, which was then allowed to begin
`immediately afterwards. These points are illustrated by the
`simple example shown in Fig. 1.
`in the Figure; at time A, talker 1 (T1) begins to speak
`until 8. T2 is idle at time A and is scheduled (by T1) to
`start speaking at C. At time B, T] stops and becomes idle
`and T2 is idle but waiting to start at time C. At C, T2
`begins to speak until E and schedules Tl, who is idle, to
`start at time D. This represents an interruption by T1 who
`will start talking before T2 has finished. At
`time D, Tl
`begins to speak until H. T2 is scheduled by T1 to start his
`next talkspurt at time F which is thus an interruption of
`T1. T2 stops talking at E and awaits a new start at F. At
`time F, T2 interrupts T1 and schedules Tl’s talkspurt to
`start at 1. T2 stops talking at G and TI carries on until H.
`Tl stops at time H and remains idle until the next start at
`I. At time I, Tl begins to speak until time M and schedules
`T2 to start at time J. T2 starts speaking at J, interrupting
`T1 and schedules T] to start at time L. T2 stops at time K.
`
`.l‘EE PROCEEDINGS, Vol. I30, Pt. G, No. 4, AUGUST 3983
`
`

`
`l I
`
`I
`I
`l
`I
`I
`
`l
`I
`1
`l
`I
`I response |
`time (-we)
`I
`I
`I
`I
`I
`
`II
`
`Fig. 1
`
`Example nfrallrer-or.-riuiry model
`
`At L, T] is still talking, so he continues the current talk-
`spurt {until M) and restarts immediately until N. Also at
`time M, T2 is scheduied to start at time 0, and so on.
`In the simulation, the following procudure was adopted.
`During talkspurts, the speech from talker’s equipment was
`taken as having been digitised with all talker pairs in the
`network having the same speech bit rates. When enough
`8-bit (byte) speech digits to fill a packet had been received
`from a talker, a packet was created at the exchange. An
`appropriate header was added to the packet which then
`went
`for
`transmission through the network. The next
`packet of the talkspurt was then filled up, and so on. At
`the end of the talkspurt, the packet which was being filled
`up was completed by filling with ‘blank’ infonnation at the
`speech bit
`rate (see Fig.
`2}. All speech packets in the
`network were thus of the same length. All packets as well
`as being of the same length were created at regular inter-
`vals during the talkspurt.
`Clearly,
`this simple model of the coarse structure of
`talker activity, and regular packet generation, makes no
`allowance for the possibility, depending on the nature of
`the interruption, of a talker stopping when interrupted. No
`allowance was made in the simulation model for the effects
`
`on talker behaviour of delay in packet creation, of cross-
`network delays. nor of buffering and speech reconstruction
`delays. All of these delays will
`in general be variable,
`except the regular packet creation delay. Delays in tele-
`phone channels do affect
`talker behaviour, as has been
`reported by Brady [20] in the case of fixed delays. The
`simple model was chosen to provide approximate conver-
`sational talker activity.
`
`_
`packet
`-_genercIlron
`l|l'l'IE5
`
`Fig. 2
`Talker activity and packet creation
`maximum data content of packet (bits)
`Speech bit rate (bills)
`
`res PROCEEDINGS. Vol. :30. Pt. 6, No. 4. A UG U51" 1933
`
`The rationale behind the simplfied approach was that if
`this model which does not allow for delays in speech, and
`operates by generating full packets at regular intervals, can
`handle more calls than a circuit switched system of the
`same trunk capacity,
`then a more complicated model,
`allowing for delays and pauses within talkspurts, may
`allow even more calls to take place.
`
`2.2 Packet-switched network model
`
`The network of the simulation model was made up of
`packet—switching exchanges
`(PSEs} connected by full-
`duplex trunks. The
`talkers were
`connected to the
`exchanges by lines which can be assumed to be either ana-
`logue or digital {operating at the speech bit rate). In all the
`examples, each talker was associated with another talker at
`another PSE in the network. These talker pairs were
`assumed to be engaged in conversation before the start of
`each experiment.
`On generation of a speech packet at a talker‘s interface,
`the required outgoing trunk was determined by consulting
`the route table. Fixed routing was used in all of the experi-
`ments. Packets entering the network from a talker could
`only be put into the queue for this trunk if there were more
`than two free queue bullets. This gives some priority to
`transit trallic, i.e. to packets which have been accepted into
`the network, for example, at node 4 in Fig. 4b, or at node 2
`in Fig. 4:’: for packets between t and 3, and between 3 and
`1. If an originating packet could not be accepted, it was
`held in a buffer associated with the talker’s interface to the
`network. That
`talker‘s identity was put
`into a queue
`associated with the trunk output queue. Whenever a
`packet was sent along the trunk and a queue bufl‘er
`became free,
`the list of talkers with waiting packets was
`inspected. If there were sufficient buffers to allow in an
`originating packet, the first one waiting joined the trunk
`output queue. If that talker had further waiting packets, he
`rejoined the list of talkers with waiting packets.
`Packets were transmitted over the trunks at the trunk
`
`transmitted packets were kept,
`rate. Copies of all
`bit
`pending acknowledgments received from the other end of
`the trunk. Associated with each copy of a packet, kept in
`the retransmission queue, was a time by which that packet
`must be acknowledged. This time was based on the worst
`possible case of acknowledgment delay. Acknowledged
`packets were deleted from the retransmission queue. If a
`I07
`
`

`
`packet were to exceed its time in the retransmission queue,
`then it would be retransmitted, followed by its successors
`(unless these had meanwhile been deleted) before any new
`packets were transmitted. However, as transmission errors
`were not simulated,
`the only condition under which the
`retransmission procedure could have been evoked was that
`in which a packet was discarded at a transit node because
`of there being no free bufi'e1's in the output queue.
`The acknowledgment process was carried out using the
`send-and-receive sequence numbers carried by all packets
`as used in the lSO‘s HDLC and in the CCITT’s X25
`recommendation [2], 22]. Any packet carrying a send
`sequence number greater than that expected was discarded
`and a REJ (Reject) packet sent
`in the reverse direction.
`This REJ packet
`indicated the last correctly received
`packet and instructed retransmission to start at the appro-
`priate point
`in the packet sequence. Only one REJ was
`allowed in a given direction until the next expected packet
`was received. If a REJ was corrupted by noise and thus
`discarded. the correct packet sequence was maintained by
`retransmission invoked by the timeout mechanism. In the
`case of no outgoing packets when one was correctly
`received. a RR {receiver ready) packet
`indicating correct
`reception was sent. This reduced the use of the timeout
`mechanism under conditions of light trunk loading.
`Packets made their way through the network to their
`destination. Here they were assumed to be passed to the
`receiver interface for conversion to speech {after any buf-
`fering, if necessary}. On arrival of every packet, the packet
`statistics were
`updated. Packet
`statistics measured
`included:
`
`(i) the number of packets received
`(ii) the mean and standard deviation of packet transit
`time for the packets of (i). Packet transit time was mea-
`sured as the difference between the arrival
`time at
`the
`destination PSE and the packet creation time at the source
`PSE
`
`(iii) the mean and standard deviation of packet inter-
`arrival
`time. Packet
`interarrival time was defined as the
`difference between arrival times of successive packets of the
`same talkspurt.
`
`The simulation program was written in Simula [23, 24]
`and was designed to be as flexible as possible. A wide
`variety of networks and conditions could be simulated by
`choosing appropriate input data for the program. The
`input data required for this were:
`(i) the number of PSEs
`(ii) the number of trunks
`the source and destination of
`(iii) for each trunk: (a)
`PSEs, (b) the trunk capacity, (:2) the bit error probability,
`and (cl) the retransmission timeout period
`(iv) the route table (this gives the next PSE en route to
`each destination)
`(v) the talkers‘ speech bit rate (the same for all talkers)
`(vi) the speech packet length {in bytes)
`(vii)
`the number of PSE. pairs with conversations
`between them
`
`(viii) for each of {vii} above, the number of talker pairs
`(ix) the duration of the simulation and intervals between
`statistics report
`(x) the seed for the random—number stream used.
`
`3
`
`Packet-switched voice network experiments
`
`3.1 General description
`Three simulation experiments were carried out, and there
`were several model parameters common to the experi-
`ments. The maximum trunk output queue length was ten
`I03
`
`
`
`packets (with two reserved for transit traffic). The talker
`speech bit rate"' was 9600 bit/s. There were no local calls
`(i.e. calls between talkers at the same PSE}. The program
`was run for 250 s in all experiments and the results of the
`first 100 s were removed in order to reduce the bias effects
`of no packets being present in the network at the start of
`the simulation. Analysis of the results has shown that a
`stable condition was reached in this time. Results were also
`collected for a single talker pair. The three experiments
`carried out were as follows:
`
`{it Two PSEs, one 144 kbitfs trunk (see Fig. 3); 128 + 8
`{speech + header) byte packets; 25 ms
`retransmission
`timeout interval; varying number of talker pairs.
`(ii) Two PSEs; one 144 kbit,r‘s trunk (see Fig. 3); 15, 20
`and 25 talker pairs, packet sizes of 32, 48, 64, 96, 128 (from
`previous experiment) 192 and 256 bytes (with 8 bytes of
`header
`in
`addition with
`correspondingly
`adjusted
`retransmission time.
`
`(a) a fully connected
`(iii) Three PSES (see Fig. 4):
`network with 144 kbit,='s trunks; (b) a star network with
`288 kbitfs trunks; and (c) a linear network with 288 kbitsfs
`trunks; 128 + 3 byte packets; 12.5 ms retransmission
`
`:
`
`PSEI
`
`trunk
`
`PSE2
`
`,
`
`to
`talkers
`
`Fig. 3
`
`2-node packer-switched uoice network
`
`C
`
`I
`
`3-node packet-switched networks
`Fig. 4
`a Three nodes: FC, in Three nodes; star. c Three nodes: linear
`
`" A 9600 bitsfs speech rate was used in order to facilitate the simulation. The
`significance of the results so obtained is not, however. restricted by this. Appropri-
`at: time scaling of packet lengths and trunk-line rates would render them applicable
`at a more realistic speech data rate o|'64 kbitfs.
`
`IEE PROCEEDINGS, Vol. 130, Pt. G, No. 4, AUGUST 1983
`
`

`
`of [07 + 20 = I2? ms (before receiver buffering]. For more
`than 22 talker pairs, the packet transit time and standard
`deviation will lead to even greater delays. It will be noticed
`that there is a difference between the packet transit time
`and standard deviation curves for all talker pairs and for
`single talker pair {see Figs. 6 and 7}. This is because of the
`effects of the smaller sample size of packets from the single
`talker pairs (see Fig. 5). The single talker pair results will
`not be considered in the rest of this paper. Packet switch-
`ing appears to be able to carry the conversations of 22
`talker pairs (under the above conditions) before delays
`become unacceptable. A 144 kbitfs trunk operating under
`circuit switched conditions can carry (l44000,i'9600) = 15
`conversations with no variable delay.
`2.0
`
`_L9l‘.€£.l!.='_1T_$_
`all
`
`.uI.
`
`
`
`
`
`SDofpndiettransittime-,5F-3.-ua0
`
`timeout period on the faster 288 kbitsfs trunks of (b) and
`(c}; number of talker pairs varied. (The abbreviation FC is
`used in the Figures to refer to the fully connected network
`configuration.)
`
`3.2 Results of experiments
`The results of the experiments will now be described.
`Related points in all Figures are joined by straight-line
`segments to identify related points in the multigraph
`Figures and to indicate trends, rather than to show exact
`behaviour, between the experimental points.
`
`3.2.? Two P.S‘Es, 128+8 byte packers. varying number
`of talker pairs: The number of packets transferred in the
`150 s (for each value of talker load} of the experiment for
`all talker pairs and for the single talker pair are shown in
`Fig. 5. The number of packets for all pairs rises almost
`linearly up to 25 talker pairs. with a smaller rise between
`25 and 27. For the single talker pair, almost
`the same
`number of packets are carried at all loads (total number of
`talker pairs). The average packet
`transit
`time of Fig. 6
`shows little increase up to 22 talker pairs but shows an
`increasing rate of increase above 22 pairs. The average
`packet transit time must be added to the packet creation
`time of(l28 x 839600) 5 = 10? ms to obtain the total delay
`between speech being uttered and becoming available for
`reconstruction on arrival at the destination PSE. Any buf-
`fering to allow for variations in arrival
`times must be
`added as well. Up to 22 talker pairs, the transit time is less
`than 20 ms. The standard deviation of packet transit time,
`shown in Fig. ‘i’, is low [less then 30 ms. suggesting receiver
`bufi‘ering of over 100 ms) up to 22 talker points, but it
`increases more rapidly as more talkers are added to the
`network. This suggests that up to 22 talker pairs with
`speech bit rate of 9600 bitfs, with 128 + 8 byte packets,
`can share a 144 kbitfs trunk with an average speech delay
`
`tulkgr $1;
`all
`
`U1
`0
`10
`
`15
`
`20
`talker pairs
`
`25
`
`30
`
`Packers transferred, varying talker
`Fig. 5
`length = £28 + 8
`
`lead.
`
`two nodes. packet
`
`tclttmpsira
`uu
`
`Fig. 3''
`
`SD of packet transit tirne. varying tuiker limits
`
`The average interpacket delays of Fig. 8 are dominated
`by the 107 ms packet creation delay and are almost equal
`to it for up to 25 talker pairs. In fact, the variation in delay
`due to queueing was found to be approximately three
`orders of magnitude less than the transit
`time. and to
`exhibit no systematic variations)‘ This indicates that. even
`with the extra transit time (queueing for transmission over
`the trunk), there is little difference between the admission
`queueing and transmission delays for successive packets of
`talkspurts. The increase in interpacket delay for more than
`25 talker pairs indicates that successive packets of each
`talkspurt take longer to reach their destination than each
`of their predecessors. A packet-switched network is clearly
`unsuitable for carrying speech traffic when operated in this
`region. The standard deviation of interpacket delay is less
`than 4 ms for less than 22 talker pairs. This is approx-
`imately an order of magnitude less than the standard devi-
`ation of packet transit time.
`
`
`
`
`
`avemgeinterpacketdr-In_
`
`I5
`
`20
`talker pairs
`
`25
`
`Irirerpacket delay, earyirly talker loads
`
`Fig. 6
`
`Pocket transit time, unrying talker load
`
`talker pairs
`
`IEE PROCEEDINGS. Vol. I30. Pt. G. No. 4. AUGUST I933
`
`1‘ Details of the effects of speech statistics on the pcreeplion of irnpairmcnls nrising
`from variable delays can be found in References 13 and 25.
`
`I09
`
`

`
`This suggests greater correlation between arrivals of
`successive packets of the same talkspurt than indicated by
`the transit-time figures. The packets are generated at
`regular intervals and, for long talkspurts {with respect to
`packet creation time), the packets of all active talkers are
`correlated. This correlation is disturbed slightly when a
`talker stops or when another joins the set of active talkers.
`Once the transit delay is determined, all the packets of the
`same talkspurt have similar transit times, and thus inter-
`packet delays are similar. The transit
`time indicates the
`delay in admission queueing and transmission, and thus
`represents the storage of packets within the network.
`0.020
`
`C’ S’ at
`
`delay,5.0ca0‘<23o
`SDofinterpacl-tel
`
`10
`
`15
`
`20
`tel ke r pa i r 5
`
`25
`
`30
`
`Fig. 9
`
`SD of tnterparlret tlelays. varying talker load
`
`the
`Variation in this for each talkspurt does not affect
`interarrival
`times of the packets and hence the speech
`reconstruction. The standard deviation of
`the packet
`transit time is thus a measure of the spread of time spent in
`the network. The standard deviation of interpacket delay is
`the measure which should be used in deciding on receiver
`buffering requirements. Several
`times this standard devi-
`ation should be allowed in the receiver buffer to minimise
`
`the number of late packets which will have to be dealt with
`in some way.
`The trunk utilisation, shown in Fig. 10, rises approx-
`imately linearly with the number of talker pairs. For less
`than 22 talker pairs, the utilisation is less than 80%. For
`more than 22, the utilisation is greater than this, with the
`associated rapid rise in transit times, as can be seen from
`the packet transit time of Fig. 6. It should be noted that
`certain parameters such as the number of packets trans-
`mitted and the trunk utilisation {which are shown in Figs.
`5 and I0, for example] can be calculated from a knowledge
`of talker activity, packet generation rate, packet
`length,
`trunk capacity and the number of talkers. It should also be
`noted that the number of talker pairs required to achieve
`100% trunk utilisation can be calculated using the packet
`generation rate, talker activity and trunk capacity. If this is
`100
`90
`80
`
`‘tn_.HaslbU’!0'!-.1so00cl:3ooo
`lineutttisation_
`
`15
`
`20
`talker pairs
`
`25
`
`30
`
`Fig.1O Trrmlr utilisation. varying talker
`It.-nytlr = £28 + 8
`
`load,
`
`two nodes. packet
`
`I10
`
`
`
`done, then it is found that 31 talker pairs are required, and
`this agrees with extrapolation of Fig. 6. However, it is clear
`that before this number of talker pairs is actually reached
`the variations in delays are such as to render speech trans-
`mission unacceptable.
`
`3.2.2 Two PSEs, varying packet sizes: The different
`packet sizes used inithis experiment, with their creation
`times and retransmission timeout periods on the 144 kbit_t‘s
`trunk, are shown in Table 1. Totals of 15. 20 and 25 talker
`pairs were used in the experiment with each value of
`packet length (except for 25 pairs 32 + 8 bytes].
`
`Table 1 2 Packet details
`
`Packet length
`
`Packet creation
`time
`
`Retransmission
`timeeut
`
`ms
`ms
`bytes
`135
`26 213
`32 + 8 (B header}
`10,29
`40
`48 + 8
`13.24
`501t3
`E-4+8
`19.12
`30
`96 + 8
`25.00
`50 2,t3
`128 + 8
`3636
`160
`192 + 8
`
`255 + B 48.53 213 1:3
`
`
`The numbers of packets carried are shown in Fig. 11. Here
`it can be seen that the number of packets carried rises as
`the packet
`length decreases. Obviously, more shorter
`packets are required to carry the same quantity of speech.
`The average packet transit times are shown in Fig. 12. As
`before, the packet transit time is large (over 100 ms} for 25
`talker pairs. Here, the packet transit time rises for packet
`lengths greater than 128 + 8 bytes. This is expected, since
`longer packets will obviously take longer to traverse the
`network. Below 128 —+- 8 bytes, the packet transit time also
`
`
`
`30
`
`CI
`
`22!.
`192
`160
`12B
`96
`6-5
`32
`data characters per packet (header = 8)
`
`256
`
`Packets transferred. all talker pairs, varying packet lengtli. two
`
` E}
`
`0
`
`128
`96
`6!.
`32
`I60
`192
`22-‘.
`data characters per packet
`[header : 3)
`
`256
`
`Fig. 12
`
`Parker transit time. all tall-ter pairs. varying packet length
`
`LEE PROCEEDINGS, Vol. I30, Pt. G, No. if, AUGUST 1983
`
`

`
`increases. This rise of transit time in this region is due to
`the increased quantity of overhead (8 bytes per packet) of
`the headers of the larger numbers of smaller packets. This
`increases the trunk utilisation above 90% (as can be seen
`in Fig.
`16} and increases the transit
`time. Fig. 12 also
`shows a rise in packet transit times below 48 + 8 bytes for
`20 talker pairs, where the packet transit times is lower than
`for 25 pairs in any case. Referring to Fig.
`[6 again, only
`below 48 + 8 bytes does trunk utilisation rise above 80%
`for 20 talker pairs. It can thus be seen that higher loading
`leads to greater packet
`length below which the header
`overhead