Network congestion prediction
To control the transmission of real time transport packets over a network, incoming packets received over the network from a transmitter are monitored. A statistical analysis is performed on the received packets to determine network congestion. A message is sent to the transmitter to control a transmission parameter in accordance with the determined network congestion.
1. A method of controlling the transmission of real time transport packets over a network, comprising:
monitoring incoming packets received over the network from a transmitter;
performing a statistical analysis of the received packets to determine network congestion; and
sending a message to the transmitter to control a transmission parameter in accordance with the determined network congestion.
2. A method as claimed in claim 1 , wherein statistical analysis is performed using Bayesian statistics.
3. A method as claimed in claim 1 , wherein the statistical analysis is performed on weighted values of packet inter-arrival times D i .
4. A method as claimed in claim 1 , wherein the statistical analysis is performed on a parameter J i defined by the expression
J
i
=
D
i
Size
of
packet
(
i
)
.
5. A method as claimed in claim 4 , wherein the statistical analysis is performed on the basis that the parameter J i follows a Pareto distribution with fixed minimum and shape parameters.
6. A method as claimed in claim 5 , comprising computing a shape parameter for the Pareto distribution, computing moving averages for the shape parameter, and sending said message when the difference between said moving averages exceed defined bounds.
7. A method as claimed in claim 6 , wherein when the ith packet arrives (i>n2) the difference is computed as
Diff
i
=
MA
i
(
n
1
)
-
MA
i
(
n
2
)
,
where
(
8
)
MA
i
(
n
)
=
∑
i
-
n
+
1
i
ρ
i
n
for
i
≥
n
(
9
)
8. A method as claimed in claim 7 , wherein the shape parameter for the Pareto distribution is updated upon the arrival of each packet, and a new estimate of the shape parameter computed in accordance with the equation:
ρ n =E (ρ)=(α 0 +n ) b n
where ρ n is the expectation value of the shape parameter after the arrival of the n th packet.
9. A method as claimed in claim 1 , wherein the transmission parameter comprises the bit rate.
10. A method as claimed in claim 1 , wherein the transmission parameter comprises a coding parameter.
11. A real time communication system using a real time transport protocol, comprising:
a receiver for receiving the packets over a network from a remote transmitter and outputting a data stream; and
a network congestion computation module at the receiver for determining network congestion based on incoming packets, said network congestion computation module being configured to perform a statistical analysis of the received packets to determine network congestion and to send a message to the transmitter to control a transmission parameter in accordance with the determined network congestion.
12. A real time communication system as claimed in claim 11 , wherein the network congestion module is configured to perform said statistical analysis is performed using Bayesian statistics.
13. A real time communication system as claimed in claim 11 , wherein the network congestion module is configured to perform the statistical analysis on weighted values of packet inter-arrival times D i .
14. A real time communication system as claimed in claim 13 , wherein the statistical analysis is performed on a parameter J i defined by the expression
J
i
=
D
i
Size
of
packet
(
i
)
.
15. A real time communication system as claimed in claim 14 , wherein the statistical analysis is performed on the basis that the parameter J i follows a Pareto distribution with fixed minimum and shape parameters.
16. A real time communication system as claimed in claim 15 , wherein the network congestion module is configured to compute a shape parameter for the Pareto distribution, compute moving averages for the shape parameter, and send said message when the difference between said moving averages exceed defined bounds.
17. A real time communication system as claimed in claim 16 , wherein when the ith packet arrives (i>n2) the network congestion module is configured to compute the difference Diff as
Diff
i
=
MA
i
(
n
1
)
-
MA
i
(
n
2
)
,
where
(
8
)
MA
i
(
n
)
=
∑
i
-
n
+
1
i
ρ
i
n
for
i
≥
n
(
9
)
18. A real time communication system as claimed in claim 17 , wherein the network congestion module is configured to update the shape parameter for the Pareto distribution upon the arrival of each packet, and a new estimate of the shape parameter computed in accordance with the equation:
ρ n =E (ρ)=(α 0 +n ) b n
where ρ n is the expectation value of the shape parameter after the arrival of the n th packet.
19. A real time communication system as claimed in any one of claims 11 to 18 , wherein the transmission parameter comprises the bit rate.
20. A real time communication system as claimed in any one of claims 11 to 18 , wherein the transmission parameter comprises a coding parameter.