KR20070023750A - Network devices and traffic shaping methods - Google Patents

Abstract

In a network device of a broadband access device, the granularity of queue speeds is improved by introducing intermediate speeds in addition to the speeds conventionally provided. For example, when an intermediate rate between the first 10 Mbps queue rate and the second 5 Mbps queue rate is desired for the traffic flow, the skip interval of the traffic shaper is configured to calculate a skip interval value according to the 5 Mbps rate. Each PDU in the traffic flow will receive an additional alternate-interval-calculation decision. Instead of the skip interval value calculated for the PDU in the traffic shaper, the skip zero queue interval (used at 10 Mbps rate) has a 50% chance, for example. Applies to PDUs. As a result, an intermediate speed such as 7.5 Mbps is obtained.

Traffic flow, data queue, traffic shaper, packet data units.

Description

Network Devices and Traffic Shaping Methods {NETWORK DEVICES AND TRAFFIC SHAPING METHODS}

TECHNICAL FIELD The present invention relates to broadband access systems and, more particularly, to network devices for digital subscriber line (DSL) access systems.

At present, the main broadband service to homes is high speed internet access, which has transformed home computers into "always-on" home appliances. Public telephone networks typically use modems that transmit data at, for example, 56 kbps. In a telephone network, modems use the same frequency range as telephones to transmit speech, which is approximately 400 Hz to 4 kHz. However, the line between the switching center and the user (ie subscriber line) allows frequencies much higher than 0 to 4 kHz to be used, thus allowing more data to be transmitted. This 4 kHz limit is due to filters located at the switching centers and filtering additional frequencies from the signal provided to the backbone network. In practice, in subscriber lines, typically made of twisted copper pairs, frequencies as high as several MHz may be transmitted.

DSL (Digital Subscriber Line) is a technology that supplies high bandwidth information, for example, to homes and small businesses over conventional copper telephone lines. A digital subscriber line is a technique that assumes that digital data does not require any change in analog form (and vice versa). Digital data is transmitted directly to the subscriber as digital data, which allows much wider bandwidth to be used in transmitting data than in traditional telephone systems. Several DSL interfaces are available, such as ADSL (Asymmetric DSL), SHDSL (Single-Pair High Speed DSL), and VDSL (Ultra High Speed DSL).

An example of an xDSL system is shown in FIG. Subscriber end-user equipment 1 (such as a router, modem, or network interface card (NIC)) located in premises owned or controlled by a customer using network services is often a customer premises equipment. It is called (CPE). On the network side, the multiplexing equipment 2, which may include dense central office distributors, xDSL modems, and other electronic equipment for connecting traffic to the data network, is often referred to as a digital subscriber line access multiplexer (DSLAM). DSLAM delivers very high speed data transmission over existing copper telephone lines and controls and routes digital subscriber line (xDSL) traffic between CPEs and network of network service providers such as ATM / IP networks.

One of the challenges for DSL operators is to select and deploy the correct deployment model for broadband services. Currently, operators are choosing different models, with a few operators using Point-to-Point Protocol (PPP) over ATM, some operators using 1483 routed, and most Operators use PPP over Ethernet or 1483 bridged. Operators have different conditions for choosing a deployment model. This selection may be based, for example, on the number of public IP addresses required, the CPE or CPE facility price, and the wholesale model. One of the key conditions today is the Layer 2 technology used in access networks.

The standard specifies the ATM used at the transport layer on top of the physical ADSL layer. If ATM is also used at the network side, connections from DSL subscribers are easily established, switched using ATM permanent virtual connections (PVCs), and no IP level intelligence is required. ATM access networks basically support all deployment models, but introduce unnecessary overhead and are difficult to manage, while Ethernet-based access networks cannot support deployment models where ATM exists. In Ethernet networks, the traditional ATM switching model needs to be replaced by packet switching, which can be performed at the Ethernet layer (layer 2) or IP layer (layer 3) depending on the chosen deployment model.

All of the access technologies have limitations as well as advantages, and support of all possibilities creates more requirements for DSLAMs. DSLAM must support simultaneous feeds from ATM and IP interfaces, which must simultaneously provide the required service levels over both QoS (ATM) and CoS (IP) connections and flows . ATM QoS (Quality of Service) allows service providers to fully utilize the available bandwidth, and provides fixed bit rate (CBR), variable bit rate real time (VBR-rt), and variable bit rate non-real time (VBR-nrt) Manage multiple service classes. IP CoS provides rules on how to manage IP flows. Typical IP classes of service classes include Expedited Forwarding (EF), Assured Forwarding (AF), and Default Forwarding (DF).

One approach to specifying the DSLAM is shown in FIG. 2. Packets and Cells ATM and IP traffic flows are entered into the DSLAM in order to ensure that they are properly handled for the type of information or services involved (sorting and marking, policing). ), Queuing, and scheduling and shaping. These operations can be performed using a network processor that can support many different communication protocols while providing wire-speed performance in a flexible programming environment. One such processor is the PayloadPlus traffic management network processor from Agere Systems. In DSLAM applications, the network processor may be configured to function as a central switch, and may supply a plurality of DSL line cards, each with many ports. Internal logic provides high-level control of network processor scheduling functions (ie, traffic manager (TM) and traffic shaper (TS) computational engine (CE)). The TM computation engine (CE) in the network processor provides globally programmable traffic management capabilities including buffer management policies through a reduced, optimized C instruction set. To enforce the buffer management policy, short scripts are run on the buffer management-processing engine. The traffic management script decides whether to discard or schedule the traffic. This decision is based on a user programmed revocation policy. In addition, the network processor has facilities for scheduling cell and frame-based traffic using multiple scheduling algorithms. The processor's configuration, traffic shaper (TS) computation engine (CE), and programmable queue definitions define traffic shaping. For example, traditional ATM fixed bit rate (CBR), variable bit rate (VBR) and non-specific bit rate (UBR) schedulers are programmable in the network processor. The traffic shaper computation engine executes the defined program on every scheduling event. The traffic shaper can also be used for variable-bit rate cell-based traffic scheduling, using dual leaky bucket algorithms, for frame-based traffic using algorithms such as weighted round robin, and each. Frame-based, smoothed, under-weighted by tracking rate credits for the queue. It can be used to support a service queue selection class for round robin scheduling.

For each logic output port (eg, mapped to xDSL port) of the traffic shaper computer engine, there may be several queues and traffic shaping schedulers. The transmission of cells or blocks from the queues to the logical output port is scheduled based on the so-called main scheduling time wheel principle. The logical output port has a predetermined port speed (eg 10 Mbps). QoS (ATM) and / or CoS (IP) connections and flows share port speed. For this purpose, the queue interval is determined dynamically for each queue. The queue interval defines the number of main scheduling time slots that will be skipped before subsequent blocks or cells are sent from a particular queue. If the number of time slots to be skipped is zero, then subsequent blocks are also sent from the same queue if the queue is not empty. This means that the queue rate is 10 Mbps. If the number of time slots to be skipped is 1, the traffic shaper CE transmits from another queue with the next highest priority before sending back from the first queue. This means that the queue rate is 5 Mbps. Similarly, when the number of time slots to be skipped is 2 or 3, a queue rate of 3.33 Mbps or 2.5 Mbps, respectively, is obtained.

The problem with this approach is inadequate granularity in QoS provisioning when using typical QoS / CoS traffic shaping. For a 10 Mbps port, only 10, 5, 3.3, 2.5, 2, 1.7, 1.3 and 1 Mbps are obtained as queue speeds. In addition, there may easily be significant latency in sufficiently recognized scheduling parameters, so QoS provisioning will not be very accurate. Additionally, due to the very limited scripting interface, sophisticated scripting is required to facilitate the synthesis of ATM and packet data traffic. Traffic schedulers are controlled through programmable scripts that are executed for each block. In an Agere network processor, this means that the script must be executed in 22 clock cycles (RSP global pulse), which strictly limits the complexity that algorithms can impose.

It is an object of the present invention to improve scheduling and traffic shaping to mitigate or overcome at least one of the above problems.

The main objects of the invention are disclosed in the appended independent claims. Further objects of the invention are disclosed in the dependent claims.

According to one aspect of the invention, the granularity of the queue speeds is improved by introducing intermediate speeds in addition to the speeds that are traditionally provided. When an intermediate speed between a relatively higher first queue rate (eg 10 Mbps) and a second lower queue rate (eg 5 Mbps) is desired for the traffic flow, the skip interval in the traffic shaper is Calculate a skip interval value according to a second lower rate (eg, 5 Mbps). This can typically result in a queue rate that matches the second rate (eg, 5 Mbps). However, according to one embodiment of the present invention, each PDU of a particular traffic flow is subject to additional override-interval-calculation decisions, and instead of skipped interval values calculated for the PDUs in the traffic shaper, a skip zero queue The interval is applied to the PDU with a predetermined probability, preferably at approximately 50%.

According to another aspect of the invention, the latency and timing problems encountered in the prior art are mitigated, whereby the PDU length is determined (in terms of the number of blocks), and the PDU with the queue interval already read from the queue. Is calculated for the first block of and written to the interval register. As a result, before the last block of the PDU is read from the queue and sent, the spacing setting in the register is performed (after a unique wait time).

These aspects are applicable to both QoS traffic and CoS traffic, allowing three different packet CoS traffic types and five different ATM QoS traffic types per port to be replaced and scheduled at the same time.

In the following, the invention will be described in more detail with reference to the accompanying drawings, in connection with preferred embodiments.

1 shows an example of an xDSL system.

2 illustrates one way of embodying the DSLAM.

3 illustrates an example of scheduling components of a routing switch processor (RSP) in accordance with an embodiment of the present invention.

4 is a flowchart illustrating one aspect of the present invention.

5 is a flowchart illustrating additional aspects of the present invention.

Although the present invention will be described below in connection with a DSL access system, this does not limit the application of the present invention to other access and communication systems.

The functional blocks shown in FIG. 2 may be distributed among different parts of the network processor in many alternative ways. For example, the Agere Payload Plus network processor includes three main units, a fast pattern processor (FPP), an Age system interface (ASI), and a routing switch processor (RSP). Classification, mark and policy 21 may be located in FPP and / or ASI, traffic management 22, queues 23, traffic shaping 24, and stream editor 25, for example, RSP It can be located at.

In general, classification identifies a PDU, policyization ensures conformance, and scheduling provides adequate service for queues, so that traffic can be networked in an appropriate manner with respect to class priorities. Can be presented. Since ATM as a connection-based protocol is different from packet-based protocols such as IP, traffic management also depends on the protocol. In ATM QoS, strict performance values can be guaranteed, while IP CoS only allows the specification of relative priority packets. This makes it impossible to specify other drop precedence for packets of constant priority.

The preferred embodiment of the present invention supports four ATM quality of service classes (ie, CBR, rt-VBR, nrt-VBR, and UBR) and three IP class service classes (ie, EF, AF, and DF). For example, services that require CBR in ATM will require EF if the traffic is IP flow. In preferred embodiments of the present invention, traffic is handled with respect to traffic parameters and priority levels regardless of whether it is ATM or IP, thus providing the same level of quality of service regardless of protocol.

Traffic policy means that active processes monitor and enforce the rules specified in the traffic contract. The policying process can be performed in the same way for both ATM and IP traffic. Incoming packets are marked with identifiers that describe their speed and priority. Traffic policy ensures that traffic received on the network does not deviate from the traffic descriptors set for the connection. Traffic policy marks PDUs for internal use based on, for example, Generic Cell Rate Algorithm (GCRA) for cells or TrTCM for packets.

In the traffic management computational engine 22, PDUs are queued or discarded based on policy results (mark) and the degree of fullness of the destination queue. If the PDU is not discarded, the PDU is added to the appropriate queue 23 for scheduling. The appropriate queue 23 is determined based on the information obtained from the classification and policy stages 21. Queue 23 is programmed to process PDUs in different ways, i.e., to specify different quality of service (QoS) and class of service (CoS) queues.

3 illustrates an example of scheduling components of a routing switch processor (RSP) in accordance with an embodiment of the present invention. RSP output port 33 is a physical interface for transmitting traffic. The RSP output port can be, for example, an xDSL interface to a subscriber connection. Port manager (PM) 32 defines the speed for one or more logical ports 34. In the example of FIG. 3, there are five port managers 32 and respective port manager blocks in each RSP unit RSP 1 to 4. Logical port 34 defines the connection for the individual traffic flow. Each logical port 34 consists of a table of schedulers. Each scheduler 35, 36, and 37 supports traffic of a particular type (s). In FIG. 3, three schedulers are configured, a first dynamic scheduler 35, a second dynamic scheduler 36, and a first-in first-out (FIFO) scheduler 37. Each scheduler maintains a table of queues. For the first and second dynamic (eg VBR) schedulers 35 and 36, and also for the FIFO scheduler 37, the table of cues is a time-slot table for assigning time slots to the queues. Can be. The first and second dynamic schedulers 35 and 36 dynamically allocate time-slots based on script programming of the traffic shaper computation engine 24, which may include traffic characteristics and RSP status. Traffic that does not require shaping, such as UBR traffic, is scheduled using the FIFO scheduler 37.

In the example of FIG. 3, five queues are configured for QoS and CoS type data flows. The first queue with queue ID Q1D1 is configured for CBR traffic flows. The second queue Q1D2 is configured for VBRrt and EF traffic flows. The third queue Q1D3 is configured for VBRnrt with w / o tag traffic flows and AF41 traffic flows. The fourth queue Q1D4 is configured for VBRnrt with w tag traffic flows. The fifth queue Q1D5 is configured for UBR and DF traffic flows.

Queues Q1D1 and Q1D2 are assigned to the first dynamic scheduler 35, queues Q1D3 are assigned to the second dynamic scheduler 36, and queues Q1D4 and Q1D5 are assigned to the FIFO scheduler 37.

In the examples described herein, the basic functionality of dynamic scheduling is performed through the Queue-64 / 48 Bytes Data Block-Time Slot Table-Queue Interval Elements and by the Main Scheduling Time Wheel. One data block of 64 or 48 bytes may be sent to a logical port in one time slot. Thus, the PDUs of the queues are scheduled in blocks of 64 or 48 bytes. Each time the RSP prepares to send another block, a global rate clock is fired, which defines the time-slot configuration. Each time the RSP prepares to send a block of data. The scheduling process begins with the selection of port manager 32 and then selects logical port 34. Thereafter, selection of the scheduler 35, 36 or 37 is performed. Typically, schedulers have a priority selected. In the example described above herein, the first dynamic scheduler 35 has the highest priority, and the FIFO scheduler 37 has the lowest priority. Finally, schedulers 35-37 have time-slot tables that define how time-slots are assigned to the queues of each scheduler.

The queue interval is determined dynamically for each queue. The queue interval defines the number of main scheduling time slots to be skipped before subsequent blocks or cells are sent from a particular queue. Thus, the interval is related to the data rate of the queue. An interval value of zero means that the number of time slots to be skipped is zero, and if the queue is not empty, then also blocks are sent from the same queue in subsequent time slots. This results in 100% bandwidth associated with logical port speed when the PDU size is smaller than one block. Similarly, the interval value day implies the skip of one time slot, and the traffic shaper CE 24 then transmits from another queue with the next highest priority before sending back from the first queue. This results in 50% speed of the port speed. In principle, velocity corresponds to the basic equation (y = 1/2 ^X ), where y is velocity and x is the interval value. Assume that the port speed is 10 Mbps. Interval values (0, 1, 2. 3. 4. 5. 6. 7, 8, and 9) are then respectively 10 Mbps, 5 Mbps, 3.33 Mbps, 2.5 Mbps, 2 Mbps, 1.7 Mbps, 1.3 Mbps, 1.1 Mbps Will generate queue speeds of, and 1 Mbps. This granularity does not allow for very precise speeds, especially for tuning the queue to higher speeds for port speeds.

The traffic manager TM and traffic shaper TS are controlled via programmable scripts that are executed for each PDU in the case of TM and for each block in the case of TS. Only one script can be used per scheduler. If the scheduler must handle ATM QoS and packet CoS traffic at the same time, a new kind of script and shaping method is needed. Traditional discrete ATM and packet traffic shaping scripts are not applicable for this purpose.

According to one embodiment of the invention, the granularity of the cue speed is improved by introducing intermediate speeds in addition to those conventionally provided. When a medium rate between a relatively higher first queue rate R1 (eg 10 Mbps) and a second lower queue rate R1 / N (eg 5 Mbps) is desired for the traffic flow. The skip interval in the traffic shaper is configured to calculate the skip interval value according to the second lower rate (eg, 5 Mbps). This will typically generate a queue rate that matches the second rate (5 Mbps). However, according to one embodiment of the present invention, each PDU of a particular traffic flow is subject to an additional alternate-interval-calculated decision, instead of the skip interval value calculated in the traffic shaper, the PDU is at a predetermined probability, preferably approximately A 50% probability is marked to receive skip zero scheduling. In other words, the traffic shaper scheduler uses a 50% probability of interval value zero for the PDU instead of the calculated interval value. Statistically, this has the effect of adding an intermediate cue bit rate value between the first queue rate and the second queue rate. If the first and second queue rates are 10 Mbps and 5 Mbps, when approximately 50% probability is used, the intermediate queue rate obtained is approximately 7.5 Mbps on average. Similarly, an intermediate queue rate may be provided between any two conventional queue rates.

In a preferred embodiment, the PDU is marked with a substitution-interval calculation decision at the classification or policy stage or at the traffic management computation engine. This is beneficial because the operation only needs to be performed once for each PDU. The result of the substitution-interval calculation decision is passed to a traffic shaper with a particular interval substitution flag that can be formed by any parameter, bit (s), byte (s), or any other suitable data structure. The alternate-interval flag may adopt a first state such as TRUE or 1 and a second state such as FALSE or 0. Any suitable random algorithm can be used to set the flag, so that the two states are shown to have the same probability. An example of a traffic manager script including instructions for providing a random 0/1 flag bit to the traffic shaper is shown in Appendix 1.

The traffic shaper may calculate the queue interval for the PDU traditionally or in accordance with the new inventive concept described below (step 41 of FIG. 4). Regardless of how the interval is calculated, the traffic shaper checks the state of the interval-alternate flag (steps 42 and 43). If the state is FALSE / 0, the traffic shaper writes a zero skip interval value in the interval register (step 45). If the state is TRUE / 1, the calculated interval value is written to the interval register (step 44). An example of a traffic shaper script including instructions for checking a random 0/1 flag bit and writing to the interval register accordingly is shown in Appendix 2. As a result, due to the presence of the interval "0.5" (50% for 0 and 1 respectively), 75% effective bandwidth is obtained. If another probability is used instead of 50%, the effective bandwidth is changed accordingly.

In the prior art, the queue interval is calculated and written to the interval register only after all blocks of the PDU have been read and counted from the queue. Before the new register's interval register is executed, there is an inherent wait time for the duration of the four blocks. Thus, after the actual PDU, four blocks from subsequent PDUs in the queue can be read and transmitted. This unwanted event distorts the timing of the traffic shaper computation engine, thus affecting other queues on the same logical port.

According to a further aspect of the present invention, these problems of the prior art are mitigated, so that the PDU length is determined (in terms of the number of blocks) (step 52 of FIG. 5), and the queue interval is calculated (step 53). Is written to the interval register for the first block (step 51) of the PDU already read from the queue (step 54). Subsequent blocks are read from the queue (step 55). As a result, before the last block of the PDU is read from the queue and sent, the spacing of the registers is performed (after four block wait times). The PDU length can be read from the TS register provided by the CE hardware.

In one embodiment applying this additional aspect of the present invention, the base interval is first calculated by software using double math, based on the effective bandwidth provided for queue speed versus port speed. This base interval is multiplied by 64 to allow greater accuracy in any additional (integer) calculations. The value 64 is chosen because it corresponds to the bit operation “left shift 6”, and bit shifting is the only execution method for performing math in traffic shaper TS scripts. This default interval is then provided as the TS parameter for the queue to be read each time the TS script is executed (for each 64 byte block).

Preferably, traffic shaping and interval calculation also consider the scheduled PDU length. In one embodiment of the invention, this is done in a TS script, whereby the PDU size changes the result of the interval multiplication (ie, length interval = fundamental interval * 2 ^{pduLengthShiftLeft} ) by the corresponding left shift count. The PDU length can be read from the TS register provided by the CE hardware. Therefore, the PDU size affects the number of left shifts in the following manner.

PDU Size Left Shift

<= 64 0

513-102 1

513-103 2

513-104 3

513-105 4

1025> 5

The actual interval value is calculated by dividing the length interval by 64, ie by the bit operation "right shift 6", thus excluding the original 64 multiplication performed by the software for the base interval. This interval value can be written to the interval register as described above. An example of a TS script that performs these steps is illustrated in Appendix 2.

This additional aspect of the invention greatly flattens the impact of different PDU sizes on the data flows of different queues and, as a related issue, affects the effect of four or so block time latencies associated with interval register processing in the RSP scheduling mechanism. Decrease.

In addition, with respect to a further aspect of the present invention, the intermediate velocity may be achieved by randomly replacing the interval value calculated as described above with respect to the first aspect of the present invention with a zero skip interval value. Thus, after the length interval is divided by 64, and before writing the interval into the TS register, it is checked whether the calculated interval value or zero is written. This may be determined by the specific bit or flag passed from the TM to the TS for all blocks of this particular PDU. As mentioned above, the TM uses a random function for bit computation, and the bits are set with a 50% probability. An example of a TS script that performs these steps is illustrated in Appendix 2.

As the technology evolves, it will be apparent to those skilled in the art that the basic idea of the invention may be implemented in a variety of ways. Accordingly, the invention and its embodiments are not limited to the examples described above but may vary within the scope of the claims.

Claims (21)

Traffic scheduling and shaping method, Buffering data blocks of packet data units (PDUs) of the traffic flow in the queue for subsequent transmission to a logical output port having a port rate R1; Reading data blocks of a PDU from the queue to a scheduler; Calculating a queue interval for the data blocks of the PDU based on the length and rate (R1 / N) of the PDU, where N is an integer equal to or greater than 1 and the queue interval is after the current PDU Indicates the number of time slots to be skipped before reading subsequent data blocks from the queue at; And Effective intermediate queue between speed R1 and speed R1 / N, by randomly overriding the calculated queue interval with a predetermined shorter or zero queue interval, with a predetermined probability, preferably at approximately 50%. Providing a rate of traffic scheduling and shaping. The method of claim 1, Generating a substitute-interval flag for the PDU at a prior classification, mark, policy or traffic management stage of the network processor, wherein the flag has a 50% probability of randomly adopting a first state; And Passing the substitute-interval flag to the scheduler; Wherein the step of randomly replacing includes checking the substitute-interval flag in the scheduler, and if the substitute interval flag is in the first state, replacing the calculated queue interval with the predetermined shorter or Substituting zero queue intervals, and if not, using the calculated queue intervals. 3. A method according to claim 1 or 2, wherein R1 = 10 Mbps, N = 2, R1 / N = 5 Mbps, and the intermediate queue rate is approximately 7.5 Mbps. As a traffic scheduling and shaping method, Buffering data blocks of packet data units (PDUs) of traffic flow in the queue for subsequent transmission to a logical output port having a port rate (R1); Reading a first data block of a PDU from the queue to a scheduler; Determining a length of the PDU and calculating a queue interval for the first data block of the PDU based on the length and speed (R1 / N) of the PDU, where N is equal to or greater than 1 A large integer, wherein the queue interval represents the number of time slots to be skipped after reading a current data block from the queue after a current PDU; And And reading a subsequent block of data of the PDU from the queue. As a traffic scheduling and shaping method, Buffering data blocks of packet data units (PDUs) of at least two traffic flows in at least two queues for subsequent transmission to a logical output port having a port rate (R1); Allocating at least two dynamic schedulers for the logical output port; When ready to transmit another data block to the output port, selecting one of the schedulers having the highest priority and the traffic to be transmitted; Selecting a subsequent queue to be serviced between the selected scheduler set of queues; Reading a first block of data from the subsequent PDU waiting in the selected queue to the selected scheduler; Determining a length of the PDU and calculating a queue interval for the first data block of the PDU based on the determined length and speed (R1 / N) of the PDU, where N equals to 1 or A larger integer, where the queue interval represents the number of time slots to be skipped after reading the next data blocks from the same queue after the current PDU; With a predetermined probability, preferably at approximately 50%, randomly replacing the calculated queue interval with a predetermined shorter or zero queue interval, thereby providing an effective intermediate queue rate between speed R1 and speed R1 / N. Making a step; And Reading subsequent data blocks of the PDU from the selected queue. Means for buffering data blocks of packet data units (PDUs) of traffic flow in a queue for subsequent transmission to a logical output port having a port rate (R1); Means for reading PDU data blocks from the queue to a scheduler; Means for calculating a queue interval for the data blocks of the PDU based on the length and rate (R1 / N) of the PDU, where N is an integer equal to or greater than 1 and the queue interval is the current PDU Then indicates the number of time slots to be skipped before reading subsequent data blocks from the same queue; And Including a means for providing an effective intermediate queue rate between rate R1 and rate R1 / N by randomly replacing the calculated queue interval with a predetermined shorter or zero queue interval with a 50% probability. Characterized by a network device. The method of claim 6, Means for generating a substitute-interval flag for the PDU at a prior classification, mark, policy or traffic management stage of the network processor, wherein the flag randomizes the first state with a predetermined probability, preferably approximately 50%. To adopt; And Means for passing the substitute-interval flag to the scheduler, Wherein the alternate means includes means for checking the alternate-interval flag in the scheduler and, if the alternate-interval flag is in the first state, the calculated queue interval to the predetermined shorter or zero queue. Means for replacing with an interval and, if not, means for utilizing the calculated queue interval. 8. A network device according to claim 6 or 7, wherein R1 = 10 Mbps, N = 2, R1 / N = 5 Mbps, and the intermediate queue rate is approximately 7.5 Mbps. The method according to any one of claims 6 to 8, At least two queue buffers buffering data blocks of packet data units (PDUs) of at least two traffic flows in at least two queues for subsequent transmission to a logical output port having a port rate R1; At least two schedulers, where at least one of the queues is assigned to each scheduler; And A main scheduling control for selecting one of the schedulers with the highest priority and the traffic to be transmitted when ready to send another data block to the output port, Wherein the selected scheduler selects a subsequent queue to be serviced in the at least one queue assigned to the scheduler and reads data blocks of subsequent PDUs waiting in the selected queue. The method of claim 9, A first queue for first quality of service (QoS) type traffic; A second queue for second QoS type traffic and first class of service (CoS) type traffic; A third queue for third QoS type traffic and second CoS type traffic; A fourth queue for fourth QoS type traffic; A fifth queue for fifth QoS type traffic and third CoS type traffic; A first dynamic scheduler for the first and second queues; A second dynamic scheduler for the third queues; And A first-in first-out (FIFO) scheduler for the fourth and fifth queues. The method of claim 10, The first QoS type traffic includes fixed bit rate (CBR) type traffic, The second QoS type traffic includes variable bit rate non-real time (VBR-nrt) type traffic, The third QoS type traffic includes variable bit rate non-real time (VBR-nrt) w / o tag type traffic, The fourth QoS type traffic includes variable bit rate non-real time (VBR-nrt) w tag type traffic, And the fifth QoS type traffic comprises non-specific bit rate (UBR) type traffic. The method according to claim 10 or 11, wherein The first CoS type traffic includes Expedited Forwarding (EF) type traffic, The second CoS type traffic includes Assured Forwarding (AF) type traffic, And the third CoS type traffic comprises Default Forwarding (DF) type traffic. At least two queue buffers buffering data blocks of packet data units (PDUs) of at least two traffic flows in at least two queues for subsequent transmission to a logical output port having a port rate R1; At least two schedulers, where at least one of the queues is assigned to each scheduler; Main scheduling control for selecting one of the schedulers having the highest priority and the traffic to be transmitted when ready to send another data block to the output port; Wherein the selected scheduler selects a subsequent queue to be serviced between the at least one queue assigned to the scheduler and reads a first data block of a subsequent PDU waiting in the selected queue, Calculation means for calculating on the first data block of the PDU based on the length and rate (R1 / N) of the PDU, where N is an integer equal to or greater than 1 and the queue interval is after the current PDU Indicate the number of time slots to be skipped before reading subsequent data blocks from the queue; And With a predetermined probability, preferably at approximately 50%, randomly replacing the calculated queue interval with a predetermined shorter or zero queue interval, thereby providing an effective intermediate queue rate between speed R1 and speed R1 / N. Network comprising a replacement means. The method of claim 13, A first queue for first quality of service (QoS) type traffic; A second queue for second QoS type traffic and first class of service (CoS) type traffic; A third queue for third QoS type traffic and second CoS type traffic; A fourth queue for fourth QoS type traffic; A fifth queue for fifth QoS type traffic and third CoS type traffic; A first dynamic scheduler for the first and second queues; A second dynamic scheduler for the third queue; And A first-in first-out (FIFO) scheduler for the fourth and fifth queues. The method of claim 14, The first QoS type traffic includes fixed bit rate (CBR) type traffic, The second QoS type traffic includes variable bit rate non-real time (VBR-nrt) type traffic, The third QoS type traffic includes variable bit rate non-real time (VBR-nrt) w / o tag type traffic, The fourth QoS type traffic includes variable bit rate non-real time (VBR-nrt) w tag type traffic, And the fifth QoS type traffic comprises non-specific bit rate (UBR) type traffic. The method according to claim 14 or 15, The first CoS type traffic includes Expedited Forwarding (EF) type traffic, The second CoS type traffic includes Assured Forwarding (AF) type traffic, And the third CoS type traffic comprises Default Forwarding (DF) type traffic. The method according to any one of claims 14 to 16, The calculating means is provided in at least one of the schedulers, The calculation means is: Means for calculating a base interval based on the available bandwidth provided for the queue rate versus the port rate; And And bit shifting means for calculating the queue interval based on the basic interval and the PDU length. The method according to any one of claims 14 to 17, And means for providing an effective intermediate queue rate between rate R1 and rate R1 / N by randomly replacing the calculated queue interval with a predetermined shorter or zero queue interval with a 50% probability. Characterized by a network device. A digital subscriber line access multiplexer comprising a network device according to any one of claims 6 to 18. A program product comprising machine-readable program code executed in a processor to perform the steps according to any of claims 1 to 6. A machine-readable data storage medium comprising program code executed in a processor to perform the steps according to any of claims 1 to 6.

Images