KR20030057594A - Method and Apparatus for weighted round-robin scheduling in ATM layer - Google Patents

Abstract

PURPOSE: A weighted round-robin scheduling method in an ATM(Asynchronous Transfer Mode) layer and an apparatus therefor are provided to effectively use a memory, using a linked-list structure, when writing ATM cells in the memory. CONSTITUTION: A VC(Virtual Connection) table checking block(110) checks a weight of a VC from a VC table(120). The VC table(120) stores weight information about each VC. A linked-list structure composing block(130) generates an address signal for forming a list structure of ATM cells stored in a linked memory according to the weight of the VC. A register control block(140) controls a register block(150) according to the weight of the VC and a linked-list, and forms an address signal indicating a cell position of a memory(160). The register block(150) is operated so that a plurality of storing media are operated as an active state or a pending state according to the weight of the VC. The memory(160) stores the ATM cells.

Description

Weighted round robin scheduling method and apparatus thereof in asynchronous transport mode layer {Method and Apparatus for weighted round-robin scheduling in ATM layer}

The present invention relates to a weighted round-robin scheduler in an asynchronous transmission mode layer, and more particularly, to a variable bit rate (VBR) service, an available transmission rate. Weighted round robin scheduling method and apparatus for asynchronous transmission mode layer suitable for ATM cell scheduling of traffic class such as Available Bit Rate (ABR) service, Unspecified Bit Rate (USR) service, etc. It is about.

In general, the weighted round robin scheduler gives a different weight to each virtual connection in memory and sequentially reads the cells of the virtual connection as many times as the weight for one frame.

If all weights are the same, then this corresponds to general round robin scheduling.

Unlike the round robin scheduler, the weighted round robin scheduler controls the amount of service by giving different weights to the queue to be served. In other words, a queue with a larger weight will serve more services in one frame than a queue with a smaller weight.

Conventionally, when processing an ATM cell, a memory area is defined for each virtual connection and weighted round robin scheduling is applied.

On the other hand, if the cell processing services in ATM networks are classified, the fixed rate (CBR) service has a fixed data rate at the peak cell rate, and the variable rate service has a strong time constraint on data delay or delay variation. The peak cell rate and average rate are specified for time-varying applications, and cell loss is allowed in the difference between the peak and the mean, and the available rate service allows for cell loss, although the demand for delay is not strict. It is designed to increase or decrease the transmission speed from the terminal according to the congestion status of the nautical network in the non-application.

However, according to the related art, since a memory area is designated for each connection in advance, the memory usage efficiency is deteriorated when traffic is concentrated on a specific connection or there is no traffic.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned conventional problems, and an object of the present invention is to provide a weighted weight in an asynchronous transmission mode layer that performs ATM cell scheduling of a traffic class in a weighted round robin method in an ATM cell processing service. A method and apparatus for round robin scheduling are provided.

A weighted round robin scheduling method in an asynchronous transmission mode layer of the present invention for achieving the above object comprises the steps of setting the memory into an active flow queue and a pending flow queue, and setting the weight for each virtual connection; Activating a scheduler when a cell input to the memory no longer occurs; If the weight of the virtual connection is equal to the number of cells read in one frame when the cell is read from memory, causing the cell to enter the pending flow queue; When all of the frames are finished, it is characterized in that it comprises the step of moving all the queues stored in the pending flow queue to the active flow queue.

A weighted round robin scheduling apparatus in an asynchronous transmission mode layer of the present invention for achieving the above object comprises: a table check block for checking a weight of a virtual connection in a virtual connection table upon input of an ATM cell; A table consisting of weight information for each virtual connection; A linked list structure building block for generating an address signal for forming a list structure of ATM cells stored in a memory connected to the linked list according to the weight of the virtual connection; A register control block configured to control the register block according to the weight and linked list of the virtual connection and form an address signal indicating a cell position of the memory; A register block configured to allow the storage medium to operate in an active or pending manner according to the weight of the virtual connection; Characterized in that it comprises a storage means for storing the ATM cell.

1 is a block diagram of a weighted round robin scheduling apparatus in an asynchronous transmission mode layer according to an embodiment of the present invention.

2 is a flow chart of a weighted round robin scheduling method in an asynchronous transmission mode layer according to an embodiment of the present invention.

3 and 4 are flowcharts illustrating a flow queue enqueuing procedure among the weighted round robin scheduling methods in the asynchronous transmission mode layer according to an embodiment of the present invention.

5 and 6 are flowcharts illustrating a flow queue dequeuing procedure among the weighted round robin scheduling methods in the asynchronous transmission mode layer according to an embodiment of the present invention.

7 is a flowchart illustrating a process of moving an active flow queue to a pending flow queue in a weighted round robin scheduling method in an asynchronous transmission mode layer according to an embodiment of the present invention.

8 through 10 are block diagrams illustrating an enqueue and a round robin scheduler of a flow queue according to an embodiment of the present invention.

11-13 are block diagrams illustrating dequeue and round robin schedulers of flow queues in accordance with an embodiment of the present invention;

14 to 16 are block diagrams illustrating a procedure of moving a flow queue according to an embodiment of the present invention.

17-19 are block diagrams illustrating dequeue and round robin schedulers of flow queues in accordance with an embodiment of the present invention.

Explanation of symbols on the main parts of the drawings

110: VC table check block 120: VC table

130: linked list structure configuration block 140: register control block

150: register block 160: memory

Hereinafter, exemplary embodiments of the present invention will be described with reference to the accompanying drawings.

1 is a block diagram of a weighted round robin scheduling apparatus in an asynchronous transmission mode layer according to an embodiment of the present invention, and FIG. 2 is a diagram of a weighted round robin scheduling method in an asynchronous transmission mode layer according to an embodiment of the present invention. 3 and 4 are flowcharts illustrating a flow queue enqueuing procedure among the weighted round robin scheduling methods in the asynchronous transmission mode layer according to an embodiment of the present invention, and FIGS. 5 and 6 are embodiments of the present invention. 7 is a flowchart illustrating a flow queue dequeuing procedure among the weighted round robin scheduling methods in the asynchronous transmission mode layer according to the present invention. FIG. 7 is an active flow in the weighted round robin scheduling method in the asynchronous transmission mode layer according to an embodiment of the present invention. 8 is a flowchart illustrating a process of moving a queue to a pending flow queue, and FIGS. 8 to 10 are embodiments of the present invention. 11 is a block diagram illustrating a dequeue and round robin scheduler of a flow queue according to an embodiment of the present invention, and FIGS. 11 to 13 are block diagrams illustrating a dequeue and round robin scheduler of a flow queue according to an embodiment of the present invention. 17 is a block diagram illustrating a flow queue movement procedure according to an embodiment of the present invention, and FIGS. 17 to 19 are block diagrams illustrating a dequeue and a round robin scheduler of a flow queue according to an embodiment of the present invention.

According to FIG. 1, the application block of this embodiment includes a VC table check block 110 that checks the weight of a virtual connection in a VC table when an input ATM cell is input, and a VC table 120 including weight information for each virtual connection. ), A linked list structure construction block 130 for generating an address signal for forming a list structure of ATM cells stored on the linked memory according to the weight of the virtual connection, and controlling the register block according to the weight and the linked list of the virtual connection. And a register control block 140 configured to form an address signal indicative of a cell position of the memory, and a register block 150 in which a plurality of storage media are operated in an active or pending manner according to the weight of the virtual connection. And a memory 160 in which an ATM cell is stored.

The register block 150 includes a base address register, a register for storing information about a previous active connection (PrevConA), a register for storing information about a previous pending connection (PrevConP), and a queue length for holding information about a queue length. It is divided into registers (Queue Length), head and tail registers (Head & Tail) for configuring the linked list, and registers (NextQ) indicating the next queue of the linked list.

The memory 160 is classified into an active flow queue and a pending flow queue.

All virtual connections form their respective VC queues in the active flow queue, and each is input in a linked-list structure. Using the memory 160 as a linked list structure prevents the inefficiency of the memory 160, which can occur when an ATM cell is crowded or vacant on a particular virtual connection.

If the cell input to the memory 160 no longer occurs, the weighted round robin scheduler enters the enable mode. When a cell is read from memory, if the weight corresponding to the virtual connection is equal to the number of cells read in one frame, it enters the pending flow queue.

If one frame ends, all queues stored in the linked list structure in the pending flow queue are put back into the active flow queue.

This repetition will only serve the weight of a particular virtual connection for one frame.

When an ATM cell is input, the virtual connection table search block reads values of the head, tail, weight, queue length, next queue, and the like of the input cell. Here, the head, tail, and nextQ values are entered into the linked list structure building block, which creates a linked list structure and writes cells to memory. The weight and queue length values of the virtual connection table are input to the register control block to determine the address of the cell to be read from the memory.

The operation of the register control block will be described in more detail as follows.

According to FIG. 2, the operation is divided into a case where a new cell arrives and a case where it does not. If a new cell arrives, a flow queue enqueue procedure is performed (S210 and S220).

On the other hand, if a new cell does not arrive, it is checked whether the weighted round robin scheduler (hereinafter, abbreviated as WRR) is enabled. If the WRR is enabled, it is also checked whether a cell in the buffer exists (S230, S240).

If it is determined that a cell in the buffer exists, the WRR scheduling operation is performed (S250).

On the other hand, when a negative determination is made in step S230 or S240 or after step S220 or S250 is performed, the process returns to step S210 to repeat the subsequent process.

Here, the flow queue enqueue procedure follows the following sequence as shown in FIGS. 3 and 4.

First, when a new cell arrives, the base address is found. The base address may be found using a memory port, a virtual path identifier (VPI), a virtual connection identifier (VCI), and the like (S31 to S32).

Next, the free cell memory is read and the queue length register is read to determine whether the queue length is '0' (S33 to S35).

If the queue length is not '0', if the base address is '0', the precell memory value is written to the next queue register pointed to by the previous connection register (PrevConn Register). Write the memory value to the base address register (S37, S38).

Next, the precell memory value is written to the head and tail registers, and the VCI value is written to the previous connection (PrevConn) register before the precell memory is read. On the other hand, when the queue length is not '0' in step S35, only the free cell memory is read (S39 to S41).

After attaching the read free cell memory value to the tail part of the ATM cell and reading the tail register value, write the ATM cell to the cell memory pointed to by the tail register, increment the queue length register by 1, and update the tail register to the free cell memory value. Returning to S31, the operation from the beginning is repeated (S42 to S46).

The flow queue enqueue procedure has been described above, and then the flow queue dequeue procedure will be described with reference to FIGS. 5 and 6.

First, if the beginning of the subframe, the base address is read to read the ATM cell indicated by the base address (S51, S52, S53).

After reading the ATM cell, the queue length is reduced by 1 and the head register value is written to the free cell memory (S54 to S55).

On the other hand, if the next queue NextQ is '0' but not at the beginning of the subframe, the process returns to step S51 (S56 to S57).

In contrast, if the first subframe is not the first and the next queue is not '0', the ATM cell pointed to by the next queue register is read, the queue length is reduced by 1, and the head register value is written to the freecell memory, and the next connection before the previous connection Check whether the queue number is the base address (S58 to S61).

If the next queue of the previous connection is the base address, the base address value is updated to a value in the tail portion of the ATM cell (S62).

After updating the base address value to the value in the tail portion of the ATM cell, or if the next queue of the previous connection is not the base address, update the head register value to the value in the tail portion of the ATM cell and following the previous connection. The queue number 1 is updated to a value in the tail portion of the ATM cell (S63 to S64).

In addition, after writing the head register value to the free cell memory in step S55, the head register value is updated to a value in the tail portion of the ATM cell, and then it is determined whether the subframe count and the weight are the same. If the subframe count and weight are the same, the base address is updated to the next queue of the current connection. On the other hand, when the subframe count and the weight are not the same, the base address is updated to a value in the tail portion of the ATM cell (S65 to S68).

After step S64, S67, or S68 is performed, it is checked whether the queue length is '0' (S69).

If the queue length is not '0', when the subframe count and the weight are not equal, the value is written to the previous connection register and the flow returns to step S51. If the subframe count and the weight are the same, the next queue of the previous connection is moved to the next one after the current connection. After changing to the second queue, the process returns to step S51 after moving to the pending flow queue according to the set procedure (S70 to S73).

If the queue length is '0', the value in the tail portion of the ATM cell is written to freecell memory, the next queue of the previous connection is replaced with the next queue of the current connection, and then the head, tail, queue length, and next queue registers. After the initialization, the process returns to step S51. This initialization is the operation to set the register value to '0' (S74 to S76).

The flow queue dequeue procedure has been described above, and the following describes a procedure of moving from an active flow queue to a pending flow queue according to FIG. 7.

First, replace the next queue of the previous connection with the next queue of the current connection, and then set the next queue of the current connection to '0'. Subsequently, the total length of the queue is updated by subtracting the queue length from the total length of the queue (TotQLen) (S71 to S73).

If the total length (TotQLen_Pending) of the pending flow queue is '0', the base address of the pending flow queue is updated to the head. If the total length (TotQLen_Pending) of the pending flow queue is not '0', the next queue of the previous pending connection (PrevConn_Pend) is updated to the head (S74 to S76).

Next, the previous pending (Prev_Pending) is updated to the current connection, and the total length (TotQLen_Pend) of the pending flow queue is updated by adding the queue length to the total length (TotQLen_Pend) of the pending flow queue (S77 to S78).

For example, as shown in FIG. 8, there is an active flow queue 811 and a pending flow queue 812 in memory. Cells arriving at the cell input are sequentially displayed in the active flow queue 811. Are stored in. For example, suppose there are w, x, y, and z connections, and weights are set for these connections, such as 3, 2, 1, and 2.

According to Fig. 9, when a cell arrives at connection w1, a and b of the free cell memory are set to head and tail, respectively, and the queue length is '1', and the next queue does not exist yet. Is updated. The active flow queue includes a connection w1 and a tail address b in the free cell memory address a.

In this case, when the cell arrives at connection x4 as shown in FIG. 10, the free cell memory address becomes q and the previous active connection indicates the previously arrived connection (here, connection z). Miscellaneous Shows compliance with the procedures described above.

FIG. 11 shows movement of cells from the active flow queue to the pending flow queue while the cells of the active flow queue are output according to the flow queue dequeue procedure. Fig. 12 shows that cells are output from the active flow queue, the corresponding freecell memory is returned, and the queue length and the next queue are updated. FIG. 13 shows that the cells of connection y should be moved from the active flow queue to the pending flow queue according to the subframe count, and FIGS. 14, 15 and 16 show a connection x, It is shown that z is moved and the previous active connection and previous pending connection registers and the active queue length and pending queue length registers are updated. If all connections are moved to the pending flow queue by continuing cell output as shown in Fig. 17 and moving to connection w as shown in Fig. 18, all connection queues in the pending flow queue are started when the new round robin scheduler (WRR) is started. Demonstrating the move to the queue.

The embodiments described above are within the scope of various changes, modifications, and equivalents of the present invention. Therefore, the present invention is not limited to the description of the examples.

According to the weighted round robin scheduling method and apparatus therefor in the asynchronous transmission mode layer of the present invention, a linked list structure can be effectively used when writing an ATM cell to a memory, and a VC queue is configured for each virtual connection. Writing to active flow queues allows for fewer registers to be moved between each queue for effective memory control.

Claims (5)

Setting a memory into an active flow queue and a pending flow queue, and setting a weight for each virtual connection; Activating a scheduler when a cell input to the memory no longer occurs; If the weight of the virtual connection is equal to the number of cells read in one frame when the cell is read from memory, causing the cell to enter the pending flow queue; And when all the frames are finished, all queues stored in the pending flow queue are moved to an active flow queue. The method of claim 1, A linked list structure is created using the head, tail, weight, queue length, and next queue value of the input cell recorded in the established virtual connection table, and the input cell is recorded in the storage means, and the weight of the virtual connection table Weighted round robin scheduling method in the asynchronous transmission mode layer, characterized in that the address number of the cell to be read from the storage means is determined by using the queue length value. The method of claim 2, wherein moving the pending flow queue to an active flow queue comprises: Replacing the next queue of the previous connection with the next queue of the current connection; Resetting the next queue of the current connection to an initial value; Updating the total length of the queue through an operation of subtracting the queue length from the total length of the queue; If the total length of the pending flow queue is an initial value, updating the base address of the corresponding pending flow queue to the head; If the total length of the pending flow queue is not an initial value, updating the next queue of previous pending connections to the head; Weighted round robin in the asynchronous transport mode layer, comprising updating the previous pending connection with the current connection and updating the total length of the pending flow queue by adding the queue length to the total length of the pending flow queue. Scheduling Method. A table check block for checking a weight of a virtual connection in a virtual connection table upon input of an ATM cell; A table consisting of weight information for each virtual connection; A linked list structure building block for generating an address signal for forming a list structure of ATM cells stored in a memory connected to the linked list according to the weight of the virtual connection; A register control block adapted to control the register block according to the weight and linked list of the virtual connection and form an address signal indicating a cell position of the memory; A register block configured to allow the storage medium to operate in an active or pending manner according to the weight of the virtual connection; Weighted round robin scheduling apparatus in an asynchronous transmission mode layer, characterized in that it comprises a storage means for storing the ATM cell. The method of claim 4, wherein the register block, A base address register; A register that stores information about a previous active connection; A register that stores information about a previous pending connection; A queue length register for holding information about the queue length; A head and tail register for constructing a linked list; Weighted round robin scheduling apparatus in the asynchronous transmission mode layer, characterized in that it is divided into registers representing the next queue of the linked list.