WO2001099342A2 - Attribution de largeur de bande de reseau numerique et procede et appareil - Google Patents

Attribution de largeur de bande de reseau numerique et procede et appareil Download PDF

Info

Publication number
WO2001099342A2
WO2001099342A2 PCT/US2001/019006 US0119006W WO0199342A2 WO 2001099342 A2 WO2001099342 A2 WO 2001099342A2 US 0119006 W US0119006 W US 0119006W WO 0199342 A2 WO0199342 A2 WO 0199342A2
Authority
WO
WIPO (PCT)
Prior art keywords
flow
flows
network link
cell
data
Prior art date
Application number
PCT/US2001/019006
Other languages
English (en)
Other versions
WO2001099342A3 (fr
Inventor
David F. Barton
Kurt J. Worrell
Original Assignee
Vitesse Semiconductor Corporation
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Vitesse Semiconductor Corporation filed Critical Vitesse Semiconductor Corporation
Priority to AU2001268384A priority Critical patent/AU2001268384A1/en
Publication of WO2001099342A2 publication Critical patent/WO2001099342A2/fr
Publication of WO2001099342A3 publication Critical patent/WO2001099342A3/fr

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04Q—SELECTING
    • H04Q11/00—Selecting arrangements for multiplex systems
    • H04Q11/04—Selecting arrangements for multiplex systems for time-division multiplexing
    • H04Q11/0428—Integrated services digital network, i.e. systems for transmission of different types of digitised signals, e.g. speech, data, telecentral, television signals
    • H04Q11/0478—Provisions for broadband connections
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L12/00—Data switching networks
    • H04L12/54—Store-and-forward switching systems 
    • H04L12/56—Packet switching systems
    • H04L12/5601—Transfer mode dependent, e.g. ATM
    • H04L2012/5629—Admission control
    • H04L2012/5631—Resource management and allocation
    • H04L2012/5632—Bandwidth allocation
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L12/00—Data switching networks
    • H04L12/54—Store-and-forward switching systems 
    • H04L12/56—Packet switching systems
    • H04L12/5601—Transfer mode dependent, e.g. ATM
    • H04L2012/5638—Services, e.g. multimedia, GOS, QOS
    • H04L2012/5646—Cell characteristics, e.g. loss, delay, jitter, sequence integrity
    • H04L2012/5651—Priority, marking, classes

Definitions

  • the present disclosure relates generally to digital networks and, more particularly, to a method and system for providing advanced levels of service in a digital network while optmizing bandwidth utilization.
  • a number of approaches have been used in the prior art to cope with traffic on digital networks.
  • Early digital network technologies did not suffer from their lack of sophistication in terms of bandwidth allocation since relatively little data required movement between interconnected components.
  • a network switch in such an early system might have employed a round robin allocation in which incoming data flows, requesting outgoing transfer via a particular network link, alternately received grants to use the particular network link.
  • Even though such a simple allocation approach necessarily wastes a portion of the available network bandwidth, the approach was simple to implement while the waste was of little or no consequence in view of the surplus of available bandwidth.
  • ATM Asynchronous Transfer Mode
  • U.S. patent number 5,515,363 issued to Ben-Nun, et al discloses improvements which address traffic shaping to improve quality of service for virtual circuits
  • ATM transfers packets containing data across the network incrementally in the form of cells. Packets vary in length depending upon the amount of data carried. For transfer, each packet is broken up into a series of cells.
  • One weakness in ATM network technology resides in a reliance upon fixed cell length. Because there is no relationship between packet length and cell length, a particular packet will rarely divide in a way which completely fills the last cell of the plurality of cells. For this reason, some space in the last cell of the plurality is often wasted, since even the last cell must be of the required fixed length. It is recognized herein that reliance on fixed cell length results in wasted bandwidth which is now valuable.
  • ATM Another weakness in ATM technology is submitted to relate to data flow bandwidth allocation.
  • flows are segregated into two general categories.
  • the first category of flows is guaranteed bandwidth on the network, while the second category of flows is relegated to any remaining bandwidth, which is termed as "leftover" bandwidth.
  • This scheme is problematic since flexibility in dealing with a flow, after assignment to one of the categories, may be limited. For example, it is difficult to allocate at least some of the leftover bandwidth to a flow in the guaranteed bandwidth category.
  • flows being transferred in a particular system may simply be a mismatch as to these two narrowly defined categories.
  • ATM response to a flow which is guaranteed some predetermined bandwidth and which fails to use its guaranteed bandwidth wastes considerable bandwidth.
  • start-up time (start-up time). Following this initial configuration, changing the transfer rate on a network link: requires ceasing all transfers over the network link, even those not involving a particular device of concern; reconfiguring the particular device with a different physical transfer rate; and then restarting transfers on the network link.
  • a network arrangement is configured for transferring a plurality of data flows using a network link, each of which flows requires a minimum bandwidth allocation on the network link.
  • the data flows are transferred by initially receiving the data flows at a first end of the network link.
  • the data flows are then divided into cells such that at least some of the cells vary in length with respect to one another. Transfer of the cells on the network link is accomplished in a predetermined way such that cells of different data flows are multiplexed and the minimum bandwidth allocation of each data flow is maintained on the network link.
  • the cells are received, including the cells which vary in length, and combined so as to reconstruct the plurality of data flows at the second end of the link.
  • the data flows are divided into cells such that the length of each cell does not exceed a maximum cell length and at least a portion of the cells include a length that is less than the maximum cell length.
  • each data flow accumulates priority credits and is debited an amount of credit corresponding only to the length of a particular cell that is transferred.
  • a plurality of different types of priority credits are simultaneously accumulated by at least one of the data flows such that priority is obtained based on having a predetermined combination of the plurality of different types of credits as to type of credits and/or amount of credits.
  • a network arrangement is configured for transferring packets originating from a plurality of data flows using a network link.
  • the packets are transferred by initially receiving the packets at a first end of the network link. For at least specific ones of the packets, each of which includes a length that is greater than a maximum cell length, the specific packets are divided into a series of cells such that each of the cells is identifiable as forming a part of that specific packet. At least the specific ones of the packets are transferred on the network link incrementally by alternating each series of cells for the respective packets. At a second end of the network link, the cells, as transferred from different packets, are received and the specific packets from which each cell originated are identified. Thereafter, that identification information is used to reconstruct the series of cells for each specific packet to produce the specific packets at the second end of the network link.
  • a network arrangement is configured for transferring a plurality of data flows using a network link, each data flow requires a priority allocation of at least a minimum bandwidth on the network link.
  • the data flows are transferred using the network link by initially receiving the data flows at a first end of the network link.
  • Cells are formed using the data flows such that each data flow is made " up of a series of one or more cells. Transfer of the data flows is initiated on the network link by transferring one or more cells on the network link which form portions of the flows consistent with the established priority allocation.
  • certain ones of the cells are transmitted consistent with the established priority allocation for a particular flow along with at least one cell transmitted irrespective of the established priority allocation for the particular flow such that the certain cells include a cell configuration and the one cell transmitted irrespective of the established priority allocation for the particular flow includes a configuration that is identical to the cell configuration of the certain cells.
  • a network arrangement is configured for using a network link having an available network link bandwidth which potentially varies over time and which is configured for incrementally transferring a plurality of data flows, for at least specific ones of which preferred minimum bandwidth allocations are specified.
  • the data flows are transferred by receiving the data flows at a first end of the network link for transfer thereon to a second end of the network link.
  • sustained percentages of the available network link bandwidth are established to be assigned, respectively, to each of the specific data flows for use during incremental transfer of the specific data flows on the network link.
  • an arrangement and method are provided for transferring the data flows on the network link.
  • a first configuration receives at least one burst as part of one of the flows at a first end of the network link with a particular assigned priority.
  • a second configuration then transfers the burst to the second end of the network link with a new priority that is higher than the particular assigned priority.
  • the new priority is based on the use of at least two different types of priority allocation credits
  • FIGURE 1 is a diagrammatic representation of a network switch manufactured in accordance with the present invention.
  • FIGURE 2 is a more detailed illustration of the network switch of Figure 1 showing further details of its structure.
  • FIGURE 3 is a block diagram showing details of a single output port of the network switch of Figures 1 and 2 illustrating details of the structure of the output port.
  • FIGURE 3a is a diagrammatic representation of a data cell formed in accordance in the present invention and configured for transfer in accordance with the present invention.
  • FIGURE 4 is a table illustrating the flow configuration parameter implementation of the present invention which- is used to fix flow parameter values in accordance with the sustained/burst priority concepts of the present invention.
  • FIGURE 5 is a diagrammatic representation, in block diagram form, showing one possible implementation of a bandwidth allocation state machine in accordance with the present invention for one of the flow paths forming a portion of the output port shown in Figure 3.
  • FIGURE 6 is a flow diagram illustrating the process carried forth in the bandwidth allocation state machine of Figure 5, illustrating one implementation of the sustained/burst priority concepts of the present invention.
  • FIGURE 7 is a block diagram illustrating an active packet priority (APP) structure for one of the flows of Figure 3 and the manner in which the structure produces APP information.
  • APP active packet priority
  • FIGURE 8 is a flow diagram illustrating one possible method by which the APP state determined by the APP structure of FIGURE 7 is calculated in accordance with the present invention.
  • FIGURE 9 is a block diagram illustrating one implementation of a bandwidth allocation arbiter produced in accordance with the present invention for use in the overall output port of Figure 3.
  • FIGURE 10 is a flow diagram illustrating one technique for updating the current selected cell register forming part of Figure 9 in accordance with the priority allocation concepts of the present invention.
  • FIGURE 11 is a flow diagram illustrating one process for updating the round robin arrangement show in Figure 9 in accordance with the present invention.
  • FIGURE 12 is a block diagram illustrating a priority encode comparator suitable for use in the bandwidth allocation arbiter of Figure 9 in accordance with the present invention.
  • FIGURE 13 is a diagrammatic representation of a first alternative encoding for the candidate and current cell registers of Figure 12 which promotes active packet priority above other priority types.
  • FIGURE 14 is a diagrammatic representation of a second alternative encoding for the candidate and current cell registers of Figure 12 which is implemented on a flow basis.
  • FIGURE 15 is a data flow diagram showing ingress data, queued data and egress data for three flows transmitted in accordance with the sustained priority concept of the present invention.
  • FIGURE 16 is a table illustrating the development of the data flow show in Figure 15, further including credit counts and pointer indications for the flows.
  • FIGURE 17 is another data flow diagram showing ingress data queued data and egress data for three flows and further illustrating the manner in which available unused bandwidth is utilized in accordance with the present invention.
  • FIGURE 18 is a table illustrating the development of the data flow shown in Figure 17 including credit counts and pointer indications for the flows.
  • FIGURE 19 is a data flow diagram shown here to illustrate the response of the present invention in a reduced network bandwidth situation.
  • FIGURE 20 is a data flow diagram shown here as part of a sustained credit counting example to illustrate the operation of the present invention in the instance of transmission of variable length multi-word cells.
  • FIGURE 21 is a table illustrating the development of the data flow shown in Figure 20 including credit counts and pointer indications for the flows.
  • FIGURE 22 is a plot of per flow sustained credit over time for each of the flows transferred in Figures 20 and 21, shown here to illustrated in detail changes in the credit counts for each flow in accordance with the present invention.
  • FIGURE 23 is a data flow diagram shown here as part of a burst fraction example to illustrate the operation of the present invention in the instance of transmission of data with a bursty flow and using multiple types of priority credits.
  • FIGURE 24 is a table illustrating the development of the data flow shown in Figure 23 including credit counts and pointer indications for the flows.
  • FIGURE 25 is a data flow diagram illustrative of a case which does not utilize the burst fraction, shown here for purposes of comparison of this result with the result of Figure 24 when one of the flows exhibits bursty ingress data.
  • FIGURE 26 is a table illustrating the development of the data flow shown in Figure 25 including credit counts and pointer indications for the flows.
  • FIGURE 27 is a data flow diagram shown here as part of an active packet priority (APP) example to illustrate the operation of the present invention in the instance of transmission of data using the APP concept of the present invention.
  • APP active packet priority
  • FIGURE 28 is a data flow diagram illustrative of a case which does not utilize active packet priority, shown here for purposes of comparison of this result with the result of Figure 24 to illustrate the advantages of APP.
  • FIGURE 29 is a table illustrating the development of the data flow shown in Figure 28, shown here to further illustrate priority allocation using the credit priority concept of the present invention.
  • Figure 1 illustrates a portion of a computer network manufactured in accordance with the present invention and generally indicated by the reference numeral 10.
  • Network 10 includes an eight port full duplex switch 12. The latter is bi-directional and is interconnected with a number of other network devices, as will be described immediately hereinafter.
  • the eight ports of network switch 12 are indicated as P 0 -P 7 .
  • Network devices connected to the ports include an a first end station 14 on P 0 , a first router 16 on Pi, a second end station 18 on P 2 , a second router 20 on P 3 , a first bridge 22 on P 4 , a third end station 24 on P 5 , a fourth end station 26 on P 6 and a second bridge 28 on P 7 .
  • the first through fourth end stations may comprise, for example, workstations or personal computers.
  • the end stations are leaf nodes of the network while the routers or bridges are internal nodes of a graph and are connected to other network devices using other links that are only partially illustrated.
  • Network switch 12 interconnects each of the eight attached devices so that any one device may communicate through the switch to any of the other devices, including a device communicating with itself through the switch.
  • each port includes an input port, PI0-PI7 and an output port PO 0 -PO 7 .
  • the connections between the ports of the network switch and the connected devices may be referred to as network links or, more simply, links.
  • the link between port 4 and first bridge 22 may be referred to as link 4.
  • Each input port is connectable to all eight output ports via an Input-to- Output port connection network 30. It should be appreciated that there are a variety of methods known to those of ordinary skill in the art for selectively connecting the input and output ports. All known techniques as well as any appropriate techniques yet to be developed are considered as being useful herein so long as each input port is selectively connectable to each output port.
  • FIG. 3 details of one output port of network switch 12 are shown.
  • PO ? is generally indicated and is shown diagrammatically.
  • Eight port inputs P 0 -P 7 are received from port connection network 30 ( Figure 2) and are separated into any flows, Fj to F N that are present.
  • flows are transferred in the form of packets each of which is made up of a series of cells.
  • the present invention contemplates the use of variable length cells in contrast to prior art network technologies such as, for example, ATM which relies on fixed length cells.
  • Cell 31 includes a header 32 and a data pay load 33.
  • Header 32 is made up of packet information 34, routing data 35, cell length data 36, a flow number 37, output link property data 38 and a channel ID 39. The latter is typically used to identify cells originating from a common packet in the reconstruction of the packets.
  • data 33 is the object of transfer on the network.
  • the specific amount of data contained in data payload 33 is specified by cell data length 33, in terms of octets.
  • ATM does not include cell size in its cell headers. Such information is not relevant in ATM since the cell size is fixed.
  • the ATM cell size is known, due to its fixed length, throughout an ATM network including at all input and output ports.
  • the present invention contemplates that adaptability of existing ATM networks, in order to utilize the advantages disclosed herein associated with variable length cells, may be limited since reliance on fixed cell size is seen in at least some ATM hardware designs. That is, the fixed ATM cell size may simplify certain aspects of hardware. Other aspects of cell 31 of the present invention will be described at appropriate points below.
  • flows F 0 -F 7 are received by output port P0 7 following pre-processing performed, for example, by each flow's originating input port.
  • the respective input port separates the header of each cell, containing control and routing information, from the data of the cell which need not be processed, but which data must be forwarded through the network switch.
  • a cell header reception circuit 40 receives the cell headers for each flow while a cell data storage circuit 42 receives the cell data from each of the flows. It is noted that the cell size, forming part of the header information, is used by cell header reception circuit 40.
  • Cell data storage circuit 42 may comprise, for example, commodity dynamic rams or any suitable storage configuration.
  • the switching configuration for physically routing the cells between ports and for separating cell header information from cell data may take on a variety of configurations.
  • Incoming data is stored by cell data storage section 42 for selective retrieval at appropriate times.
  • Techniques for identifying the correct cell data, based on header information, are known in the art and readily adaptable to the present application by one having ordinary skill in the art.
  • the cell header data arriving on one of the flow inputs to cell header reception circuit 40 is analyzed by the latter in a way which identifies the appropriate flow for each cell.
  • Each output port including exemplary port PO 7 , includes a flow cell header queue arrangement 50 (indicated by a dashed line) having a flow cell header queue corresponding to each flow.
  • a flow cell header queue arrangement 50 indicated by a dashed line
  • flow cell header queue arrangement 50 having a flow cell header queue corresponding to each flow.
  • flow 0 cell header queue 60 the flow cell header queue corresponding to each flow.
  • flow 1 cell header queue 61 the flow cell header queue corresponding to each flow.
  • flow N-l While the latter is specified as being associated with flow N-l, it should be appreciated that the highest order designated flow in the present example is flow 7. The use of flow N-l emphasizes that the present invention is not restricted to any specified number of flows.
  • a flow configuration parameter section stores priority allocation information associated with the various flows.
  • Information provided initially by an external agent (not shown) is used in determining specific values for the parameters to be described.
  • the specific parameters associated with each flow are illustrated. For purposes of brevity, only those parameters associated with flow 0 will be described, since all other flows include parameters having the same designations. Accordingly, F 0 has an F 0 bandwidth allocation (hereinafter BWA) sustained fraction 72, an F 0 BWA sustained limit 74, an F 0 BWA burst fraction 76 and an F 0 BWA burst limit 78.
  • BWA bandwidth allocation
  • the BWA parameters for each flow are used by associated BWA state machines that are indicated by the reference numbers 80, 82 and 86 for flows F 0 , F] and F N _ J , respectively.
  • Each of the BWA state machines stores a BWA state value.
  • Further associated with each flow are Active Packet Priority (hereinafter APP) computation elements. Specifically, each flow has an associated APP section, indicated as APP 0 , APPi and APP N ., respectively.
  • APP Active Packet Priority
  • the BWA parameters in flow configuration parameter section 70 may be determined with the objective of maintaining flow control, for instance, based on any number of considerations including, but not limited to: 1. Ingress data rates for traffic destined to be transmitted on a particular output port/link (based upon the setting of sustained fraction 72).2. The burst characteristics of a flow; i.e., is the flow a steady flow of N megabytes per second, uniformly distributed over time or is the flow bursty, needing low bandwidth over certain intervals and needing significant bandwidth over other intervals (based on the setting of burst fraction 76). 3. The amount of credit that can be accumulated by a flow without adversely affecting the other flows present on the same output link.
  • the APP sections are interposed between their particular flow queues and a bandwidth allocation arbiter 90. These elements cooperate in a highly advantageous way for allocating network link bandwidth to the various flows, as will be described.
  • lines 92, 94 and 98 extend from the flow queue headers through the APP components of each flow to the bandwidth allocation arbiter. These lines are configured for transferring priority control information as well as the actual cell headers, once selected for transmission.
  • data passed from the flow cell header queues is augmented with APP information in route to arbiter 90.
  • the remaining structure in the figure includes a transmitter 100 interconnected to the bandwidth allocation arbiter via a transmit header queue 110.
  • a cell is selected for transmission by bandwidth allocation arbiter 90, its corresponding cell header is placed or "released" into transmit header queue 110.
  • Cell headers line up in the transmit header queue awaiting transfer by transmitter 100 on a physical network link 120. It is noted that the latter may be connected at an opposing end (not shown) to another network switch or, for example, to another network switch or, for instance, to an station or bridge (see Figure 1). If network link 120 is connected at its opposing end to another network switch, the network link would simply be connected to an input port, for example, such as one of the input ports shown in Figure 2.
  • one network link may carry a plurality of flows, each of which may be directed to a different destination. That is, each flow in Figure 3 does not represent a network link. Rather, the flows may be derived from any of the input ports shown in Figures 2 and 3, but all are directed to link 120, in this example, and thus P0 7 , as represented in Figure 3.
  • Transmitter 100 is configured for sequentially transmitting cells when their associated cell headers are received, in turn, from the transmit header queue since the priority allocation method of the present invention has, at this point, been accomplished by preceding elements.
  • Transmitter 100 includes a cell data fetcher section 122 which receives cell headers and then retrieves associated data from cell data storage arrangement 42 based on a known relationship between each cell header and the corresponding stored data.
  • a cell is chosen from a particular flow (e.g., flow 1) the cell header is dequeued from the flow queue (e.g., Flow 1 cell header queue 61) and is placed on the transmit header queue 110 for subsequent data retrieval and transmission.
  • Select line 124 serves two purposes.
  • bandwidth allocation arbiter 90 provides an indication to the bandwidth allocation arbiter from the transmitter whenever a cell is transmitted.
  • the select line is comprised of a combination of signals further including the flow ID of a transmitted cell and the priority with which a cell was transmitted. Accordingly, bandwidth allocation arbiter 90 then updates the BWA State for each flow. At times when output network physical link 120 is available, cells are sequentially transmitted as per the physical link protocol in first-in-first-out order from transmit header queue 110.
  • the cell header is retrieved from the transmit header queue 110 by cell header fetcher 122 and the fetcher locates the data associated with the cell, as well as any output physical link parameters stored with the data.
  • output physical link parameters which may be stored with the data include but are not limited to physical link header information, physical link error detection/correction information or physical link timing information. If the input physical link and output physical link are identical physically, the input link physical data may be applied without change to the output physical link. Alternatively, physical network information may be modified or created by the network switch itself. Irrespective of the specific nature of data such as physical network information, the objective is to ensure that all information necessary to accomplish transmission of a particular cell is to hand using the combination of a cell header queue with referring to cell data storage section 42. The recombined cell is then transferred on network link 120.
  • Each flow cell header queue promotes a candidate cell to bandwidth allocation arbiter 90.
  • the candidate cell (not shown), represented by its corresponding cell header within each cell header queue, is first in line in the flow cell header queue.
  • the values present in flow configuration parameter section 70 are used by the BWA State components in determining the transmission order of outgoing cells in attempting to meet desired outgoing bandwidth guarantees. As will be described below, the BWA sustained and burst bandwidth fractions do not actually specify a guaranteed bandwidth on the outgoing network link for certain reasons.
  • sustained and burst fractions represent a portion or percentage of outgoing available bandwidth present on the outgoing network link.
  • the use of such parameters as representing a percentage of available bandwidth is considered as being highly advantageous and has not been seen heretofore by Applicants.
  • this percentage allocation approach is so advantageous resides in the fact that some systems allow devices connected to a network link to control the clock rate used on the Jink.
  • a downstream device is able to conform the network link to its own capabilities.
  • the downstream device can instantaneously and selectively change the clock rate either up or down depending upon its own bandwidth requirements.
  • the down stream device may even set the clock rate to zero which halts the link and then restart the link at will.
  • the present invention preserves the priority state of each flow since priority credits are neither issued nor debited in the absence of link transmissions.
  • bandwidth allocation arbiter 90 utilizes flow configuration arrangement 70, the status of output link 120, the BWA State value and APP information for each flow in allocating priority for the use of the outgoing network link bandwidth. Further details regarding use of the APP information will be provided at an appropriate point below.
  • F 0 the values relevant to F 0 are BWA F 0 sustained fraction 72, BWA F 0 sustained limit 74, BWA F 0 burst fraction 76 and BWA burst limit 78. While the manner in which these values are used in the overall method of the present invention will be described in detail hereinafter, it is noted for the moment that BWA F 0 sustained fraction 72 limits allocation of available network link bandwidth in an average sense over time.
  • the BWA F 0 burst fraction is used in the instance of flows exhibiting "bursty" behavior to limit the percentage of available network bandwidth usurped by a flow during a particular burst.
  • flow configuration 70 is generally semi-static as supplied by the external agent, however, the flow parameters can be changed by the external agent at any time to provide new desired minimum bandwidth guarantees once the configuration change is complete. It is noted that the present invention implements its priority allocation scheme using credits accumulated by the flows and, thereafter, decremented in a highly advantageous way. In this regard, two different types of credits are allocated to each flow: sustained credits are allocated relative to the
  • BWA sustained fraction and sustained limit parameters while burst credits are allocated relative to the BWA burst fraction and burst fraction parameters.
  • BWA F 0 sustained limit 74 limits the number of sustained word credits that can be accumulated by the flow while BWA F 0 burst limit 78 specifies the maximum number of burst word credits that can be accumulated by the flow. Either of the sustained or burst limits may be fixed values. While the allocation of two different types of credits to each flow is considered as being highly advantageous in and by itself, more detailed features of the present invention, concerning allocation and use of these different types of credits, are considered to provide still further advantages, as will be described.
  • FIG. 5 a diagram of a BWA flow state machine is illustrated. Consistent with previous examples, the present example illustrates F 0 BWA state machine 80. All other BWA state machines are identical with the exception of storing parameter sets for different flows. Moreover, the BWA state machines for all of the flows operate in parallel. The flow 0 parameter configuration values are taken from flow configuration 70 ( Figure 4) and used only in this state machine. BWA F 0 sustained fraction 72 and sustained limit 74 are used as inputs to an ALU S c while BWA F 0 burst fraction 76 and burst limit 78 are used as inputs to an ALU BC - The use of these parameters will be described more fully below.
  • a sustained priority register 130 stores a concatenated value in which SC (Sustained Credit) 132 indicates a number of maximum sized cell credits that have been accumulated for sustained transfers. It is noted that the term "sustained transfer" contemplates a transfer that could be continuous over a substantially long period of time.
  • SC Word 134 indicates a fraction of one maximum sized cell credit, in terms of 16 bit words, accumulated for sustained transfers. SC 132 and SC Word 134 cooperatively indicate a total number of sustained word credits that are available to F 0 .
  • SC fraction 136 indicates a fractional part of a single word (a 16 bit word is used in the present example, but other word sizes may just as readily be employed) that has been accumulated for sustained transfers.
  • SC 132, SC Word 134, and SC Fraction 136 concatenated in sustained priority register 130, cooperatively indicate a number of sustained fractional word credits that have been accumulated.
  • the register tracks not only a fraction of the maximum cell length, but a fraction of a single word from which the cells are made up.
  • SC fraction 136 is made up of 7 bits and is, therefore, capable of indicating fractional word amounts in units of 1/128* of a word. The significance of this capability will be made evident in continuing discussions. For the moment, however, it is noted that a fraction of 1 (128/128) is not permitted since this is equivalent to allocating the entire link to one flow.
  • a burst priority register 140 stores a concatenated value in which BC (Burst Credit) 142 indicates a number of maximum sized cell credits that have been accumulated for burst transfers.
  • BC Word 144 indicates a number of fractional maximum sized cell burst credits that have been accumulated.
  • BC 142 and BC word 144 together indicate a total number of burst word credits accumulated for use by F 0 , noting that the units in BC 142 are maximum sized cells while the units in BC word 144 are words.
  • BC fraction 146 indicates a fractional part of a single word that has been accumulated for burst transfers.
  • BC 142, BC Word 144 and BC Fraction 146 together indicate a total number of burst fractional word credits that are available to F 0 .
  • BC fraction 146 indicates a fraction of a word at a resolution of 1/128 th of a word.
  • ALU S c performs all computations associated with sustained transfer credits while ALU B c performs all arithmetic computations associated with burst transfer credits. Additional information is provided to these ALU's for use in the subject computations, as will be described at appropriate points below.
  • a sustained priority threshold value is stored in an SC threshold register 152 for comparison with the combined value of SC 132 and SC word 134. The result of this comparison determines whether flow 0 has sustained priority.
  • the sustained priority determination is used internal to the Fo state machine as well as being supplied to other components on a line 154.
  • the threshold sustained priority value in SC compare register 152 is typically the cell size, in words, for the candidate cell of F 0 that is being considered for transmission. It should be emphasized that this comparison establishes that F 0 has sufficient sustained credits based on the actual length of its candidate cell in words.
  • one optimization for expediting computation is to verify that there are at least sufficient sustained credits to transfer a maximum sized cell, if such a maximum exists. That is, the present invention contemplates an implementation having no upper limit on cell size. In view of these alternatives, it is considered that a wide range of optimizations may be produced by one having ordinary skill in the art. In this regard, one of ordinary skill in the art may utilize any number of the bits present in sustained priority comparator 150 or in the burst priority comparator to be described. Normally, some selection of the upper order bits is used in the comparison. By adding a single lower order bit to the comparison, accuracy in the comparison is increased by a factor of one-half.
  • a burst priority threshold value is stored in a BC threshold register 160.
  • the burst priority threshold value is compared with the combined value of BC 142 and BC word 144.
  • the burst priority threshold may, for example, be the size in words of the F 0 candidate cell or the size in words of a maximum size cell, again, if such a maximum exists.
  • this burst priority determination is output on a line 164 for use internally by the state machine and is also supplied to other device components.
  • burst priority comparison can be modified in any number of different ways using various combinations of the bits present in burst priority register 140 such as, for example, in optimizations intended to increase the accuracy of the comparison by adding more lower order bits to the comparison.
  • Each state machine further uses indicators including: a "word transmitted flow k" received on a line 166 (derived from select line 124 in Figure 3), which indicates that a word has been transmitted in association with any of these flows and wherein the flow identification number is given by the letter "k"; word transmitted had SP (sustained priority) 168 (also derived from select line 124); and transmitted word event on a line 170, which indicates that a word is transmitted from any flow (also derived from select line 324).
  • Burst ALU B c additionally uses a Flow Queue k Not Empty 172 indication in consideration of adjustment of the value in BC register 140, also to be described at an appropriate point.
  • step 201 an indication arrives that a word has been transmitted for any flow via line 170 in Figure 5. In other words, the steps shown in Figure 6 are only performed responsive to transmission of a word.
  • step 204 determines whether the word was for the flow associated with the particular BWA component and the whether sustained priority was associated with that transmitted word.
  • step 208 is performed to debit this flow's credits amounting to the cost of transmitting a word of data.
  • the credits in sustained priority register 130 and burst priority register 140 are each debited by one word of credit. These values are not permitted to go negative. If the transmitted word was not from this flow on step 206 and/or did not have sustained priority, step 208 is not performed. In this regard, it should be appreciated that this process is performed on each clock cycle and that one word of data is transmitted on each clock cycle. It is for this reason that the flow is debited only for one word; the amount of bandwidth that was actually used.
  • a maximum size cell is made up of 64 words.
  • step 209 it is determined if any additional cells are present in the flow queue for this flow (i.e., queue 80 in the example of F 0 ).
  • step 209 is performed immediately after step 206 if the transmitted word was from a different flow. If there is at least one pending cell in the F 0 flow queue, step 210 adds BWA burst fraction 76 to the number of available credits in burst credit register 140. Step 210 is performed irrespective of the originating flow of the transmitted word. On the other hand, if the F 0 flow queue is empty, F 0 is not allocated additional burst credits and the method proceeds immediately to step 212.
  • F 0 is credited by an amount equal to BWA sustained fraction 72 in sustained credit register 130.
  • receiving the sustained credit allocation is based solely on the fact that a word was transmitted from any flow, as detected in step 202. That is, each flow receives the sustained credit fraction allocation, even the flow which transferred the word that initiated step 201.
  • the sustained and burst fractions may be different for each flow, as is readily specifiable per the separate entries available for each flow in Figure 4. On this basis, flows may accumulate either sustained or burst credits at very different rates.
  • either the burst or sustained credit fraction may be specified in the range from 1/128 of a word up to 127/128 of a word, creating significant control in the rate at which credits accumulate (i.e., in increments of approximately 0.8 %), as will be described in further detail. It should be appreciated that additional bits can be added and relied on to further reduce the size of the smallest available fractional increment of a word, thereby allowing still further accuracy in specifying the allocation fractions disclosed herein.
  • step 214 it is determined whether the burst temporary value is less than F 0 burst limit 78, shown in Figure 4. If not, the burst temporary value is set to the burst limit and the latter is stored in burst credit register 140 in step 216. If the burst temporary value is less than F 0 burst limit 78, the burst temporary value is stored in burst credit register 140 as the number of sustained credits available to F 0 in step 218. That is, the prior number of burst credits is incremented by F 0 BWA burst fraction 76. At step 220, it is determined whether the sustained temporary value is less than F 0 BWA sustained limit 74, shown in Figure 4.
  • the sustained temporary value is set to the sustained limit and the latter is stored in sustained credit register 130 by step 222. If the sustained temporary value is less than F 0 sustained limit 74, the sustained temporary value is stored by step 224 in sustained credit register 130 as the number of sustained credits available to F 0 . With step 226, monitoring is returned to step 201.
  • the foregoing discussion utilizes temporary values at least in part for purposes of clarity. A particular embodiment may or may not rely on the use of such temporary values, as desired. In view of the foregoing, it is worth emphasizing that a flow is debited by a word's worth of credit for the word that is transferred from that flow on a particular clock cycle.
  • the effect of the sustained credit allocation is to permit an over time average allocation to a flow while the burst credit allocation serves to limit or control short term dominance of a network link by a bursty flow. Examples of flows which exhibit various behaviors are illustrated hereinafter for purposes of demonstrating at least some appropriate settings for the burst and sustained fractions as well as the corresponding limits.
  • FIG. 7 illustrates APPo ( Figure 3) including a representative cell having the information described with regard to Figure 3a contained by cell header 32.
  • packet info 34 is included in packet info 34.
  • This information identifies the cell as one of either start of packet, middle of packet or end of packet. Accordingly, a packet is considered as being active if the indication is middle or end of packet. While a start of packet indication may signal the beginning of a multi- cell packet, such that additional cells are to follow, it is also possible that the packet consists only of a single "start" cell.
  • the present invention implements an APP (Active Packet Priority) state identification process, for example, within each of the APP sections. It should be appreciated that the APP section associated with each flow is essentially identical. The resulting states are indicated for each flow to bandwidth allocation arbiter 90 as part of lines 92, 94 and 98 by augmenting cell header information as it passes through the APP section of each of the flows. In the instance of F 0 line 92 is used.
  • APP Active Packet Priority
  • the latter illustrates steps performed in conjunction with the APP process of the present invention that are generally indicated by the reference number 302 and which is implemented within an APP calculator 304 shown in Figure 7. Again, the present example refers to F 0 .
  • the APP cell type is extracted from cell header data 34 as the cell header data passes through APPo-
  • the APP calculator 304 determines the APP cell type for use in determining the APP value to be presented to arbiter 90 by augmenting line 92.
  • step 312 determines if the cell type is "start of packet.” If this is the case, it is implied that a packet is not in progress because the start of packet indication is given only to the first cell of a packet. In this regard, it should be appreciated that the cell being considered has not yet been transmitted. Therefore, a middle or end of packet indication represents a cell that is needed to complete transmission of a packet that is already underway. Conversely, a start of packet indication is the first cell of that packet and, hence, transmission of the entire packet is pending. For this reason, monitoring only for start of packet is necessary in step 314 to set the active packet value to false.
  • step 316 sets the APP value to true.
  • Step 318 then returns monitoring to step 310.
  • a cell could be "marked" by an external agent as being a cell that should be given APP. That is, a cell is marked with APP at some upstream point.
  • the APP as utilized by BWA 90 will function nonetheless.
  • an integer may represent the packet offset from the first word of data in a particular cell. In this case, APP is true when the offset is equal to zero. That is, the first word of a packet is pending transmission. All of these modifications are considered as being within the scope of the present invention so long as APP is determinable in accordance with the teachings herein. Specific examples illustrating the use of the APP value will be given below.
  • each flow cell header queue supplies cell information to bandwidth arbiter 90 by way of its corresponding APP section.
  • APP 0 provides packet priority information on line 92 which passes through the APP section and which is augmented by raw cell header information in accordance with the active packet priority process described with reference to Figures 7 and 8.
  • Line 92 is provided to a compare information multiplexer 400 along with other information from the BWA State component of the corresponding flow comprising a sustained priority (SP) and a burst priority (BP) indication.
  • SP sustained priority
  • BP burst priority
  • a flow counter 402 stores a flow count that is incremented on a per clock cycle basis so as to sequentially select the information from each flow for routing to a priority encode comparator 404.
  • the latter includes a Candidate Cell input from multiplexer 400 and a Current Cell input. Output of the priority encode comparator is indicated as Selected Cell Update Result.
  • a flow ID 405 for the selected flow is added to the priority indications.
  • the flow ID is derived from factors such as, for example, if a cell is to be grouped (for bandwidth allocation purposes) with other cells such that all cells in that group have the same numerical flow value or if a cell is not to be grouped with any other cells.
  • a new (unused) flow ID is assigned to start another "group of cells.”
  • the latter is typically all cells "owned” by some single paying customer or all cells of a particular characteristic/type (e.g., all internet telephone calls have similar sustained rate and burst rate characteristics and can be grouped so long as the sustained fraction is increased for every "call" added to the particular flow). It is noted that these considerations reflect the parameters recited for use in flow control described above with regard to Figures 3 and 4 since per flow differences are defined in terms of those 4 parameters.
  • bandwidth allocation arbiter 90 includes a current selected cell register 406 and a round robin group 408, indicated within a dashed line.
  • the current selected cell register receives input from the Selected Cell Update Result output of priority encode comparator 404 and provides an output, in the form of the stored values, to the Current Cell input of the priority encode comparator.
  • the round robin group is connected to priority encode comparator 404 and itself includes a sustained credit round robin (SRR) 410, a burst credit round robin (BRR) 412 and a no credit round robin (NCRR) 414.
  • SRR sustained credit round robin
  • BRR burst credit round robin
  • NCRR no credit round robin
  • the objective of the selection process is to select the next cell to be transmitted while the transmission of a previously selected or released cell is already underway.
  • one of the previously selected cells being transmitted is likely to be composed of a plurality of words such that its transmission extends over a series of clock cycles. Therefore, during transmission of such a previously selected cell, the selection of the next cell to be transmitted is continuously updated on each clock cycle.
  • the selected cell is not immediately transmitted onto the network link, but is immediately transferred or released into fransmit header queue 110 of Figure 3 from current selected state register 406. Insofar as the handling of a cell for priority purposes, however, the cell is, for practical purposes, considered as being transmitted upon its release into the transmit header queue.
  • priority encode comparator 404 reads the Candidate Cell input from compare info mux 400 and also reads the Current Cell input from current selected cell register 406. The priority encode comparator then selects between the two inputs based on the inputs themselves and potentially based upon round robin arrangement 408. The input which wins the comparison, either the flow data read from the compare information multiplexer 400, determined by flow counter 402, or the current selected cell read from current selected cell register 406 is then saved in the current selected cell register for use at least in the next comparison. Potentially, the current selected cell may change on each clock cycle. Over the duration of transmission of a previously selected cell, therefore, the current selected cell in register 406 should represent the candidate cell having the highest priority, as selected from among the various flows.
  • transmitter 100 provides select line 124 signals immediately upon de-queue of a cell header from transmit header queue 110, indicating that the cell identified at that instant by current selected cell register 406 is to be transmitted on its output lines connected to transmit header queue 110.
  • This arrangement is considered as being advantageous since it serves to maintain the transmit header queue in a relatively full state.
  • header information for that cell is released into transmit header queue 110, under the requirement that the data is valid. Released data remaining in current selected cell register 406 is immediately invalidated, as will be described.
  • a de-queued header will remain in cell data fetcher 122 until a sufficient portion of data has been retrieved from cell data storage section 42 so as to insure uninterrupted transfer of the cell.
  • step 502 receives an internal clock signal 504.
  • flow counter 402 ( Figure 9) is updated using equation 1 to increment modulo the flow indicated by the counter based on the number of flows.
  • Compare info mux 400 receives the updated flow indication from flow counter 402 and promotes the candidate cell for the indicated flow at the candidate cell input of priority encode comparator 404.
  • Step 506 monitors select line 124 signals to determine if a cell transmission is initiated for any flow. If a cell is transmitted, step 508 is performed to determine if the data in current selected cell register 406 is valid.
  • step 510 is performed, as will be described below. If, on the other hand, the data is valid, step 522 is performed in which data for the current selected cell is released into transmit header queue 110 and to round robin arrangement 408. Step 514 then marks the contents of current selected cell register 406 as invalid.
  • a transmit decoder 524 receives information related to the transmitted word on a line 526 which includes information provided from select line 124 of Figure 3. This information includes the flow queue ID associated with the transmitted word and an indication of the sustained priority state of the flow at the time the cell was selected.
  • These unique signals are derived from the combination of lines 166 and 168 of Figure 3, while data from selected queue decoder 516 and transmit decoder 524 is transferred from arbiter 90 to the respective flow queues on lines 536, 538 and 540.
  • step 510 is performed.
  • the contents of current selected state register 406 and compare info mux 400 are fed into priority encode comparator 404 at its Candidate Cell and Current Cell inputs, respectively.
  • the two priority encode comparator inputs are compared to select the cell flow having the best or highest priority data.
  • Step 544 determines if the Candidate Cell is selected. In this case, the contents of the current selected cell register are to be replaced.
  • Step 546 is then performed to load the Candidate Cell data into current selected cell register 406 via the Selected Cell Update Result output. This information includes cell information, priorities and Flow ID.
  • step 544 determines that the current cell data, already contained by current selected cell register 406, has a relatively higher priority
  • step 548 loads the Current Cell data back into current selected cell register 406 via the Selected Cell Update Result output.
  • step 550 validates the data now stored in current selected cell register 406.
  • step 552 execution moves to the next clock cycle. If, alternatively, step 508 determines that the current selected cell data in register 406 is invalid, the contents of register 406 are replaced with the candidate cell in step 553 and the method then moves to step 542, continuing as described above.
  • Process 600 follows step 512 of Figure 10. Initially, step 602 verifies that a cell was selected. If transmit selection step 602 indicates "NO" on a particular clock cycle, step 604 follows and represents that no change is made in the priority allocation states. Step 605 therefore follows step 604, to return the process to step 602. That is, Current Selected Cell register 406, SRR 410, BRR 412, and NCRR 414 remain unchanged. If, however, transmit selection step 602 indicates that a cell has been selected, step 608 then determines if the released selected flow/cell has sustained credit.
  • step 610 If the flow has no sustained credit, then no credit round robin NCRR 411 is updated in step 610. If the step 608 test indicates that the selected flow has been transmitted with sustained credit, sustained credit round robin, SRR 410 is updated in step 612. Following step 612, step 614 determines if the flow has burst credits. If so, burst credit round robin, BRR 412 is updated in step 616. When a round robin is updated, it is set to the flow number just selected plus one. For example, if a round robin indicates flow 0, but flow 1 was selected since flow 1 did not have the appropriate priority, the round robin would be set to flow 2. If flow 2 were selected in the same scenario, the round robin would be set to flow 0.
  • step robin if the round robin points to the selected flow, the round robin is set to the next highest flow number.
  • steps 610, 616 or after a negative determination on step 614 operation returns to step 602 via step 605, to continue monitoring for a transmit selection on a clock cycle. Specific details as to the use of round robin arrangement 408 in making priority determinations will be described below.
  • priority encode comparator 404 is diagrammatically illustrated, generally indicated by the reference number 404a.
  • the reader is referred to Figure 9 for component interconnections illustrated therein.
  • a different set of data is presented on the Candidate Cell input from compare information multiplexer 400. This data is to be compared with the data present on the Current Selected Cell input from register 406. The latter input is used to construct a current integer 640, for example, within an appropriate register.
  • a candidate integer 650 is constructed, for example, in another register based on the candidate cell input from multiplexer 400. Thus, the comparison is made between current integer 640 and candidate integer 650. Details relating to the construction of these values will be provided at an appropriate point below.
  • multiplexer 654 includes a candidate cell input 656, a current cell input 658 and a selection or result output 660. The data from the input selected in accordance with the indication on line 655 is presented at selection output 660. The selection is then routed to current selected cell register 406 ( Figure 9).
  • current integer 640 is computed as the bitwise concatenation of data presented at current cell input 658 from priority encode comparator 404.
  • current integer 640 is made up of a cell info valid field 642a; a current sustained priority indicator 642b, comprising a single bit; a current burst priority field indicator 642c, also comprising a single bit; a current active packet priority indicator 642d; and a distance (DS) indicator, 642e.
  • Cell info valid field 642a is manipulated by steps 514 and 550 of Figure 10.
  • the current sustained and burst priority indicators are originally obtained from lines 154 and 164, respectively, of Figure 5.
  • APP indicator 642d is derived from the APP computation section of each flow as shown in Figure 3 and described with regard to Figures 7 and 8.
  • the value for distance indicator, DS is produced using, for instance, a current flow distance calculator section 662 in conjunction with a current round robin selector 664.
  • Round robin selector 664 receives inputs from each round robin within round robin group 408 of Figure 9 and also receives the SP and BP indications forming portions of the current selected cell input data.
  • the SP round robin (SRR) input is received on a line 666
  • the BP round robin (BRR) input is received on a line 668
  • the no credit round robin (NCRR) input is received on a line 670.
  • the particular round robin used in determining the value of DS is established in view of SP and BP as received on lines 672 and 674, respectively. If the SP value is set, indicating that the current selected cell has sustained priority while the BP value is zero, indicating that the current selected cell has sustained priority, SP round robin 410 is used in contributing to the value of DS. If SP and BP are both set, indicating that the current selected cell has both sustained and burst priority, round robin selector 664 chooses the burst priority round robin value to be passed to calculator 662. If SP and BP are both zero, indicating that the selected cell has no credit, round robin selector 664 chooses the no credit round robin value to pass to calculator 662.
  • DS calculator 662 determines the distance value in accordance with Table 1, using the round robin pointer value passed from round robin selector 664 along with the flow ID provided on a line 680 from current integer register 640.
  • the DS value is determined based on the flow ID of the cell being represented by current integer 640 and the flow pointed to by an appropriate one of the round robins, as selected by round robin selector 664. For example, if the flow ID is 1, the selected cell has only sustained priority and the sustained round robin points to flow 2, a DS value of 3 is used. A relative priority boost is provided with an increase in the value of DS, as a result of the addition of DS to current integer 640.
  • Candidate register 650 includes essentially identical components for use in establishing the DS value for the candidate cell, consistent with the foregoing discussion relating to current integer 640. These components have been identified using identical reference numbers having an appended prime (') mark and have not been described for purposes of brevity.
  • the underlying purpose of round robin arrangement 408 and associated round robin pointers and distance calculators is to break priority ties between the current cell and the candidate cell. That is, with all other data being equal as between current integer 640 and candidate integer 650, DS will break ties between the flows in a fair manner alternating from one flow to the next.
  • the order of the round robin pointer values that appear in the round robin column of Table 1 for each flow ID may be reordered so long as each of the round robin pointer values is included for each of the flow ID's.
  • the encoded priority scheme shown in Figure 12 relies on the use of APP on a low order priority basis. That is, if the current and candidate registers are tied in all positions above the APP position, packet priority will break the tie in the event that the represented cells possess different APP values, the formulation of which has been described above. In this manner, latency as to transferring a packet for which a first cell has already occurred can be improved.
  • the round robins are updated only when a cell is transmitted.
  • the contents of candidate cell register 650 always replace the contents of candidate register 640 since the priority information relating to that cell may have changed subsequent to the comparison which previously placed its information into current cell register 640.
  • Figure 13 illustrates a first alternative encoding of the candidate and current integers which is generally indicated by the reference numbers 640a/650a.
  • the leftmost bit is the most significant bit.
  • this first alternative encoding prioritizes active packet selection above all other priority factors. That is, active packet priority, APP, cell selection is given higher priority than bandwidth allocation represented as the combination of SP and BP. This encoding is useful, for example, when minimizing latency is more important than maintaining actual bandwidth allocation goals.
  • Figure 14 represents a second alternative encoding which is generally indicated by the reference numbers 640b/650b.
  • This encoding uses a per cell priority (PCP) field 700 in place of the DS field.
  • PCP per cell priority
  • ties are broken in accordance with a strict priority scheme. That is, fairness in breaking ties is given essentially no weight.
  • the priority is per cell and can either be extracted from the cell itself or the priority may be inferred from the flow number (e.g., flow 0 is the lowest priority while flow N-l is the highest priority).
  • the present invention contemplates that an unlimited number of alternative encodings are possible in view of this overall disclosure and that all of these alternative encodings are within the scope of the present invention.
  • the encoding can be arranged to do best effort active packet priority while absolutely maintaining minimum bandwidth allocations.
  • a suitable encoding for this is: [Valid, SP, BP, APP, DS] arranged in the manner of figures 13 and 14.
  • Another example is an encoding which does best effort bandwidth allocation but absolutely maintains minimum packet latency using an encoding: [Valid, APP, SP, BP, DS].
  • Figure 15 is a sustained priority transfer example generally indicated by the reference number 800.
  • the horizontal or X axis of the figure measures time in units labeled successively from 0-9 and A-Z with the letters "O" and "Q" omitted from the series for the purpose of avoiding confusion.
  • Each time unit may be, for example, the amount of time required to transmit a single word, the amount of time required to transmit a single cell of a given size, or the amount of time required to transmit some arbitrary amount of data such as 1500 bytes which could comprise multiple cells of variable individual length.
  • the time units are illustrated as being of equal width across the horizontal axis, but this constraint is not a requirement.
  • the vertical or Y axis denotes units of bandwidth.
  • bandwidth block Only fully consumed blocks of bandwidth are considered. For example, if the X axis units are "time for 128 bytes of data to arrive" then the Y axis unit value indicates the "number of 128 byte portions of data to arrive.” A partially filled bandwidth block is not considered. It should be emphasized that this constraint is imposed for illustrative purposes only and that such a constraint is, likewise, not required in an actual implementation. For example, if a bandwidth block is a byte, less than 8 bits could be transferred, since such a transfer may be required at times.
  • Figure 15 contemplates three flows Fo-F 2 passing through switch 12 of the present invention.
  • the upper portion of the figure illustrates ingress data 802 representing the arrival of the data from each flow at the network switch.
  • the lower portion of the figure illustrates queued data 804 for each flow awaiting transmission along with a lowermost egress data map 806 showing the order of transmission of the data obtained from queued data 804.
  • a block of data takes up the entire link for that unit of time. That is, for the ingress data, a single block indicates that one of the input ports was fully utilized for that time period by the corresponding flow's data or, for the egress data, that one of the output ports was completely utilized for that time period.
  • Ingress data may represent data inbound from any of the input ports of switch 12, the combination of which determines the depth of data blocks stacked vertically along the Y axis. For an 8 port switch, as an example, the maximum stack depth is therefore 8.
  • Each data block is labeled with the notation ⁇ Flow ID (IngressTime) >. ⁇ Time(Ingress#)>. ⁇ Packet #(Channel ID) > where ⁇ Time > is the Y position of the block in the ingress stack. Packet # is not relevant to the present example, but will be considered in conjunction with at least one subsequent example. Therefore, every block/data unit is labeled uniquely for purposes of tracking its progress through the example. It should be appreciated that a block arriving on a flow at time 0 will be queued at time 1.
  • This same data unit will then be delayed by at least one additional time period prior to being transmitted on the egress data map. See, for example, block 0.0.0 in time period 0 on ingress data F 0 , time period 1 on F 0 queued data 804 and time period 2 on egress data map 806.
  • the blocks are skewed to the right in moving down the figure from an ingress position, to a queued position, to an egress map position.
  • Sustained priority ties are broken round robin in the order 0,1,2,0,1,2... in cases where all three flows are able to consistently maintain sustained priority.
  • the round robin corresponding to the tied priority level is relied on to break the tie and is subsequently updated.
  • the former illustrates sustained fraction assignments invoked in the present example. It is noted that the sustained limit, burst fraction and burst limits are not used since the purpose of this example is to facilitate an understanding of the sustained fraction.
  • the latter appears in Figure 4 designated by the reference number 72. Accordingly, F 0 , F] and F 2 are respectively assigned sustained fractions of 50%, 25% and 25%. All data units arrive for the three flows in succession, as shown. Due to skewing, the first data unit to be transmitted at time 2 is 0.0.0.
  • each flow begins with 1.0 SP credits.
  • the SP priority level is set to 2.0 credits. It should be appreciated that this is the value specified in SC threshold register 152 (see Figure 5) of the BWA state machine for each flow.
  • NCRR and SP pointers values are also shown and are both initially set to point to flow 0 at time 0.
  • flow 0 again wins the arbitration.
  • the SRR is updated by setting it to the ID number of the flow transmitted plus one. That is, the SRR is again set to flow 1.
  • the flows are all tied at SP. Because the SRR points to flow 0, its data unit should be transmitted at time E; however, the flow 0 queue is empty. Therefore, the next higher flow with SP (flow 1) is permitted to transmit at time E. Accordingly, the SRR is set to the value 2 in the time E row.
  • flow 0 still has nothing to transmit while flows 1 and 2 are in a no credit tie. In this case, the NCRR is used to break the tie, permitting flow 1 to transmit data unit 1.4.0 at time G.
  • the NCRR is then set to point to flow 2 in the time G row. Over times J-M only flow 2 has data to transmit.
  • the first transmit, at time J, is accomplished with SP causing the SRR to be set to 0 in the time I row.
  • the times K and L transmits are made with no credit causing the NCRR to be set to 0 in each of these rows.
  • Flow 2 regains sustained priority for the transmission of 2.8.0 at time M, but the SRR is once again re-set to flow 0.
  • all flows are out of queued data such that idles occur (no transmission). Throughout these idles, the priority state of each flow is maintained. The maintenance of the states would also occur in situations where the flow queues still contain data but, for some reason, no transmissions are taking place. Such a situation could result, for example, if a downstream link has failed or due to downstream flow control which can cause flow constriction via clock control.
  • Figure 17 illustrates an available priority bandwidth allocation example performed in accordance with the present invention and generally indicated by the reference number 820.
  • Ingress data is designated by the reference number 822
  • queued data is designated by the reference number 824
  • the egress data map is designated as 826.
  • Figure 17 illustrates flows F 0 -F 2 passing through switch 12 of the present invention under the SP configuration parameters specified by table 3. It is noted that F 0 is assigned a sustained allocation fraction of 33%. This value is satisfied by the average arrival of a data unit in one out of every three time periods.
  • each flow once again begins with 1 SP credit and the SP credit level is set at 2 credits.
  • Figure 18 is similar to previously described Figure 16, showing sustained credits and pointer values. Other similar figures will be provided, as appropriate.
  • the NCRR decides the data unit to be transmitted since no flow has reached the SP level of 2 credits.
  • data unit 0.3.0 is transmitted with SP.
  • data units 0.6.0 and 0.6.1 arrive on the flow 0 queue.
  • Flow 0, at this time has SP and, therefore, data unit 0.6.0 is immediately transmitted at time 8.
  • Data unit 0.6.1 remains in the flow 0 queue.
  • the credit situation among the three flows at time 8 is such that there is a three-way no credit tie.
  • data unit 2.R.0 is transmitted at time T and the SRR is again updated. Also at time T, the priority determination finds flow 0 alone having SP. Accordingly, data unit O.P.O is transmitted at time U. Data unit O.P.I is then transmitted at time V, since this is the only data unit ready to transmit in the flow queues.
  • Any flow having queued data may transmit so long as no other flow has superior priority and in accordance with selection using the NCRR, if some other flow is tied at no priority.
  • a flow receiving the leftover bandwidth is not debited either type of priority credit (sustained or burst) as a result of making this leftover bandwidth transmission. In essence, such a flow has been given a "free ticket.”
  • prior art priority allocation schemes are devoid of this scheme. For example, while ATM provides an Available Bit Rate (ABR) concept that is intended to account for leftover bandwidth, ABR is not capable of such instantaneous response. In effect, ATM must readjust allocation parameters to "after the fact" account for leftover bandwidth. Some of the leftover bandwidth is inherently wasted in this manner.
  • ABR Available Bit Rate
  • Figure 19 illustrates a reduced network bandwidth example carried forth in accordance with the present invention and generally indicated by the reference number 830.
  • Ingress data is designated by the reference number 832
  • queued data is designated by the reference number 834
  • the egress data map is designated as 836.
  • this example is identical to aforedescribed Figure 15 with respect to ingress of the data on the three flows.
  • the present example is also identical to that of Figure 15 over the first 7 time periods.
  • the data units within queued data 834 are identical to those in queued data 804 of Figure 15.
  • time period 8 it is assumed that the available bandwidth on the output network link is reduced by 50%, as is represented by doubling the clock period for egress data map 836 at time 8. This condition persists until time P.
  • ATM uses flow configuration parameters typically in a strict data rate sense such as, for example, a particular Flow A gets 10000 bytes/second.
  • a strict data rate sense such as, for example, a particular Flow A gets 10000 bytes/second.
  • Flow A gets 10000 bytes/second.
  • the present invention maintains the desired flow percentages even when the bandwidth approaches constriction at 0 bytes/second.
  • both the sustained fraction and sustained limit values are specified, for instance, in units of words. These parameters appear in Figure 4 designated by the reference numbers 72 and 74, respectively. Only queued data is illustrated due to drawing space constraints and because illustration of ingress data is not necessary to an understanding of the present example.
  • the sustained limit for flows 0 and 1 is set to 128 words while the sustained limit for flow 2 is set to 65 words.
  • the sustained priority/no-priority boundary in the example is 64 word credits, specified in SC threshold register 152 of Figure 5. Therefore a flow having a number of sustained credits greater than or equal to 64 is given sustained priority. As mentioned previously, this limit may correspond to the number of credits required to transmit a full sized cell (i.e., 64 words transmitted in 64 consecutive time periods).
  • the notational convention within the data blocks of Figure 22 is
  • flow 0 has sustained credit (since it's SP credit value > 64) while flow 1 does not have sustained priority at 63.75 credits. Accordingly flow 0 is selected for transmission (flow 0, cell 0). As flow
  • flow 0, cell 2 is selected and its 2 cells are transmitted at times B and C.
  • the NCRR is set to flow 1 at time B.
  • another no credit arbitration is performed between flows 0 and 2 with flow 1 winning since the NCRR indicates flow 1.
  • Flow 1, cell 1 is made up of five cells that are transferred over times D-H.
  • flow 1 has regained SP and is selected to transmit a two word cell, 1.2.0 and 1.2.1 at times I and J, respectively.
  • flows 0 and 1 have no priority.
  • flow 0 indicates flow 2
  • flow 0 is selected to transmit a three word cell: 0.3.0, 0.3.1 and 0.3.2. Since flow 0 was selected the NCRR is set to flow 1 at time K.
  • the network link is idle during times N, P and R since no data is available to transmit.
  • flow 1 has SP and is selected to transmit a three word cell: 0.4.0, 0.4.1 and 0.4.2 at times S, T and U, respectively.
  • flow 1 is the only flow with data to transmit and is selected to transmit its third cell made up of four words: 1.3.0, 1.3.1, 1.3.2 and 1.3.3.
  • the sustained credit limit is reached at time 4 and grows no more after this point.
  • the limit prevents a flow from accumulating a great number of sustained credits which would translate into a long period of sustained priority. If no limit were provided, there is a potential, unwanted burst of traffic (from flow 2) that could be quite long.
  • the sustained limit is set to limit the accumulation of sustained credits allowing only manageable burst lengths. In this example, the manageable burst tolerance is very low. Hence, if flow 2 were to become active, its cells would be highly interleaved with the other active flows.
  • a burst fraction example is illustrated, generally indicated by the reference number 850.
  • the sustained and burst fractions are both utilized. These parameters appear in Figure 4 designated by the reference numbers 72 and 76, respectively.
  • the network link output is 100% utilized such that there is always queued data in order to transmit.
  • Flows 1 and 2 have completely uniform arrival patterns.
  • Flow 0 is fully bursty. That is, flow 0 fully provides its 50% egress bandwidth but does so by bursting at 4x the configured egress bandwidth allocation, but only bursts one quarter of the time.
  • this goal is accomplished by setting the burst fraction for flow 0 to 100%. That is, whenever flow 0 has sustained credit, it will always have burst credit which, in effect, insures that there is never an SP tie between flow 0 and any other flow; flow 0 will always win in the instance where it has credit.
  • flows 1 and 2 have their burst fractions set to zero so that they never accumulate burst credits.
  • the need to use the burst fraction in this example resides in the requirement that all cells on flow 0 must be transmitted as soon as possible on the output bus, prior to cells of flows 1 or 2. If this requirement is violated, the subsequent ingress of cells on flow 0 will ultimately be delayed such that the upstream device is unable to supply data at the 50% rate. In this event, the network switch executing the present example will not be able to sustain the output bandwidth of 50%.
  • the sustained priority threshold is set to 2 credits. The reader is reminded that flow 0 will have burst priority at all times that it has sustained priority. Flow 0 first achieves sustained/burst priority at time 3. Flows 0 and 1 are still at no credit, below the 2 credit threshold. Therefore, flow 0 transmits cell 0.0.1 at time 4.
  • Flow 0 again reaches sustained/burst credit at time 5 in what would be a three-way SP tie without the burst priority held by flow 0. Accordingly, flow 0 transmits cell 0.0.2 at time 6. Flow 0 then transmits cell 0.0.3 at time 8 with burst/sustained priority, completing transmission of the first burst of four cells initially received at time 0 in the flow 0 ingress data. Flow 0 accumulates burst/sustained credits over times 9, A and B, since no queued data is available to flow 0. Thus, at time B, when queued data is available to flow 0, 3.0 burst/sustained credits are available
  • the last cell of the second burst, 0.A.3 is transferred at time G following flow 0 reaching burst/sustained priority at time F.
  • This pattern essentially repeats in egress data 856 at times K and U for the remaining two bursts which appear in the ingress data at times I and S. It is noted that the burst round robin value is not shown in Figure 24 for the reason that flow 0 will always be selected if it has burst/sustained credit since no other flow attains burst credit. Each time flow zero transmits, the burst round robin is reset to point to flow 1. The BRR value is not relied on when only one flow has burst credit.
  • the BRR operates in essentially the same manner as the SRR in cases where more than one flow has burst credit. Specifically, the flow pointed to by the BRR is selected if that flow has burst credit. If the pointed to flow does not have burst credit, the next highest flow with burst credit is selected. The BRR is then set to the selected flow ID plus one.
  • burst priority example of Figures 22 and 23 the use of two types of credits assigned to a single flow is considered to be highly advantageous.
  • the use of burst credits permits a priority configuration that boosts the priority of flow 0 in a way that assures transmission of flow 0.
  • flow 0 is assigned a sustained fraction of 0.5 or 50%, it is observed that over the eight time periods from time 2 to time 9, flow 0 is allocated the output bandwidth on four of these time periods.
  • one-half of the output bandwidth is again given to flow 0.
  • time periods I through R and S through Z one-half of the output bandwidth is dedicated to flow 0.
  • the priority configuration of this example provides one extra time period between completion of one flow 0 burst and starting the next flow 0 burst, implying that a smaller burst fraction will also work. Since there is one extra time period, the burst fraction can be reduced to allow the one additional intervening cell. Accordingly, a burst fraction of 0.667 or greater will accomplish the contemplated goals in this example. In other words, even at a setting of 0.667, burst credits will accumulate to a degree that is sufficient to insure timely transmission of the flow 0 data.
  • the following example serves to illustrate the result when burst credits are not used, as in the immediately foregoing example.
  • Delay is imposed with receipt of the flow 0 ingress bursts at times B, K and V in Figure 25 as compared with times A, I and S, respectively, for the same bursts in Figure 23.
  • the delay in receiving the ingress data bursts on flow 0 could be attributed to a number of factors.
  • One factor is a switching device that aggravates an upstream device throughput issue by virtue of having limited buffering. That is, only a limited amount of data buffering is available in the switch for ingress traffic. Once the buffers have been filled, the switch must apply some sort of flow control back to the upstream device to prevent that device from sending more data for which no storage buffers exist, if that data were to arrive prior to a buffer being freed (typically via transmission). This flow control may cause the upstream device to incur additional latency once a buffer does free up in the switch and the upstream device receives the "ok to send" message.
  • the upstream device may be bursty, avoiding the latency as often as possible by bursting out enough data to fill the switching device's buffers and then delaying further transmission to allow the switching device buffers to clear.
  • the present invention contemplates matching the network switch or device to an input device such that data is bursted out at the same rate as the incoming burst. This match is achieved using the combined sustained and burst fraction settings.
  • the present invention contemplates setting up an inbound burst for transmission upon receipt with a higher priority than the priority at which the originating, upstream device transferred the burst. This is, in essence, achieved by using the burst fraction as illustrated in connection with Figure 23.
  • the benefits of this arrangement reside in reduced latency through maintaining flow of the burst in order to avoid the aforedescribed startup latency.
  • hardware costs may be reduced since there is a corresponding reduction in the amount of buffering that is required. That is, a buffer capable of storing a full sized packet is not necessary.
  • Upstream device requirements are satisfied by efficiently buffering only portions of packets in conjunction with the highly advantageous priority allocation concepts of the present invention. In contrast, the typical prior art approach is to provide ever more buffering.
  • FIG e illustrates an active packet priority example performed in accordance with the present invention and generally indicated by the reference number 870.
  • the present example contemplates the application of the priority encoding described above with regard to Figure 13. That is, the active packet priority data is given seniority for the use of output network link bandwidth above all other considerations.
  • Three packets are illustrated, one of which arrives on each of flows 0-2. The packets are all of the same length and arrive simultaneously for illustrative purposes. The packets are scheduled such that, as long as there is data for an active packet to be sent, that data is prioritized over sending any data from a new (non-active) packet.
  • the flow 0 packet is given the use of the network link to transfer cell 0.0.0, for example, as a result of the NCRR pointing to flow 0.
  • Each successive cell of the flow 0 packet then has priority for transfer based on APP. Therefore, the flow 0 packet is transferred over times 2-7.
  • an arbitration is performed based on either the sustained or no credit round robin, depending upon the sustained fraction setting.
  • the application of burst credits and the burst round robin may also be an arbitration factor, but is not considered in the present example for purposes of clarity.
  • the flow 1 packet is selected for transfer, for example, if the SRR indicates flow 0 and both flows 1 and 2 have gained sustained priority during the transfer of the flow 0 packet, flow 1 would be selected as having the next highest flow ID. The SRR would then be set to flow 2. Alternatively, if flows 1 and 2 do not have sustained priority, the NCRR would indicate flow 1 as a result of selection of cell 0.0.0 at time 1. Therefore, flow 1 would be selected using the NCRR. In either instance, the flow 1 packet is transferred over times 8-D. The flow 2 packet is then given the use of the network link since the flow queues of the remaining flows are empty. The flow 2 packet is transferred over times E-J.
  • the last cell of the flow 0 packet appears in egress data 874 at time 7; the last cell of the flow 1 packet appears in the egress data at time D; and the last cell of the flow 2 packet appears in the egress data at time J.
  • Transfer latency is indicated for each of the flow packets in terms of the total time taken to complete packet transfer as measured from time 0. Specifically, the latency for packet 1 is 8 time units while the latencies for the flow 1 and 2 packets are 14 and 20 time periods, respectively. Specific advantages of the APP feature will be discussed in conjunction with the next example.
  • the latency of the flow 2 packet is at 20 time periods in both figures.
  • the advantage of the APP feature is evident, however, in view of the relative latencies of the flow 0 and flow 1 packets. The latter experiences a latency of 19 time periods in Figure 28 and a latency of only 14 time units in Figure 27: a reduction of 5 time periods.
  • the latency of the flow 0 packet is reduced from 18 to only 8 time periods: a reduction of 10 time periods. Accordingly, it is submitted that the APP feature of the present invention is highly advantageous in environments where it is desired to transfer at least certain packets with the lowest possible latency. Average latency has been significantly reduced while worst case latency is unaffected.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

La présente invention concerne une attribution perfectionnée de largeur de bande de réseau numérique. Dans un mode de réalisation de l'invention, les flux nécessitent une attribution de largeur de bande minimum sur la liaison du réseau et on transfère ces flux en utilisant des cellules de longueurs variables multiplexées à partir de différents flux. Dans un autre mode de réalisation de l'invention, les paquets en provenance de flux de données sont divisés en une série de cellules de façon que chacune de ces cellules soit identifiable comme partie de paquets spécifiques. On transfère au moins ces paquets spécifiques sur la liaison de réseau de façon incrémentale en alternant chaque série de cellules de paquets respectifs. Au niveau d'une seconde extrémité de la liaison de réseau, on utilise les informations d'identification pour reconstruire la série de cellules de chaque paquet spécifique de façon à produire les paquets spécifiques au niveau de cette seconde extrémité de liaison de réseau. Dans encore un autre mode de réalisation de l'invention, on forme les cellules en utilisant les flux de données de sorte que chaque flux de données soit constitué d'une ou de plusieurs séries de cellules. A un moment donné, lorsqu'on ne peut attribuer la liaison de réseau à aucun des flux de données en respectant les attributions de priorité établies pour les flux de données, on transfère au moins une cellule additionnelle faisant partie d'un de ces flux, sans tenir compte de l'attribution de priorité affectée à ce flux. Enfin, d'autres modes de réalisation de l'invention, comprennent une réponse à la réaffectation de priorité des salves et de largeur de bande variable de liaison de réseau.
PCT/US2001/019006 2000-06-19 2001-06-13 Attribution de largeur de bande de reseau numerique et procede et appareil WO2001099342A2 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
AU2001268384A AU2001268384A1 (en) 2000-06-19 2001-06-13 Digital network bandwidth allocation method and apparatus

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US59660900A 2000-06-19 2000-06-19
US09/596,609 2000-06-19

Publications (2)

Publication Number Publication Date
WO2001099342A2 true WO2001099342A2 (fr) 2001-12-27
WO2001099342A3 WO2001099342A3 (fr) 2002-09-19

Family

ID=24387966

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2001/019006 WO2001099342A2 (fr) 2000-06-19 2001-06-13 Attribution de largeur de bande de reseau numerique et procede et appareil

Country Status (2)

Country Link
AU (1) AU2001268384A1 (fr)
WO (1) WO2001099342A2 (fr)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6763025B2 (en) 2001-03-12 2004-07-13 Advent Networks, Inc. Time division multiplexing over broadband modulation method and apparatus
US6891841B2 (en) 2001-03-12 2005-05-10 Advent Networks, Inc. Time division multiple access over broadband modulation method and apparatus
WO2007093879A3 (fr) * 2006-02-13 2007-12-06 Nokia Corp Appareil, procédé et logiciel permettant la sélection d'une segmentation de paquets
EP2575302A1 (fr) 2011-09-30 2013-04-03 British Telecommunications Public Limited Company Procédé de gestion de bande passante pour la livraison de contenu et aspects correspondants
WO2015038504A1 (fr) * 2013-09-11 2015-03-19 Nec Laboratories America, Inc. Service sans à-coups dans un transpondeur optique à débit variable

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5323395A (en) * 1991-12-18 1994-06-21 Alcatel N.V. Method and system for sending data over a transmission link in variable length blocks in asynchronous transfer mode
EP0961519A2 (fr) * 1998-05-28 1999-12-01 Newbridge Networks Corporation Division de Bande, des types de connection, et des categories de service pour des services des joncteurs commutés
EP0977405A1 (fr) * 1998-07-31 2000-02-02 Alcatel Méthode, planificateur, mémoire tampon intelligente, processeur et système de télécommunication pour la distribution de la bande passante disponible

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5323395A (en) * 1991-12-18 1994-06-21 Alcatel N.V. Method and system for sending data over a transmission link in variable length blocks in asynchronous transfer mode
EP0961519A2 (fr) * 1998-05-28 1999-12-01 Newbridge Networks Corporation Division de Bande, des types de connection, et des categories de service pour des services des joncteurs commutés
EP0977405A1 (fr) * 1998-07-31 2000-02-02 Alcatel Méthode, planificateur, mémoire tampon intelligente, processeur et système de télécommunication pour la distribution de la bande passante disponible

Non-Patent Citations (6)

* Cited by examiner, † Cited by third party
Title
FRANK M ET AL: "Practical experiences with a transport layer extension for end-to-end bandwidth regulation" LOCAL COMPUTER NETWORKS, 1997. PROCEEDINGS., 22ND ANNUAL CONFERENCE ON MINNEAPOLIS, MN, USA 2-5 NOV. 1997, LOS ALAMITOS, CA, USA,IEEE COMPUT. SOC, US, 2 November 1997 (1997-11-02), pages 337-346, XP010252439 ISBN: 0-8186-8141-1 *
KAWASAKI T ET AL: "A STRATEGY OF QUALITY CONTROL ON ATM SWITCHING NETWORK - QUALITY CONTROL PATH (QCP)" COMMUNICATIONS: THE KEY TO GLOBAL PROSPERITY. GLOBECOM 1996. LONDON, NOV. 18 - 22, 1996, GLOBAL TELECOMMUNICATIONS CONFERENCE (GLOBECOM), NEW YORK, IEEE, US, vol. 1, 18 November 1996 (1996-11-18), pages 432-436, XP000742193 ISBN: 0-7803-3337-3 *
LAMTI L ET AL: "An algorithm for traffic enforcement in multi-tasks operating systems" COMPUTERS AND COMMUNICATIONS, 1997. PROCEEDINGS., SECOND IEEE SYMPOSIUM ON ALEXANDRIA, EGYPT 1-3 JULY 1997, LOS ALAMITOS, CA, USA,IEEE COMPUT. SOC, US, 1 July 1997 (1997-07-01), pages 609-613, XP010241422 ISBN: 0-8186-7852-6 *
PAREKH S ET AL: "Some performance trade-offs associated with ATM fixed-length vs. variable-length cell formats" IEEE, XP010071779 *
SALAHI J: "Modeling the performance of a broadband network using various ATM cell formats" IEEE, XP010081249 *
YONG-HEE JEON ET AL: "Priority-based dynamic access policies to WAN from LAN/MAN" LOCAL COMPUTER NETWORKS, 1991. PROCEEDINGS., 16TH CONFERENCE ON MINNEAPOLIS, MN, USA 14-17 OCT. 1991, LOS ALAMITOS, CA, USA,IEEE COMPUT. SOC, US, 14 October 1991 (1991-10-14), pages 214-222, XP010025763 ISBN: 0-8186-2370-5 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6763025B2 (en) 2001-03-12 2004-07-13 Advent Networks, Inc. Time division multiplexing over broadband modulation method and apparatus
US6891841B2 (en) 2001-03-12 2005-05-10 Advent Networks, Inc. Time division multiple access over broadband modulation method and apparatus
US7194001B2 (en) 2001-03-12 2007-03-20 Advent Networks, Inc. Time division multiplexing over broadband modulation method and apparatus
WO2007093879A3 (fr) * 2006-02-13 2007-12-06 Nokia Corp Appareil, procédé et logiciel permettant la sélection d'une segmentation de paquets
EP2575302A1 (fr) 2011-09-30 2013-04-03 British Telecommunications Public Limited Company Procédé de gestion de bande passante pour la livraison de contenu et aspects correspondants
WO2013045870A1 (fr) 2011-09-30 2013-04-04 British Telecommunications Plc Gestion de largeur de bande pour distribution de contenu
US9794189B2 (en) 2011-09-30 2017-10-17 British Telecommunications Public Limited Company Bandwidth management for content delivery
WO2015038504A1 (fr) * 2013-09-11 2015-03-19 Nec Laboratories America, Inc. Service sans à-coups dans un transpondeur optique à débit variable

Also Published As

Publication number Publication date
WO2001099342A3 (fr) 2002-09-19
AU2001268384A1 (en) 2002-01-02

Similar Documents

Publication Publication Date Title
EP0763915B1 (fr) Dispositif et méthode de transfert de paquets, appropriés pour un grand nombre de portes d'entrée
US6434155B1 (en) Weighted round robin engine used in scheduling the distribution of ATM cells
US6205118B1 (en) Adaptive time slot scheduling apparatus and method for end-points in an ATM network
CN100405344C (zh) 用于在交换结构中分发缓冲区状态信息的装置和方法
US7813365B2 (en) System and method for router queue and congestion management
KR100326789B1 (ko) 패킷통신시스템,패킷통신모듈,메모리공간할당시스템및메모리공간할당방법
US8599870B2 (en) Channel service manager with priority queuing
EP0678996A2 (fr) Appareil et méthode pour le traitement des besoins en largeur de bande passante dans un commutateur ATM
EP0981228A2 (fr) Planificateur de bande passante à deux composants, avec application dans des systèmes de communication digitales à plusieurs classes
EP0712220A1 (fr) ContrÔle de débit noeud à noeud dans un réseau ATM
WO1995030294A1 (fr) Architecture de mode de transfert asynchrone et element de commutation
US7292576B2 (en) ATM switch having output buffers
US10063478B2 (en) Switching device and control method of switching device
EP1111851B1 (fr) Système ordonnanceur pour l'ordonnancement de cellules ATM
WO2001099342A2 (fr) Attribution de largeur de bande de reseau numerique et procede et appareil
KR100565110B1 (ko) 데이터 패킷의 분배 및 처리 유닛
US7330475B2 (en) Method for sharing the bandwidth available for unicast and multicast flows in an asynchronous switching node
EP1258114A1 (fr) Procede et dispositif de mise en forme du trafic de donnees
EP1111854A2 (fr) Modes minimum et maximum pour la transmission de cellules de données ATM sur des bancs de canaux de réseau dans un système de communications
EP1111852A2 (fr) Pile de plans avec méchanisme de contrepression pour la régulation de courants de cellules de données ATM
EP1111853A2 (fr) Méthode pour provisioner une fréquence de trames garantie dans un allocateur hiérarchique
EP1158731A2 (fr) Améliorations dans ou liées à un commutateur de paquets
JPH03280644A (ja) 動的スロット割当て多重化方式
KR20010048125A (ko) 다중 단계 큐를 이용한 전송 지연시간 보장형 에이에이엘2셀 패키징 장치 및 방법
JPS63260234A (ja) バ−スト多重化方法

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A2

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT TZ UA UG UZ VN YU ZA ZW

AL Designated countries for regional patents

Kind code of ref document: A2

Designated state(s): GH GM KE LS MW MZ SD SL SZ TZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE TR BF BJ CF CG CI CM GA GN GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
AK Designated states

Kind code of ref document: A3

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT TZ UA UG UZ VN YU ZA ZW

AL Designated countries for regional patents

Kind code of ref document: A3

Designated state(s): GH GM KE LS MW MZ SD SL SZ TZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE TR BF BJ CF CG CI CM GA GN GW ML MR NE SN TD TG

REG Reference to national code

Ref country code: DE

Ref legal event code: 8642

122 Ep: pct application non-entry in european phase
NENP Non-entry into the national phase in:

Ref country code: JP