Method and system for controlling TCP traffic with random early detection and window size adjustments
View Patent ↗A method for controlling data traffic with random early detection and window size adjustments including performing random early detection on incoming data packets, calculating a simple moving average of packet dropping probabilities for the data packets as calculated when performing random early detection, decreasing an advertised window size if the simple moving average is greater than a probability target plus a tolerance factor, increasing the advertised window size if the simple moving average is less than the probability target minus a tolerance factor, and not adjusting the window size if the simple moving average is not greater than a probability target plus a tolerance factor and not less than a probability target minus a tolerance factor.
1. A method for controlling data traffic with random early detection and window size adjustments comprising:
performing random early detection on incoming data packets;
calculating a simple moving average of packet dropping probabilities for the data packets;
decreasing an advertised window size if the simple moving average is greater than a probability target plus a tolerance factor;
increasing the advertised window size if the simple moving average is less than the probability target minus a tolerance factor; and
not adjusting the window size if the simple moving average is not greater than a probability target plus a tolerance factor and not less than a probability target minus a tolerance factor.
2. The method of claim 1 , wherein performing random early detection on incoming data packets comprises:
calculating an average aggregate queue size;
calculating an average link capacity;
estimating an average queuing delay using the average aggregate queue size and the average link capacity;
enqueuing an incoming data packet and setting the packet dropping probability for the incoming data packet to 1 if the average queuing delay is less than a delay minimum;
dropping the incoming data packet and setting the packet dropping probability for the incoming data packet to 0 if the average queuing delay is greater than a delay maximum; and
calculating the packet dropping probability for the packet if the average queuing delay is not less than the delay minimum and not greater than the delay maximum and using the packet dropping probability to determine if the packet should be dropped.
3. The method of claim 2 , wherein the average aggregate queue size is calculated according to:
Q ( t )= aQ ( t )+(1 −a ) {circumflex over (Q)} ( t− 1)
wherein Q(t) is the summation of queue size of each flow queried at the beginning of a scheduling period, a is a forgetting factor wherein 0<a≦1, {circumflex over (Q)}(t) is the average aggregate queue size, and t is a time tick of a scheduling clock.
4. The method of claim 2 , wherein the average link capacity is calculated according to:
Ĉ ( t )= bC ( t )+(1 −b ) Ĉ ( t− 1)
wherein C(t) is the instant offered capacity at time t, t is a time tick of a scheduling clock, b is a forgetting factor wherein 0<b≦1, and Ĉ(t) is the average link capacity.
5. The method of claim 2 , wherein the average queuing delay is calculated according to:
{circumflex over (D)} ( t )=α D ( t )+(1−α) {circumflex over (D)} ( t− 1)
wherein D(t)={circumflex over (Q)}(t)Ĉ(t), Ĉ(t) is the average link capacity, {circumflex over (Q)}(t) is the average aggregate queue size, t is a time tick of a scheduling clock, and α is a forgetting factor wherein 0<α≦1.
6. The method of claim 2 , where the packet dropping probability of the incoming data packet when the average queuing delay is not less than the delay minimum and not greater than the delay maximum is calculated according to
P
RED
=
(
P
max
-
P
min
)
·
D
^
(
t
)
-
D_min
D_max
-
D_min
wherein P max is the maximum dropping probability wherein 0≦P max ≦1, P min is the minimum packet dropping probability, D_min is the delay minimum, D_max is the delay maximum, {circumflex over (D)}(t) is the average queuing delay, t is a time tick of a scheduling clock, and P RED is the packet dropping probability.
7. The method of claim 1 , wherein the simple moving average is calculated according to:
P
RED
,
SMA
=
∑
t
0
=
1
K
P
RED
(
t
-
t
0
+
1
)
K
wherein K is the number of samples used for the simple moving average, P RED is the packet dropping probability for an incoming data packet, t is a time tick of a scheduling clock, and P RED,SMA is the simple moving average.
8. The method of claim 1 , wherein the incoming data packets are TCP packets.
9. A system for controlling data traffic with random early detection and window size adjustments comprising:
an IP gateway connected to a gateway;
wherein the IP gateway is configured to receive incoming data packets from packet sources,
perform random early detection on the incoming data packets,
calculate a simple moving average of packet dropping probabilities for the incoming data packets, wherein the packet dropping probabilities are calculated by the IP gateway performing random early detection,
decrease a size of a window advertised to the packet sources if the simple moving average is greater than a probability target plus a tolerance factor,
increase the size of the window advertised to the packet sources if the simple moving average is less than the probability target minus a tolerance factor, and
not adjust the size of the window if the simple moving average is not greater than a probability target plus a tolerance factor and not less than a probability target minus a tolerance factor.
10. The system of claim 9 , wherein the IP gateway is configured to perform random early detection on the incoming data packets by
calculating an average aggregate queue size at the IP gateway;
calculating an average link capacity of the IP gateway;
estimating an average queuing delay using the average aggregate queue size and the average link capacity;
enqueuing an incoming data packet to be transmitted and setting the packet dropping probability for the incoming data packet to 1 if the average queuing delay is less than a delay minimum;
dropping the incoming data packet and setting the packet dropping probability for the incoming data packet to 0 if the average queuing delay is greater than a delay maximum; and
calculating the packet dropping probability for the packet if the average queuing delay is not less than the delay minimum and not greater than the delay maximum and using the packet dropping probability to determine if the packet should be dropped.
11. The system of claim 10 , wherein the average aggregate queue size is calculated according to:
{circumflex over (Q)} ( t )= aQ ( t )+(1 −a ) {circumflex over (Q)} ( t− 1)
wherein Q(t) is the summation of queue size of each flow queried at the beginning of a scheduling period, a is a forgetting factor wherein 0<a≦1, {circumflex over (Q)}(t) is the average aggregate queue size, and t is a time tick of a scheduling clock.
12. The system of claim 10 , wherein the average link capacity is calculated according to:
Ĉ ( t )= bC ( t )+(1 −b ) Ĉ ( t− 1)
wherein C(t) is the instant offered capacity at time t, t is a time tick of a scheduling clock, b is a forgetting factor wherein 0<b≦1, and Ĉ(t) is the average link capacity.
13. The system of claim 10 , wherein the average queuing delay is calculated according to:
{circumflex over (D)} ( t )=α D ( t )+(1−α) {circumflex over (D)} ( t− 1)
wherein D(t)={circumflex over (Q)}(t)Ĉ(t), Ĉ(t) is the average link capacity, {circumflex over (Q)}(t) is the average aggregate queue size, t is a time tick of a scheduling clock, and α is a forgetting factor wherein 0<α≦1.
14. The system of claim 10 , where the packet dropping probability of the incoming data packet when the average queuing delay is not less than the delay minimum and not greater than the delay maximum is calculated according to
P
RED
=
(
P
max
-
P
min
)
·
D
^
(
t
)
-
D_min
D_max
-
D_min
wherein P max is the maximum dropping probability wherein 0≦P max ≦1, P min is the minimum packet dropping probability, D_min is the delay minimum, D_max is the delay maximum, {circumflex over (D)}(t) is the average queuing delay, t is a time tick of a scheduling clock, and P RED is the packet dropping probability.
15. The system of claim 9 , wherein the simple moving average is calculated according to:
P
RED
,
SMA
=
∑
t
0
=
1
K
P
RED
(
t
-
t
0
+
1
)
K
wherein K is the number of samples used for the simple moving average, P RED is the packet dropping probability for an incoming data packet, t is a time tick of a scheduling clock, and P RED,SMA is the simple moving average.
16. The system of claim 9 , wherein the incoming data packets are TCP packets.
17. A system for controlling data traffic with random early detection and window size adjustments comprising:
a terminal;
wherein the terminal is configured to receive incoming data packets from packet sources,
perform random early detection on the incoming data packets,
calculate a simple moving average of packet dropping probabilities for the incoming data packets, wherein the packet dropping probabilities are calculated by an IP gateway performing random early detection,
decrease a size of a window advertised to the packet sources if the simple moving average is greater than a probability target plus a tolerance factor,
increase the size of the window advertised to the packet sources if the simple moving average is less than the probability target minus a tolerance factor, and
not adjust the size of the window if the simple moving average is not greater than a probability target plus a tolerance factor and not less than a probability target minus a tolerance factor.