摘要:
A method and apparatus for controlling the data transfer rates of data sources in an asynchronous transfer mode-based network utilizing intelligent marking to identify those data sources with a data transfer rate exceeding a marking threshold and to direct those thus-identified data sources to reduce their data transfer rates, and to identify those data sources with a data transfer rate lower than a marking threshold and to direct those thus-identified data sources to increase their data transfer rates when the maximum data transfer rate at a communications switch exceeds a first threshold When the maximum data transfer rate at the communications switch exceeds a second thresholds directing all data sources to reduce their data transfer rates. The first and the second thresholds are based on a congestion indicator of available bit rate traffic stored in a buffer in the communications switch. The marking threshold equals a first function of at least a communications switch maximum data transfer rate multiplied by a second function of at least a first congestion indicator.
摘要:
The total implementation complexity of packet schedulers which aim at approximating the Generalized Processor Sharing (GPS) policy is the combination of the complexity of their system-potential function and the complexity involved in sorting the timestamps in order to select the packet with minimum timestamp for transmission. Given that several scheduling algorithms which use a system-potential function of O(1) complexity have been introduced (among them, the Minimum-Delay Self-Clocked Fair Queuing (MD-SCFQ) algorithm achieves optimal delay and excellent fairness properties), the major contribution to the total complexity comes from the task of sorting the timestamps every time a packet is transmitted or received, which is common to all GPS-related schedulers. A method and an apparatus, called the Logarithmic Calendar Queue (LCQ), is disclosed which can achieve a dramatic reduction of the implementation complexity of sorting the timestamps in a GPS-related scheduler, at the cost of very small and controllable degradation of the guaranteed delay bounds. This is achieved by using a plurality of calendar subqueues, each subqueue associated with different granularity for the representation of timestamp values.
摘要:
A scheduler and method for use in packet communication systems apply a generalized discrete-rate scheduling technique which removes the limitation of the linear increase in sorting complexity with the number of supported service rates. The set of supported service rates may be increased without increasing the number of timestamps that need to be sorted. Conversely, the generalized discrete-rate scheduler supports a given number of service rates using a smaller number of rate FIFO queues, thus further reducing complexity. Such improved performance is achieved by splitting, for scheduling purposes only, a connection or session into multiple sub-connections or sub-sessions. The technique can be applied to per-connection-timestamp and no-per-connection-timestamp discrete-rate schedulers, as well as to any other discrete-rate scheduler.
摘要:
Multicasting of an ingress cell received at one of a plurality of input ports is provided through a multistage network to a plurality of output ports. The received ingress cell at a first input port includes a data payload and information identifying one or more routing queues to which a single copy of the data payload is to be outputted. A backpressure queue assignment is identified for the first ingress cell, the backpressure queue assignment identifying a backpressure indicator indicative of the congestion level at at least one routing queue of a subsequent stage. Using the backpressure indicator, it is determined if the first ingress cell is to be outputted from one or more of the routing queues. Thereafter, the first ingress cell is outputted from one or more of the routing queues when the congestion level is below or equal to a first threshold.
摘要:
A method and apparatus for computing a congestion indicator in a communications switch including monitoring at least one first congestion indicator in the communication switch, computing an arrival rate and a first and a second service rate, incrementing a second congestion indicator using the arrival rate, and decrementing the second congestion indicator using the first and second service rates. The method and apparatus are used for deciding to store the data and electronic code that arrives at a communications switch when the second congestion indicator is below a first threshold, and for deciding not to store the data and electronic code that arrives at a communications switch when the second congestion indicator equals or exceeds the first threshold. The method and apparatus are also used for controlling the data transfer rates of data sources in an ATM-based network when both available bit rate and other traffic is present in a communications switch. The method and the apparatus utilize intelligent marking to identify data sources with data transfer rates exceeding a marking threshold and direct those data sources to reduce their data transfer rates when the second congestion indicator exceeds a second threshold. When the second congestion indicator exceeds a third threshold or the first congestion indicator exceeds a fourth threshold, the method and apparatus direct all the data sources to reduce their data transfer rates. The marking threshold equals a function of the maximum data transfer rate multiplied by a function of the second congestion indicator.
摘要:
A method and apparatus for controlling the data transfer rates of data sources in an asynchronous transfer mode-based network utilizes maximum and minimum data transfer rates of sources in the network. A switch instructs data sources within the network to modify their data transfer rates by detecting "potential congestion" and "congested" states.
摘要:
Multicasting of an ingress cell received at one of a plurality of input ports is provided through a multistage switching network to one or more of a plurality of output ports. An ingress cell including a data payload and information identifying output ports of the network is received at one of the input ports of a first stage of the network. A determination is made of the minimum number of first output ports of the first stage of the network to which a single copy of the ingress cell is to be outputted to enable the ingress cell to reach all of the identified output ports. The single copy of the ingress cell is then routed to the minimum number of first output ports of the first stage of the network.
摘要:
A scheduler and method for use in packet communication systems apply a generalized discrete-rate scheduling technique which removes the limitation of the linear increase in sorting complexity with the number of supported service rates. The set of supported service rates may be increased without increasing the number of timestamps that need to be sorted. Conversely, the generalized discrete-rate scheduler supports a given number of service rates using a smaller number of rate FIFO queues, thus further reducing complexity. Such improved performance is achieved by splitting, for scheduling purposes only, a connection or session into multiple sub-connections or sub-sessions. The technique can be applied to per-connection-timestamp and no-per-connection-timestamp discrete-rate schedulers, as well as to any other discrete-rate scheduler.
摘要:
A scheduler and method for use in ATM and packet communication systems applies a no-per-connection-timestamp discrete-rate scheduling technique which does not require the computation and storage of one timestamp per connection, and only maintains a single timestamp per supported service rate. The elimination of the per-connection timestamps has no negative effect on the delay bounds guaranteed by the scheduler. The total implementation cost of such schedulers which approximate the Generalized Processor Sharing (GPS) policy is reduced, since there is less complexity involved in maintaining and sorting the timestamps for all connections.
摘要:
A system is disclosed which services a plurality of queues associated with respective data connections such that the system guarantees data transfer rates and data transfer delays to the data connections. This is achieved by associating each connection having at least one data packet waiting in its associated queue (such a connection called a backlogged connection) with a timestamp generated as a function of system parameters including (a) the number of queues that are backlogged, (b) the data transfer rate guaranteed to each connection, (c) the sum of data transfer rates guaranteed to all backlogged connections, (d) the previous timestamp of the connection, and (e) the weighted sum of the timestamps of all backlogged connections, each timestamp weighted by the data transfer rate guaranteed to the corresponding connection. The backlogged connection associated with the timestamp having the smallest value among all of the backlogged connections is then identified and a data packet is transmitted from the queue corresponding to that connection. A new timestamp is then generated for that connection if it is still backlogged. Once the transmission of the data packet is completed, the foregoing determination of the connection with the minimum timestamp is then repeated to identify the next queue to be serviced.