Reducing Network Congestion by Preferentially Dropping Packets Sent by High Bandwidth Sources

Patent metadata

Field	Value
Publication number	US-2016269305-A1
Application number	US-201514645091-A
Country	US
Kind code	A1
Filing date	Mar 11, 2015
Priority date	Mar 11, 2015
Publication date	Sep 15, 2016
Grant date	—

Abstract

Some embodiments provide a method for reducing congestion in a network stack that includes a series of components that send data packets through the network stack to a network. At a first component of the network stack, the method receives a data packet from a second component of the network stack. The method identifies a usage indicator value for a flow to which the data packet belongs. The usage indicator value is based on a comparison of a size of the flow to a size of a queue for a third component of the network stack. The method determines whether to send the data packet based on a comparison of the usage indicator value to a threshold usage value. The method sends the data packet to a next component of the network stack only when the usage indicator value is less than the threshold usage value.

First claim

What is claimed is:

1 . A method for reducing congestion in a network stack comprising a series of components that send data packets through the network stack to a network, the method comprising:
at a first component of the network stack, receiving a data packet from a second component of the network stack; identifying a usage indicator value for a flow to which the data packet belongs, the usage indicator value based on a comparison of a size of the flow to a size of a queue for a third component of the network stack; determining whether to send the data packet based on a comparison of the usage indicator value to a threshold usage value; and sending the data packet to a next component of the network stack only when the usage indicator value is less than the threshold usage value.

2 . The method of claim 1 , wherein the threshold usage level is generated randomly for each packet.

3 . The method of claim 1 , wherein the usage indicator value is based on a ratio of unacknowledged data for the flow to which the data packet belongs to the queue size of a physical network interface card (PNIC).

4 . The method of claim 3 , wherein the unacknowledged data for the flow comprises packets sent from the network stack for the flow that have not (i) received acknowledgment receipts or (ii) timed out.

5 . The method of claim 1 , wherein when the usage indicator value is greater than the threshold usage value, the first component drops the data packet.

6 . The method of claim 1 , wherein the first component only performs the determination when the queue of the third component of the network stack is full past a threshold level.

7 . The method of claim 6 , wherein the first component is a virtual network interface card (VNIC), the second component is a TCP/IP stack, and the third component is a physical network interface card (PNIC).

8 . The method of claim 1 , wherein the data packet is a first data packet, the flow is a first flow, the usage indicator value is a first usage indicator value, and the threshold usage value is a first threshold value, the method further comprising:
receiving from the second component of the network stack a second data packet; identifying a second usage indicator value for a second flow to which the second data packet belongs, the second usage indicator based on a comparison of a size of the second flow the size of the queue for the third component of the network stack; determining whether to send the second data packet based on a comparison of the second usage indicator value to a second threshold usage value; and sending the second data packet to the next component of the network stack only when the second usage indicator value is less than the second threshold usage value.

9 . The method of claim 8 , wherein the first threshold usage value and the second threshold usage value are different values.

10 . The method of claim 9 , wherein the first and second flows are the same.

11 . A non-transitory machine readable medium storing a program which when executed by at least one processing unit reduces congestion in a network stack comprising a series of components that send data packets through the network stack to a network, the program comprising sets of instructions for:
at a first component of the network stack, receiving a data packet from a second component of the network stack; identifying a usage indicator value for a flow to which the data packet belongs, the usage indicator based on a comparison of a size of the flow to a size of a queue for a third component of the network stack; determining whether to send the data packet based on a comparison of the usage indicator value to a threshold usage value; and sending the data packet to a next component of the network stack only when the usage indicator value is less than the threshold usage value.

12 . The machine readable medium of claim 11 , wherein the threshold usage level is generated randomly for each packet.

13 . The machine readable medium of claim 11 , wherein the usage indicator value is based on a ratio of unacknowledged data for the flow to which the data packet belongs to the queue size of a physical network interface card (PNIC).

14 . The machine readable medium of claim 13 , wherein the unacknowledged data for the flow comprises packets sent from the network stack for the flow that have not (i) received acknowledgment receipts or (ii) timed out.

15 . The machine readable medium of claim 11 , wherein the program comprises a set of instructions for dropping the packet when the usage indicator value is greater than the threshold usage value.

16 . The machine readable medium of claim 11 , wherein the first component only performs the determination when the queue of the third component of the network stack is full past a threshold level.

17 . The machine readable medium of claim 6 , wherein the first component is a virtual network interface card (VNIC), the second component is a TCP/IP stack, and the third component is a physical network interface card (PNIC).

18 . The machine readable medium of claim 1 , wherein the data packet is a first data packet, the flow is a first flow, the usage indicator value is a first usage indicator value, and the threshold usage value is a first threshold value, the program further comprising sets of instructions for:
receiving from the second component of the network stack a second data packet; identifying a second usage indicator value for a second flow to which the second data packet belongs, the second usage indicator based on a comparison of a size of the second flow the size of the queue for the third component of the network stack; determining whether to send the second data packet based on a comparison of the second usage indicator value to a second threshold usage value; and sending the second data packet to the next component of the network stack only when the second usage indicator value is less than the second threshold usage value.

19 . The machine readable medium of claim 18 , wherein the first threshold usage value and the second threshold usage value are different values.

20 . The machine readable medium of claim 19 , wherein the first and second flows are the same.

Assignees

• Nicira Inc

Inventors

• Sreeramoju Akshay K
• Ravi Madhusudhan
• Koller Ricardo

Classifications

• H04L47/12

  Avoiding congestion; Recovering from congestion · CPC title

• H04L47/32 Primary

  by discarding or delaying data units, e.g. packets or frames · CPC title

• H04L47/11

  Identifying congestion · CPC title

• H04L47/30

  in combination with information about buffer occupancy at either end or at transit nodes · CPC title

• H04L47/2483

  involving identification of individual flows · CPC title