Abstract: A communication terminal (1) on a transmission side comprises means (1-10) for detecting congestion, means (1-7) for detecting a sign of congestion with reference to an ideal congestion window size and a current congestion window size, means for estimating a ratio of priority traffic with respect to traffic within a network, and means (1-8) for dynamically changing a threshold value for detection of the sign of congestion in response to the estimated ratio. When the congestion or the sign of congestion is detected, a congestion window size is changed into the idea congestion window size.

Abstract: In a communication system for transmitting and receiving data between a plurality of terminals via a network through a session relay for relaying communication between the terminals, each of the terminals or the session relay measures a one-way or a round-trip delay time in the network, determines a threshold delay time at which network congestion is judged. With reference to a delay time upon detection of packet loss and the threshold delay time, estimation is made about the possibility that the packet loss results from congestion. A congestion window is changed with reference to the possibility upon detection of the packet loss.

Abstract: A method and apparatus for congestion control of a Transmission Control Protocol (TCP) connection. A TCP source monitors received acknowledgments (ACKs) and uses header information associated with the ACKs to determine an amount of data delivered at a TCP receiver. The amount of delivered data, divided by the time requited to deliver the data, gives a measure of instant data rate. The instant data rate is input to a filtering and averaging process to generate an estimate of the end-to-end bandwidth share or data rate of the connection. For the entire duration of the connection, the TCP source updates its estimate of the end-to-end bandwidth share every time an ACK is received. The estimate of the end-to-end bandwidth share is used the reset a congestion window and a slow start threshold whenever packet loss is indicated such as whenever a timeout expires or whenever three (3) duplicate ACKs are received.

Abstract: A communication terminal (1) on a transmission side comprises means (1-10) for detecting congestion, means (1-7) for detecting a sign of congestion with reference to an ideal congestion window size and a current congestion window size, means for estimating a ratio of priority traffic with respect to traffic within a network, and means (1-6) for dynamically changing a threshold value for detection of the sign of congestion in response to the estimated ratio. When the congestion or the sign of congestion is detected, a congestion window size is changed into the idea congestion window size.

Abstract: A method and apparatus for congestion control of a Transmission Control Protocol (TCP) connection. A TCP source monitors received acknowledgments (ACKs) and uses header information associated with the ACKs to determine an amount of data delivered at a TCP receiver. The amount of delivered data, divided by the time requited to deliver the data, gives a measure of instant data rate. The instant data rate is input to a filtering and averaging process to generate an estimate of the end-to-end bandwidth share or data rate of the connection. For the entire duration of the connection, the TCP source updates its estimate of the end-to-end bandwidth share every time an ACK is received. The estimate of the end-to-end bandwidth share is used the reset a congestion window and a slow start threshold whenever packet loss is indicated such as whenever a timeout expires or whenever three (3) duplicate ACKs are received.

Abstract: A computer performance modeling system is provided. The modeling system includes a first and a second submodel which simulate system operations of a particular system component in minute and simple detail, respectively. A performance analysis mechanism is operatively coupled to the first and the second submodel to execute a simulation of the particular system component using the first and the second submodel. The performance analysis mechanism has a switching device for switching between the first and the second submodel during the simulation at a predetermined point in the execution of the simulation. In an alternative embodiment, the second submodel derives a specific behavior from simulation statistics collected by the first submodel. Also, the performance analysis mechanism executes a simulation of a system model having several of the particular system components by using a plurality of the first and the second submodels. A method for modeling computer performance also is provided.

Abstract: A computer performance modeling system is provided. The modeling system includes a first and a second submodel which simulate system operations of a particular system component in minute and simple detail, respectively. A performance analysis mechanism is operatively coupled to the first and the second submodel to execute a simulation of the particular system component using the first and the second submodel. The performance analysis mechanism has a switching device for switching between the first and the second submodel during the simulation at a predetermined point in the execution of the simulation. In an alternative embodiment, the second submodel derives a specific behavior from simulation statistics collected by the first submodel. Also, the performance analysis mechanism executes a simulation of a system model having several of the particular system components by using a plurality of the first and the second submodels. A method for modeling computer performance also is provided.