Multiplex communications – Communication techniques for information carried in plural... – Adaptive
Reexamination Certificate
1998-05-07
2001-10-02
Vu, Huy D. (Department: 2664)
Multiplex communications
Communication techniques for information carried in plural...
Adaptive
C370S468000
Reexamination Certificate
active
06298070
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to the data processing field. More particularly, this invention relates to a method and apparatus for adaptively transmitting data packets in a train.
BACKGROUND
Computer networks that facilitate data processing are becoming increasingly common. Such networks include multiple nodes, which are typically computers, that may be distributed over vast distances and connected by communications links, such as telephone wires. Nodes communicate with each other using packets, which are the basic units of information transfer. A packet contains data surrounded by control and routing information supplied by the various nodes.
Sending, receiving, and processing of packets have an overhead, or associated cost. That is, it takes time to receive a packet at a node, to examine the packet's control information, and to determine the next action. One way to reduce the packet overhead is a method called packet training. This packet-training method consolidates individual packets into a group, called a train, so that a node can process the entire train of packets at once. The word “train” comes from a train of railroad cars. It is less expensive to form a train of railroad cars pulled by a single locomotive than it is to give each railroad car its own locomotive. Analogously, processing a train of packets has less overhead, and thus better performance, than processing each packet individually.
In a typical training method, a node will accumulate packets until the train reaches a fixed target-length. Then the node will process or retransmit the entire thin at once. In order to ensure that the accumulated packets are eventually handled since the packet arrival rate at the node is unpredictable, the method will start a timer wizen the node receives the train's first packet. When the timer expires, the node will end the train and process it even if train has not reached its target length.
This training method works well in times of heavy packet-traffic because the timer never expires. But in times of light packet-traffic, the packets that the node accumulates experience poor performance while waiting in vain for additional packets to arrive, and the ultimate timer expiration introduces additional processing overhead.
Thus, there is a need for a packet-training mechanism that will overcome the disadvantages of the prior art and provide improved performance even in times of a light, variable, or unpredictable packet-traffic rate.
SUMMARY OF THE INVENTION
The invention dynamically adjusts the number of packets sent in a train from a node to reflect the rate-of-packets arriving at a node in a network. In the preferred embodiment, the node has a packet controller that determines the opium train-length, that is the optimum number-of-packets to send in a train. The node also has a timer interval, which is the maximum time-to-wait before sending the next train. The packet controller samples the packet arrival-rate and calculates the elapsed time to receive a configurable-constant number-of-packets in a train. This elapsed time is referred to as a sampling interval. The packet controller only calibrates the optimum train-length when the sampling interval changes significantly from the historic sampling-interval. A significant change is a predetermined percentage greater or less than the historic sampling-interval.
When the timer interval expires (referred to as a timeout), the packet controller sets the optimum train-length to be the number-of-packets accumulated prior to the timeout, which lets the packet arrival-rate determine the number of packets that should be trained. Furthermore, timeouts occurring without a corresponding increase in the optimum train-length cause the packet controller to first lower the optimum train-length by a small amount, and then on back-to-back timeouts without a received packet, the packet controller causes more drastic drops in the optimum train-lengths down to the current number-of-packets accumulated prior to the timeout. This timeout processing rides out small changes in the packet arrival-rate.
REFERENCES:
patent: 4517641 (1985-05-01), Pinheiro
patent: 4736369 (1988-04-01), Barzilai et al.
patent: 4805167 (1989-02-01), Leslie et al.
patent: 4979184 (1990-12-01), Bottoms et al.
patent: 5040176 (1991-08-01), Barzilai et al.
patent: 5063562 (1991-11-01), Barzilai et al.
patent: 5127051 (1992-06-01), Chan et al.
patent: 5166930 (1992-11-01), Braff et al.
patent: 5179660 (1993-01-01), Devany et al.
patent: 5199028 (1993-03-01), Arnold
patent: 5260942 (1993-11-01), Auerbach et al.
patent: 5285442 (1994-02-01), Iwamura et al.
patent: 5343465 (1994-08-01), Khalil
patent: 5400328 (1995-03-01), Burren et al.
patent: 5402416 (1995-03-01), Cieslak et al.
patent: 5537438 (1996-07-01), Mourot et al.
patent: 5566208 (1996-10-01), Balakrishnan
patent: 5606580 (1997-02-01), Mourot et al.
patent: 5859853 (1999-01-01), Carlson
patent: 7066837 (1995-03-01), None
“TCP Vegas: New Tehniques for Congestion Detection and Avoidance”, Lawrence S. Brakmo, Sean W. O'Malley, and Larry L. Peterson, Dept. of Computer Science, Univ.of Arizona, Tucson, AZ 85721, pp. 24-35.
“Delay Analysis For Packet Trains Over Computer Communication Networks”, Ihab A. Jacobs, and Eric B. Hall, Dept. of Electrical Engineering, Southern Methodist University, Dallas, Texas 75275, pp. 13-7 -13-14.
Patent application Ser. No. 08/283,889, Filed Aug. 01, 1994, “Method and Apparatus For Grouping Packets Using Timed Windows”, Brad L. Brech, Gary S. Delp and Albert A. Slane.
Patent application Ser. No. 08/472,368, filed Jun. 07, 1995, “Method and System For Enhanced Communication In A Multisession Packet Based Communication System”, Gary S. Delp and Albert A. Slane.
Patent application Ser. No. 08/670,795, filed Jun. 21, 1996, “Adaptive Packet Training”, David Glenn and Lee Sendelbach.
Article entitled “Efficient Data Transmission in Remote Operation System By Intentional Transmission Delay”, IBM Technical Disclosure Bulletin, vol. 37, No. 08, Aug. 1994.
Carlson David Glenn
Sendelbach Lee Anton
Bussan Matthew J.
Gamon Owen J.
International Business Machines - Corporation
Jones Prenell
Vu Huy D.
LandOfFree
Packet training with an adjustable optimum number of packets does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Packet training with an adjustable optimum number of packets, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Packet training with an adjustable optimum number of packets will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2587350