Methodology for improving TCP throughput over lossy communication links

• US 6,975,591 B1
• Filed: 11/22/2000
• Issued: 12/13/2005
• Est. Priority Date: 11/22/2000
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Shorey, Rajeev, Gupta, Vishu, Saran, Huzur, Chrungoo, Abhay
• Primary Examiner(s): Rao, Seema S.
• Assistant Examiner(s): SCHEIBEL, ROBERT C
• Application Number:
  US09/718,679
• Time in Patent Office: 1,847 Days
• Field of Search: 370229-336, 370/389, 709230-237, 709/250, 714/748
• US Class Current: 370/229
• CPC Class Codes: H04L 1/0002 by adapting the transmission rate
• CPC Class Codes: H04L 1/187 Details of sliding window management
• CPC Class Codes: H04L 41/5025 by proactively reacting to service quality change, e.g. by reconfiguration after service quality degradation or upgrade
• CPC Class Codes: H04L 43/0829 Packet loss
• CPC Class Codes: H04L 69/16 Implementation or adaptation of Internet protocol [IP], of transmission control protocol [TCP] or of user datagram protocol [UDP]
• CPC Class Codes: H04L 69/163 In-band adaptation of TCP data exchange; In-band control procedures
• CPC Class Codes: H04W 80/06 Transport layer protocols, e.g. TCP [Transport Control Protocol] over wireless transmission control protocol/Internet protocol [TCP/IP] or user datagram protocol [UDP] H04L69/16

Abstract

A method for improving TCP throughput over lossy communication links without affecting performance over non-lossy links comprises determining lookahead-loss which is the number of lost packets in a given loss-window; using the loss-window and lookahead-loss to detect congestion in the communication links; and controlling transmission under congestion conditions and under normal conditions, wherein the controlling transmission comprises controlling a size of the loss-window by beginning in a slow-start phase; advancing to a congestion avoidance phase when a slow-start threshold is reached; entering a halt growth phase when the first level of packet loss has been recovered; returning to the congestion avoidance phase when a first level of packet recovery occurs; entering a k-recovery phase when a second level of packet loss occurs. The loss window and slow-start threshold are reduced in half and returned to the congestion avoidance phase when the second level of packet loss has been recovered.

Claims

1. A method for improving TCP throughput over lossy communication links without affecting performance over non-lossy links comprising:

determining lookahead-loss which is the number of lost packets in a given loss-window;

using said loss-window and said lookahead loss to detect congestion in said communication links; and

controlling transmission under congestion conditions as well as under normal conditions,

wherein said controlling transmission comprises controlling a size of said loss-window by;

beginning in a slow-start phase;

advancing to a congestion avoidance phase when a slow-start threshold is reached;

entering a halt growth phase when a first level of packet loss occurs;

returning to said congestion avoidance phase when the first level of packet loss has been recovered;

entering a k-recovery phase when a second level of packet loss occurs,

wherein said loss window and said slow-start threshold are reduced in half and returning to said congestion avoidance phase when the second level of packet loss has been recovered.

2. The method as claimed in claim 1, wherein said determining of lookahead-loss is for identifying the number of packets transmitted by a sender in said loss-window for which at least one of the following conditions is true:

said sender has received at least max-dupacks duplicate cumulative acknowledgements, and

said sender has neither received acknowledgement nor selective acknowledgment for said packets, while it has received selective acknowledgements for at least max-dupsacks packets with higher sequence numbers.

3. The method as claimed in claim 1, wherein said detecting of congestion is for identifying when the number of packets lost in said loss-window is greater than an appropriately selected preset number.

4. The method as claimed in claim 1, wherein said controlling transmission is a TCP k-SACK protocol which is a modification of a fast retransmit algorithm of a basic congestion control algorithm of TCP to include entering a ‘

halt growth phase’

whenever said lookahead loss is greater than zero and congestion is not detected, and entering ‘

a k-recovery phase’

whenever the congestion is detected.

5. The method as claimed in claim 4, wherein during said ‘

halt growth phase’

a sender freezes a congestion window and maintains it in that state.

6. The method as claimed in claim 4, wherein said entry into ‘

k-recovery phase’

reduces a congestion window to half its original size, while a slow-start threshold is reduced to half only on a first occasion of entry into the k-recovery phase during a packet loss recovery cycle.

7. A computer program product comprising computer readable program code stored on computer readable storage medium embodied therein for improving TCP throughput over lossy communication links without affecting performance over non-lossy links comprising:

a computer readable program code configured for determining lookahead-loss which is the number of lost packets in a given loss-window;

a computer readable program code configured for using said loss-window and said lookahead loss to detect congestion in said communication links; and

a computer readable program code configured for controlling transmission under congestion conditions as well as under normal conditions,

wherein said computer readable program code configured for controlling transmission comprises a computer readable program code configured for controlling a size of said loss-window by;

beginning in a slow-start phase;

advancing to a congestion avoidance phase when a slow-start threshold is reached;

entering a halt growth phase when a first level of packet loss occurs;

returning to said congestion avoidance phase when the first level of packet loss has been recovered;

entering a k-recovery phase when a second level of packet loss occurs,

wherein said loss window and said slow-start threshold are reduced in half and returning to said congestion avoidance phase when the second level of packet loss has been recovered.

8. The computer program product as claimed in claim 7, wherein said computer readable program code configured for determining of lookahead-loss comprises a computer readable program code configured for identifying the number of packets transmitted by a sender in said loss-window for which at least one of the following conditions is true:

said sender has received at least max-dupacks duplicate cumulative acknowledgements, and

said sender has neither received acknowledgement nor selective acknowledgment for said packets, while it has received selective acknowledgements for at least max-dupsacks packets with higher sequence numbers.

9. The computer program product as claimed in claim 7, wherein said computer readable program code configured for detecting of congestion comprises a computer readable program code configured for identifying when the number of packets lost in said loss-window is greater than an appropriately selected preset number.

10. The computer program product as claimed in claim 7, wherein said computer readable program code configured for controlling transmission is a TCP k-SACK protocol which is a modification of a fast retransmit algorithm of a basic congestion control algorithm of TCP to include entering a ‘

halt growth phase’

whenever said lookahead loss is greater than zero and congestion is not detected, and entering ‘

a k-recovery phase’

whenever the congestion is detected.

11. The computer program product as claimed in claim 10, wherein during said ‘

halt growth phase’

a sender freezes a congestion window and maintains it in that state.

12. The computer program product as claimed in claim 10, wherein said entry into ‘

k-recovery phase’

reduces a congestion window to half its original size, while a slow-start threshold is reduced to half only on a first occasion of entry into the k-recovery phase during a packet loss recovery cycle.