Method for avoiding network congestion and an apparatus thereof
View Patent ↗A method for avoiding a network congestion at a network includes an aggregator and a plurality of workers, wherein the plurality of workers share the same bottleneck link to the aggregator. The method includes counting a number of connections between the aggregator and the worker, and adjusting a window size of each of the workers. The window size is based on the number of connections. The method further includes adding an ACK delay to an ACK, and sending the ACK with the window size to the worker. The ACK delay is based on the window size.
1. A method for avoiding network congestion in a network, the network including an aggregator and a plurality of workers, the plurality of workers sharing a same bottleneck link to the aggregator, the method comprising:
counting a number of connections between the aggregator and the plurality of workers;
adjusting a window size of each of the plurality of workers;
adding an ACK delay to an ACK; and
sending the ACK with the window size to the plurality of workers,
wherein
the ACK delay is based on the window size,
the window size is based on the number of connections, and
the ACK delay is
Δ
=
(
n
Link
capacity
/
MSS
×
advertise
)
-
RTT
where advertise is the window size, MSS is a maximum segment size, RTT is a minimum round trip time, and n is a total number of concurrent connections.
2. The method of claim 1 , wherein the counting the number of connections between the aggregator and the plurality of workers comprises:
initializing the number of connections.
3. The method of claim 1 , further comprising:
increasing the number of connections between the aggregator and the plurality of workers when a connection is created, and
decreasing the number of connections when a connection is closed.
4. The method of claim 1 , wherein the counting the number of connections further includes sending a query request to each of the plurality of workers.
5. The method of claim 1 , wherein
the window sizes of all connections are the same,
the adjusting the window size further includes setting the window size to a maximal value, and
the maximal value of the window size is
advertise
=
max
(
⌊
Link
capacity
/
MSS
×
RTT
n
⌋
,
1
)
.
6. The method of claim 1 , wherein instead of a first one of the ACK delay (Δ), a random value from 0 to (RTT+Δ) that follows uniform distribution is chosen, where, RTT is a minimum round trip time.
7. The method of claim 1 , wherein the sending the ACK with the window size to the plurality of workers is executed after a delay timer expires, if the ACK delay is larger than zero.
8. The method of claim 1 , wherein the sending the ACK with the window size to the plurality of workers is immediately executed without a delay, if the ACK delay is negative.
9. The method of claim 1 , wherein the method is executed in a Transmission Control Protocol (TCP) layer.
10. The method of claim 1 , wherein the method is a rate-based congestion control algorithm.
11. An aggregator for avoiding network congestion in a network, the network including the aggregator and a plurality of workers, the plurality of workers sharing a same bottleneck link to the aggregator, the aggregator comprising:
a processor configured to,
count a number of connections between the aggregator and the plurality of workers,
adjust a window size of each of the plurality of workers, the window size being based on the number of connections, and
add an ACK delay to an ACK, the ACK delay being based on the window size, the ACK delay being
Δ
=
(
n
Link
capacity
/
MSS
×
advertise
)
-
RTT
where advertise is the window size, MSS is a maximum segment size, MT is a minimum round trip time, and n is a total number of concurrent connections; and
a transceiver configured to,
send the ACK with the window size to the plurality of workers.
12. The aggregator of claim 11 , wherein the processor is further configured to
initialize the number of connections.
13. The aggregator of claim 11 , wherein
the number of connections between the aggregator and the plurality of workers is increased when a connection is created, and
the number of connections is decreased when a connection is closed.
14. The aggregator of claim 11 , wherein the transceiver is further configured send a query request to each of the plurality of workers.
15. The aggregator claim 11 , wherein
the window sizes of all connections are the same,
the processor is further configured to set the window size to a maximal value, and
the maximal value of the window size is
advertise
=
max
(
⌊
Link
capacity
/
MSS
×
RTT
n
⌋
,
1
)
.
16. The aggregator of claim 11 , wherein instead of a first one of the ACK delay (Δ), a random value from 0 to (RTT+Δ) that follows uniform distribution is chosen, where, RTT is a minimum round trip time.
17. The aggregator of claim 11 , wherein the transceiver is further configured to
send the ACK with the window size to the plurality of workers, the ACK being executed after a delay timer expires, if the ACK delay is larger than zero.
18. The aggregator of claim 11 , wherein the transceiver is further configured to
send the ACK with the window size to the plurality of workers, the ACK being immediately executed without a delay, if the ACK delay is negative.
19. A network system for avoiding network congestion, the network system comprising the aggregator according to claim 11 and a plurality of workers, the plurality of workers share the same bottleneck link to the aggregator.
20. A non-transitory computer readable storage medium storing computer readable instructions that, when executed by at least one processor, cause the at least one processor to perform a method of claim 1 .