Paper Topics

Traffic Characterization

1) Aggregate Voice Traffic:
a) Requests for network resources from a large population of users.

Fundamental properties of these requests:

- Arrival Rate, Number of requests/time unit, e.g., calls/sec.

- Holding Time, Length of time the request will use the network resources, e.g., min/call.

c) Traffic Intensity is the product of the average holding time and the arrival rate,

d) The traffic intensity is measured in Erlangs or CCS. A CCS is 100 call sec/hr and:

1 Erlang = 36 CCS

e) Typically the traffic intensity is specified for the 'Busy Hour'

f) Both the holding time and the number of requests/time unit are random. Usually it is assumed that the probability density function for the holding time and the time between requests is exponential, i.e.,

2) Individual voice sources: the properties of speech signals

a) Inactivity factor

- Talkspurt duration is exponentially distributed with a mean of 350 ms.

- Silence period is exponentially distributed with a mean of 650 ms.

- Digital Speech Interpolation (DSI) systems take advantage of the inactivity factor to effectively double their capacity. [The analog version developed initially was called of Time Assignment and Speech (TASI) Interpolation]

b) Signal redundancies.


- Highly variable

- Not well known

- Likely to change as new services and applications evolve.

- Highly bursty, where one defination of burstyness is:
Burstyness = Peak rate/Average rate

- Example: During a typical remote login connection over a 19.2kb/s modem a user types at a rate of 1 symbol/sec or 8 bits/sec and then transferes a transfers a 100 kbyte file. Assume the total holding time of the connection is 10 min.

What is the burstyness of this data session?

The time to transfer the file is about 41 sec. so for 600 - 41sec = 559 sec the data rate is 8 bits/sec or 4, 472 bits were transfered.

Thus in 600 sec. 4,472 + 100X8X103 bits were transfered, yeilding a average rate of:

804,472 bits/600 sec = 1,340 bits/sec.

The peak rate was 19.2 Kb/s so the burstyness for this data session was:

19,200/1,340 = 14.3

- Data traffic is characterized by a message ( or packet ) length in bit (or bytes) and an arrival rate.

- The traffic intensity is then:

- The standard assumption is that the packet length and interarrival times have exponential pdf's.