JP2006019860A - Frame read scheduling circuit and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a circuit and a method for impartially distributing a band to each user in a frame read scheduling and bringing a read delay of a high priority connection frame to a low delay. <P>SOLUTION: The method includes: a first step of sharing received frames with each user and each class; a second step of monitoring and/or shaping the high priority class frames on the basis of preset information and aborting the frames in excess of a setting permissible amount; a third step of reading the frames of each class of each user shared to a low priority class by a preset scheduling method; and a fourth step of reading the low priority class frames and the frames shared to the high priority class, and the method stops reading of the read high priority class frame by the users for a time equivalent to the length of the high priority class frames read in the fourth step. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a frame read scheduling circuit and method for determining a frame read order on a connection basis when a plurality of connections are concentrated or multiplexed in a switch on a network.

  When a plurality of connections are concentrated or multiplexed in a switch that handles Ethernet (registered trademark) frames on the network, fair bandwidth distribution is performed for each user, and each connection within the user is In some cases, the read delay time is differentiated according to the set priority.

  In order to perform fair bandwidth distribution among users, for example, a method of giving a token that is a right to read a frame to each connection by using fair read control such as WFQ (Weighted Fair Queuing) and WRR (Weighted Round Robin). There is. For example, Non-Patent Document 1 describes a method of separately preparing a FIFO buffer for a plurality of connections and reading a variable-length frame stored in the buffer.

  Further, as a method of setting a priority related to a frame read delay time for each connection and performing a read earlier for a frame of a higher priority connection, for example, priority read control based on the priority is available.

  In order to achieve both the fair read control of the bandwidth distributed among users and the priority read control among a plurality of connections within the user, the in-user frame priority read unit 14 that performs the priority read control as shown in FIG. This can be realized by combining two control units of the inter-user frame fair reading unit 15 that performs fair read control.

Yukihiro Yoshida, Yasunobu Kasahara, “Examination of bandwidth fair allocation function in Ether demultiplexer”, Shingaku Techniques, CS2002-103, November 2002

  However, when the above method is used, it is possible to preferentially read the frame of the high priority connection with respect to the frame of the low priority connection within the user, but give priority to the frame of the low priority connection of a different user. It cannot be read. For this reason, when the number of users accommodated in the switch increases, a situation occurs in which the frame of the high priority connection cannot be read for a long time due to the frame reading of the low priority connection of another user. For example, in a switch having a reading speed of 1 Gbit / s, when 5000 users are concentrated and other users are simultaneously sending frames of a low priority connection of 1500 bytes, the frame of the high priority connection is a maximum of 60 ms. It will wait for reading. In this case, traffic that can only tolerate a delay of about several ms cannot be accommodated.

  Therefore, according to the present invention, even when a large number of users are accommodated in the frame readout scheduling, the bandwidth allocated to each user is distributed fairly based on the preset weight, and the readout delay of the high priority connection frame is reduced. It is an object of the present invention to provide a circuit and a method that have a lower delay than conventional methods.

  In addition to the above object, another object of the present invention is to prevent band breakage and increase band use efficiency.

  It is another object of the present invention to provide a method and a circuit for ensuring quality when accommodating a large number of users and establishing communication dynamically for communication.

According to the frame readout scheduling method of the present invention,
To accommodate one or more users who use one or more connections, determine the bandwidth to be allocated to each connection based on the priority set for each connection and the bandwidth distribution ratio set for each user, and for frame reading In the first scheduling method, the input frame is sorted for each user and each class by referring to information assigned to the input frame and / or a preset attribute for the connection. A second step of monitoring and / or shaping the frames allocated to the high priority class based on information set in advance and discarding frames exceeding the set allowable amount; The frames for each user class assigned to the priority class are determined according to a preset scheduling method. A third step of reading, and a fourth step of reading out the low priority class frame read in the third step and the frame allocated to the high priority class by a preset scheduling method In the frame reading in the third step, reading of the user of the read high priority class frame is stopped for a time corresponding to the high priority class frame length read in the fourth step, or the fourth step In this step, the entire third step is stopped for a time corresponding to the read high priority class frame length.

According to another embodiment of the frame readout scheduling method of the present invention,
The frame read in the third step is also preferably stored in a buffer.

According to another embodiment of the frame readout scheduling method of the present invention,
It is also preferable to suppress or accelerate the reading operation in the third step according to the number of frames accumulated in the buffer.

Furthermore, according to another embodiment of the frame readout scheduling method of the present invention,
In the second step, it is also preferable to perform high priority class frame monitoring, shaping, and discarding control by external call control.

According to the frame readout scheduling circuit of the present invention,
To accommodate one or more users who use one or more connections, determine the bandwidth to be allocated to each connection based on the priority set for each connection and the bandwidth distribution ratio set for each user, and for frame reading In the circuit that performs the scheduling, the frame allocation that distributes the input frame for each user and each class with reference to the information given to the input frame and / or the attribute set in advance for the connection. The high-priority class frame distributed by the distribution means and the frame distribution means is input, and the input frame is monitored and / or shaped based on preset information, exceeding the set allowable amount A traffic monitoring / shaping unit for each user that discards a frame, and the frame distribution unit The divided low priority class frames are input, and the input frames are stored in a FIFO buffer prepared for each class, and stored in the user individual frame storage means for each user and the user individual frame storage means. The frames from the low priority class frame reading means for each user, and the traffic monitoring / shaping means of each user, which are sequentially read out in accordance with a preset scheduling method, are input to the frames. Inter-user low-priority class frame that sequentially reads frames by means of a preset scheduling method from a shared frame accumulation unit that accumulates in a FIFO-type buffer prepared for each class and the intra-user low-priority class frame reading unit of each user Read means and the shared frame storage means A virtual frame insertion unit that monitors the length of each frame to be generated, generates a virtual frame for each user from the acquired frame length information, and inserts the generated virtual frame into the inter-user low priority class frame reading unit And inter-class frame reading means for reading frames by a preset scheduling method from the shared frame storage means and the inter-user low priority class frame reading means, and the shared frame storage means and / or the inter-class frame reading means Frame reading inhibiting means for inhibiting the reading operation of the inter-user low priority class frame reading means based on the traffic information obtained from the above.

According to another embodiment of the frame readout scheduling circuit of the present invention,
The inter-user low-priority class frame reading unit includes a read reservation frame storage unit that stores a frame read, and the inter-class frame reading unit is preset by the shared frame storage unit and the read reservation frame storage unit. It is also preferable to read out the frame by the scheduling method.

According to another embodiment of the frame readout scheduling circuit of the present invention,
The read reservation frame number monitoring means for monitoring the number of frames stored in the read reservation frame storage means and the reading operation of the inter-user low priority class frame reading means are accelerated by an instruction from the read reservation frame number monitoring means. It is preferable that the frame reading acceleration means suppresses the reading operation of the inter-user low priority class frame reading means even in response to an instruction from the read reserved frame number monitoring means.

Furthermore, according to another embodiment of the frame readout scheduling circuit of the present invention,
It is also preferable to have a connection establishment control means for determining whether or not the high priority class frame can be controlled by call control from the outside and instructing the traffic monitoring / shaping means.

  Even in the case of accommodating a large number of users in frame readout scheduling, the bandwidth allocated to each user is distributed fairly based on preset weights, and the readout delay of high-priority connection frames is reduced compared to the conventional method. This has the effect of reducing both.

  In addition, it is possible to prevent the occurrence of the band splitting because the determination of the read target user is not in time, and to increase the band use efficiency.

  The quality can be ensured even when a large number of users are accommodated and the connection is dynamically established and communication is performed while the total bandwidth of the high priority class traffic is limited to a certain level.

  The best mode for carrying out the present invention will be described in detail below with reference to the drawings.

  (First Embodiment) In this embodiment, the types of connections are classified into a high priority class that requires a short read delay time and a low priority class other than that. Unlike the technology, it is not subject to fair read control, but is directly subject to priority read control, thereby shortening the read delay time of the high priority class frame. Further, the present invention is characterized in that the fairness of bandwidth distribution among users is ensured by reflecting the passing amount of the high priority class frame in the fair read control for the low priority class frame.

  FIG. 2 is a diagram showing a first embodiment of a frame readout scheduling circuit according to the present invention. According to FIG. 2, the frame read scheduling circuit includes a frame distribution unit 1, a traffic monitoring / control unit 2, a user individual frame storage unit 3, an intra-user low priority class frame read unit 4, and a shared frame storage unit 5. An inter-user low-priority class frame reading unit 6, a virtual frame insertion unit 7, an inter-class frame reading unit 8, and a frame reading inhibiting unit 9.

  The frame distribution unit 1 distributes the frame for each user and each class with reference to information given to the frame input to the frame read scheduling circuit and / or an attribute set in advance for the connection. Classes are classified into high priority classes and low priority classes. The high priority class frame is input to the traffic monitoring / shaping unit 2 for each user, and the low priority class frame is input to the user individual frame storage unit 3 for each user. A plurality of high priority classes and low priority classes can be set for each user.

  The traffic monitoring / shaping unit 2 performs monitoring (policing) and / or shaping (shaping) on the input high-priority class frame based on preset traffic setting information. When an excessive frame exceeding the set allowable amount is input, the excessive amount is discarded.

  The user individual frame accumulating unit 3 has an individual FIFO buffer provided for each class, and temporarily accumulates the input low priority class frames.

  When there are a plurality of low priority classes in the user, the user low priority class frame reading unit 4 arranged after the user individual frame storage unit 3 sequentially reads the frames from the user individual frame storage unit 3. Then, concentration or multiplexing is performed. For the read control from the user individual frame storage unit 3 performed by the low priority class frame reading unit 4 within the user, priority read control or fair read control is used according to the QoS (Quality of Service) policy of the class.

  The shared frame storage unit 5 has a FIFO buffer, and temporarily stores the high priority class frames from the traffic monitoring / shaping unit 2 of each user.

  The inter-user low-priority class frame reading unit 6 reads a frame from the intra-user low-priority class frame reading unit 4 of each user by fair read control based on a preset bandwidth distribution ratio.

  The virtual frame insertion unit 7 monitors the length of the frame stored in the shared frame storage unit 5 for each user, and generates a virtual frame based on the acquired frame length information. In the generated virtual frame, the inter-user low-priority class is given priority over the frame from the intra-user low-priority class frame reading unit 4 at the timing when the frame that is the source of the virtual frame is read by the inter-class frame reading unit 8. It is inserted into the frame reading unit 6.

  The interclass frame reading unit 8 reads frames from the shared frame accumulating unit 5 and the low priority class frame reading unit 6 between users by priority read control or fair read control.

  The frame reading inhibiting unit 9 stops or inhibits the operation of the inter-user low priority class frame reading unit 6 based on the traffic information acquired from the shared frame accumulating unit 5 and / or the interclass frame reading unit 8. The traffic information includes the length of the high-priority class frame that passes through the shared frame storage unit 5 or the amount when the inter-class frame reading unit 8 reduces the output bandwidth.

  The relationship between the bandwidth allocated to the user by the frame readout scheduling circuit and the priority class will be described. FIG. 3 is a diagram illustrating an example of bandwidth distribution in a case where the bandwidth of all classes is fairly distributed among users. In this distribution example, a band distribution ratio is determined for each user, and a distribution band to each user is determined based on the band distribution ratio. Then, the bandwidth allocated to each user is assigned with priority over the high priority class, and the remaining bandwidth is assigned to the low priority class. When the maximum bandwidth is set in the high priority class, the excess part of the higher priority class traffic exceeding this is discarded.

  The band distribution method is preferably applied when a user contracts a line including a plurality of classes as a single service.

  FIG. 4 is a flowchart illustrating an operation example of the inter-user low priority class frame reading unit 6 when the bandwidth of all classes is fairly distributed among users. In FIG. 4, T is the token length (bytes) given at one time, M is the minimum number of bytes that allows the next search to be advanced, XHi is the number of read bytes of the high priority class of user #i, and XLi is It is the number of read excess bytes of the low priority class of user #i, and TC is the remaining token length that can be used. Here, the number of read bytes XHi of the high-priority class is for reflecting the use band of the high-priority class in the operation of the inter-user low-priority class frame reading unit 6. Specifically, the virtual frame insertion unit 7 By this, the inter-user low priority class frame reading unit 6 is notified. Further, the number of excessive read bytes XLi of the low priority class represents a portion of the frames read up to the previous time that exceeds the given token length.

  First, a user to be read next is determined by scheduling means such as WRR, and a token having a fixed length of T bytes is given. When user #i is selected, the flow shown in FIG. 4 is executed.

  (S401) It is confirmed that there is a frame to be read from the in-user low priority class frame reading unit 4 of the user #i.

  (S402) When there is no frame to be read, the sum of the number of read bytes XHi of the high priority class of user #i and the number of read excess bytes XLi of the low priority class of user #i is the minimum that allows the next search to be advanced Confirm that the number of bytes is larger than M. Since it takes a certain time to search for the next frame to be read, M must be larger than the number of bytes that can be read during the longest search time.

  (S403) The token length T is substituted for the remaining remaining token length TC.

  (S404, S405, S406) TC and XHi are compared. If TC is less than or equal to XHi, the token given to the user is depleted by subtracting XHi, and the process ends. If TC is greater than XHi, XHi is subtracted from TC. As a result, XHi becomes zero.

  (S407, S408, S409) TC and XLi are compared, and when TC is less than or equal to XLi, the token given to the user is depleted by subtracting XLi, and the process ends. When TC is larger than XLi, XLi is subtracted from TC. As a result, XLi becomes zero.

  (S410, S411) It is confirmed that there is a read target frame from the low priority class frame reading unit 4 for the user #i, and if there is a read target frame, the length of the read target frame is substituted for L. Read out the frame.

  (S412, S413, S414) TC and L are compared. If L is less than TC, L is subtracted from TC, and the process returns to S410. If L is equal to or greater than TC, the value obtained by subtracting TC from L is substituted into XLi as the number of read excess bytes in the low priority class.

  (S415) In S410, when there is no frame to be read by user #i, the timing of the next search is advanced by the read time corresponding to TC.

  By executing the above flow, it is possible to realize fair bandwidth distribution among users including the use bandwidth of the high priority class that does not pass through the inter-user low priority class frame reading unit 6 shown in FIG.

  In addition, the flow can be simplified by adding some restrictions to the flow of FIG. FIG. 5 is a flowchart illustrating a simplified operation example in the inter-user low priority class frame reading unit 6 when the bandwidth of all classes is fairly distributed among users. Here, in the flow of FIG. 4, the token length T given at one time is the same as the minimum frame length, and the minimum number of bytes M that allows the next search to be carried forward is the same as T.

  Compared with FIG. 4, S410, S412, S413, and S415 can be deleted. Since it is not necessary to advance the search time in S415 in FIG. 4, the search cycle can be made constant, and the mounting is easy.

  Next, let us consider a case where the use band of the high priority class is not reflected in the inter-user low priority class frame reading unit 6. FIG. 6 is a diagram illustrating an example of bandwidth distribution in the case where the low-priority class bandwidth is fairly distributed among users. In this distribution example, after assigning the high-priority class band of each user, the remaining band is set as the low-priority class band, and the band to be distributed among users is determined based on the band distribution ratio. Similar to the distribution example in FIG. 3, when the high priority class traffic exceeds the maximum bandwidth set for each user, the excess portion of the frame is discarded. Therefore, the total of the high priority class traffic does not exceed the maximum bandwidth.

  This band distribution method is desirably applied when a user contracts a high priority class line and a low priority class line independently.

  When the inter-user low-priority class frame reading unit 6 ensures fairness only in the low-priority class band, the distribution band is determined from all the distributable bands so as not to exceed the read band. It is necessary to subtract the bandwidth used in the high priority class. For this reason, when reading a frame in the high priority class, the frame reading suppression unit 9 stops the operation of the inter-user low priority class frame reading unit 6 for the time required for reading.

  FIG. 7 is a flowchart illustrating an example of a stop operation example of the low priority class frame reading unit 6 between users when the low priority class bandwidth is fairly distributed among users.

  (S701, S702, S705) Whether or not the high priority class frame is read is determined. When reading, the length L of the read target frame is added to the read byte count XHi of the high priority class of user #i. When there is no reading, the inter-user low priority class frame reading unit 6 executes a low priority class read operation.

  (S703, S704) The inter-user low priority class frame reading unit 6 stops the low priority class read operation for a time corresponding to XHi. Thereafter, 0 is substituted into XHi.

  As shown in FIG. 5, implementation can be simplified by performing the processing of the inter-user low priority class frame reading unit 6 in increments of token length. When the processing of the low priority class frame reading unit 6 between users is performed in increments of token length, the flow of FIG. 7 can also simplify time management.

  FIG. 8 is a flowchart showing an example of a simplified stop operation in the low priority class frame reading unit 6 between users when the low priority class bandwidth is fairly distributed among users. Here, int (x) is a function that converts the real number x into an integer by rounding down the fractional part.

  Compared with FIG. 7, FIG. 8 shows that the multiple of T not exceeding XHi is set to S in S803, and the operation of the inter-user low priority class frame reading unit 6 is stopped for the reading time corresponding to S in S804. Yes. In S805, S is subtracted from XHi, and the rest is reflected at the next high priority class frame reading.

  FIG. 9 is a flowchart showing an operation example of the low priority class frame reading unit 6 between users when the low priority class bandwidth is fairly distributed among users.

  Compared with FIG. 4, there is no part to reflect the number of read bytes of the high-priority class from the given token, and the fairness of bandwidth distribution is ensured only in the low-priority class.

  Next, let us consider a case where the presence / absence of an operation for classifying the high priority class into two types and reflecting the use band in the inter-user low priority class frame reading unit 6 is mixed within the same user. Of the two types of high-priority classes, a class that does not reflect the used bandwidth in the inter-user low-priority class frame reading unit 6 is referred to as a high-priority class # 1, and a class that reflects the used bandwidth is referred to as a high-priority class # 2.

  FIG. 10 is a diagram illustrating an example of bandwidth distribution when the bandwidths of the high priority class # 2 and the low priority class are fairly distributed among users. In this distribution example, after allocating the band of the high priority class # 1 of each user, the remaining band is set as a band for the high priority class # 2 and the low priority class, and distributed among users based on the band distribution ratio. The band is determined. Of the bandwidths distributed to each user, the bandwidth is assigned with priority over the high priority class # 2, and the remaining bandwidth is assigned with the low priority class. Similar to the case of FIG. 3, when the traffic of the high priority class # 1 and the traffic of the high priority class # 2 exceed the maximum bandwidth set for each user, the excess portion of the frame is discarded.

  It is desirable to apply this band distribution method when the user contracts the high priority class # 1 line and the high priority class # 2 + low priority class line independently.

  The flowchart showing the operation of the inter-user low-priority class frame reading unit 6 when the high-priority class # 2 + and the low-priority class bandwidth is fairly distributed among users shows the operations performed on the high-priority class in FIG. , Replaced with an operation performed on the high-priority class # 2. In addition, since the use band of the high priority class # 1 is not reflected in the operation of the inter-user low priority class frame reading unit 6, the use band of the high priority class # 1 corresponds to the use band of the high priority class # 1. It is necessary to stop operation. The flowchart showing the stop operation of the low priority class frame reading unit 6 between users when the bandwidth of the high priority class # 2 + low priority class is fairly distributed among users is the operation performed for the high priority class of FIG. Is replaced with an operation performed on the high priority class # 1.

  According to the present embodiment, even when a large number of users are accommodated in the frame read scheduling, the bandwidth allocated to each user is distributed fairly based on a preset weight, and the read delay of the high priority connection frame is reduced. There is an effect of both reducing the conventional method.

  (Second Embodiment) In the configuration of the frame readout scheduling circuit of FIG. 2, when a high-priority class frame passes through the shared frame storage unit 5, a low-priority class frame readout unit between users for a time corresponding to the passed frame length. 6 is stopped. If the read target is a virtual frame when the inter-user low priority class frame reading unit 6 reads a frame, it assumes that a frame corresponding to the length of the virtual frame has been read and consumes the read right given as a token. . However, since there is actually no frame entity, in order not to cause a bandwidth breakdown, it is necessary to instantaneously search and determine the next user to be read. That is, the user to be read is determined twice or more within one search time. Therefore, the reading timing of the high-priority class frame in the inter-class frame reading unit 8 and the timing of the reading stop operation in the low-priority class frame reading unit 6 between users are made asynchronous, and the search for the user to be read within one search time is performed. By limiting to one time, it is possible to secure a sufficient search time for the user to be read.

  FIG. 11 is a diagram showing a second embodiment of the frame readout scheduling circuit according to the present invention. A read reserved frame storage unit 10 is arranged between the inter-user low priority class frame reading unit 6 and the interclass frame reading unit 8 having the configuration shown in FIG. The present embodiment is characterized in that the frame from the inter-user low priority class frame reading unit 6 is buffered in the read reserved frame storage unit 10 until the inter-class frame reading unit 8 reads the frame.

  The inter-user low priority class frame reading unit 6 reads frames at the same speed as the inter-class frame reading unit 8. When the frame of the high priority class passes through the shared frame storage unit 5, the virtual frame insertion unit 7 generates a virtual frame. When the same user as the virtual frame is granted a token and obtains the right to read the frame, the virtual frame inserting unit 7 inserts the virtual frame into the inter-user low priority class frame reading unit 6 and performs a frame reading process. . At this time, since there is no entity to read out the virtual frame, the inter-user low priority class frame reading unit 6 is stopped for the reading time corresponding to the virtual frame length.

  FIG. 12 is a diagram showing an example of scheduling in the case where there is a frame only in the low priority class in the band distribution method in which the bandwidth of all classes is distributed fairly among users. (A) is a block diagram of a simplified frame readout scheduling circuit, and (b) is an example of scheduling.

  The buffer prepared for each user in FIG. 12A corresponds to the user individual frame storage unit 3 and the in-user low priority class frame reading unit 4 in FIG. The WRR in FIG. 12 (a) corresponds to the inter-user low priority class frame reading unit 6 in FIG. 11, and in the example of FIG. 12 (a), scheduling is performed by WRR (weighted round robin) control. Yes. The PQ in FIG. 12A corresponds to the shared frame storage unit 5 in FIG. 11, and the RQ corresponds to the read reservation frame storage unit 10. The SP in FIG. 12A corresponds to the interclass frame reading unit 8 in FIG. 11, and scheduling is performed by priority read control in the example of FIG. Each user traffic can be classified into a high priority class and a low priority class. The frames of the high priority class are accumulated in the PQ, and the frames of the low priority class are read out by the WRR and accumulated in the RQ. The frames accumulated in PQ and RQ are scheduled and read by SP. Frames are stored in the buffer for each user provided in the preceding stage of WRR, and the numerical value in the frame indicates the frame length.

  FIG. 12B is an example of a read schedule of WRR and SP, and the horizontal axis is the time axis. A dotted line indicates a token (reading right) time interval given to each user. Here, it is assumed that the token time is equal to 1500 in frame length.

  When the frame transmission passes the time indicated by the dotted line, the excess is subtracted from the token given to the user next time. In FIG. 12B, the excess is indicated by a negative number below the frame. Also, the read ratio of each user (weight given to each user by WRR) is assumed to be equal.

  It is assumed that user #A is first selected by WRR. From the user #A, 800 + 1500 = 2300 frames are read by the first reading, and thus 800 frames are exceeded. Frames read by WRR are stored in RQ. Next, user #B is selected, 1200 + 1500 = 2700 frames are read and 1200 frames are exceeded, then user #C is selected, 800 + 800 = 1600 frames are read, and 100 frames are exceeded.

  In reading from the next user #A, the reading operation is stopped for a time corresponding to the previous excess 800. Although 700 tokens remain for user #A, since user #A has no frame to be read, the timing of the next read user search by WRR is advanced. Similarly, in the reading of the next user #B, the reading operation is stopped for a time corresponding to the previous excess 1200, and the remaining tokens are 300. Since 1500 frames are read out using this, it exceeds 1200 frames. Similarly, in the next user #C reading, the reading operation is stopped for a time corresponding to 100 of the previous excess, and the remaining tokens are 1400 frames. Using this, 1200 frames are read out. As a result, there is no frame to be read by user #C, so the WRR search is advanced by a time corresponding to 200 frames.

  The next user #A skips because there is no frame to be read, and reads from the next user #B. Since user #B has the previous excess of 1200 frames, the token value is 300 frames.

  On the other hand, the SP sequentially reads the frames accumulated in the RQ. FIG. 12B shows a state of a frame read from the SP. Since the WRR frame read end time and the SP frame read end time coincide with each other, no band splitting occurs.

  FIG. 13 is a diagram showing an example of scheduling in the case where there is a frame only in the high priority class in the bandwidth distribution method for fairly distributing the bandwidth of all classes among users. (A) is a block diagram of a simplified frame readout scheduling circuit, and (b) is an example of scheduling.

  When the high priority class passes the PQ, it is reflected in the WRR as a virtual frame. However, when a plurality of frames are reflected for the same user, the frame length is added and reflected. Since a total of 4200 frames are reflected as virtual frames for user #A, reading is stopped for a time corresponding to 1500 frames, which is the right to be read at one time. Similarly, a total of 1600 frames are reflected as a virtual frame for user #B, and a total of 2300 frames are reflected as a virtual frame for user #C. Only reading is stopped.

  Since the next user #A still has 2700 virtual frames remaining, reading is stopped for a time corresponding to 1500 frames, but since the remaining virtual frames of user #B are only 100 frames, this corresponds to 100 frames. The reading is stopped only for the time to be performed. Since user #B has no frame to be read, the WRR search is advanced by a time corresponding to 1400 frames. Similarly, the user #C and the user #A can reflect the high priority class traffic on the WRR by processing the remaining virtual frames.

  FIG. 14 is a diagram illustrating an example of scheduling in a case where there are frames in both the high priority class and the low priority class in the band distribution method in which the bandwidth of all classes is distributed fairly among users. (A) is a block diagram of a simplified frame readout scheduling circuit, and (b) is an example of scheduling.

  For user #A, a total of 2000 frames in the high-priority class have passed PQ, so reading is stopped for a time corresponding to 1500 frames in the first WRR search. The next user #B reads 1200 + 1500 = 2700 frames and exceeds 1200 frames. The next user #C reads 800 + 800 = 1600 frames and exceeds 100 frames.

  The next user #A stops reading for a time corresponding to 500 frames, and reads 800 + 1500 = 2300 frames using the remaining 1000 frames. As a result, it exceeds 1300 frames. Here, when the high priority class frame 800 of user #A newly arrives at PQ, the excess of user #A is 2100 frames. Similarly, when the user #B 1500 high priority class frame arrives at the PQ anew, the excess of the user #B becomes 2700 frames. In reading of the next user #B, reading is stopped only for a time corresponding to 1500 frames of the excess, and the excess is 1200 frames. The next user #C stops reading for a time corresponding to the excess 100 frames, and reads 1200 frames using the remaining 1400 frames for reading. Although the user has the right to read 200 frames, the user #C has no frame to be read, so the WRR search is advanced by a time corresponding to 200 frames.

  The next user #A stops reading for the time corresponding to 1500 frames of the excess, and the user #B stops reading for the time corresponding to 1200 frames which is the excess. From user #B, 1500 frames are read out using the remaining 300 frames of read right, and the number exceeds 1200 frames.

  Next, when the user #A is selected, the reading is stopped for a time corresponding to the excess 600 frames, and since the user #A has no other frames to be read, the WRR search is advanced by a time corresponding to 900 frames. . Similarly, the user #B stops reading for a time corresponding to 1200 frames, and the user #B has no other read target frame, so the WRR search is advanced by a time corresponding to 300 frames.

  In the second embodiment, the reading operations of the inter-user low priority class frame reading unit 6 and the inter-class frame reading unit 8 can be made asynchronous, and the inter-user low priority class frame reading unit 6 should read next. Frames can be determined in advance. This has the effect of preventing the occurrence of band breakage due to the inability to determine the read target user in time.

  Note that the read reservation frame storage unit 10 does not necessarily store actual frames, and may hold only frame information to be read. With such an implementation, it is possible to reduce the amount of memory required as a buffer and the processing delay time associated with frame accumulation.

  (Third Embodiment) This embodiment is characterized in that the frame storage state of the read reservation frame storage unit 10 is monitored and the frame reading speed in the low priority class frame reading unit 6 between users is adjusted.

  FIG. 15 is a diagram showing a third embodiment of the frame readout scheduling circuit according to the present invention. According to FIG. 15, the frame readout scheduling circuit has a readout reserved frame number monitoring unit 11 and a frame readout acceleration unit 12 in addition to the configuration of FIG. The read reservation frame number monitoring unit 11 monitors the frame storage state of the read reservation frame storage unit 10. The frame reading acceleration unit 12 instructs the inter-user low priority class frame reading unit 6 to accelerate the frame reading speed based on the instruction from the reading reserved frame number monitoring unit 11. Similarly, the frame readout inhibiting unit 9 instructs the inter-user low priority class frame readout unit 6 to inhibit the frame readout rate based on the instruction from the readout reserved frame number monitoring unit 11.

  In FIG. 15, the interclass frame reading unit 8 reads a high priority class frame from the shared frame storage unit 5 and a low priority class frame from the read reservation frame storage unit 10. For example, when priority reading control is applied as a scheduling method of the interclass frame reading unit 8, when the high priority class frame is in a burst state, the non-priority low priority class frame is read by the read reserved frame storage unit 10. I will wait. As shown in the second embodiment, when the continuity of the frame of the high priority class and the reading stop operation in the low priority class frame reading unit 6 between users become asynchronous, the low priority class frame is stored in the read reserved frame storage unit 10. Will accumulate. Since the length of the read reservation frame storage unit 10 is finite, if the number of frames to be stored exceeds the length of the read reservation frame storage unit 10, frame discarding occurs. Even if the inter-user low priority class frame reading unit 6 reads a frame by performing fair bandwidth distribution for each user. When frame discard occurs, fairness is lost. Further, when the inter-user low priority class frame reading unit 6 reads out a frame at the same speed as the transmission band, the band is broken due to frame discard.

  In order to avoid the above problem, the read reservation frame accumulation unit 10 is provided with a read inhibition start threshold, a read inhibition end threshold, a read acceleration start threshold, and a read acceleration end threshold, and is compared with the frame accumulation amount in the read reservation frame accumulation unit 10. As a result, the state of the inter-user low priority class frame reading unit 6 is changed.

  FIG. 16 is a state transition diagram of the inter-user low priority class frame reading unit 6 based on the state of the read reservation frame accumulating unit 10. The relationship between the threshold values is: read inhibition start threshold ≧ read inhibition end threshold ≧ read acceleration end threshold ≧ read acceleration start threshold.

  As states of the low priority class frame reading unit 6 between users, a normal state, a suppression state, a suppression release state, and an acceleration state are defined. The state depends on the speed at which the inter-user low-priority class frame reading unit 6 reads a frame from the intra-user low-priority class frame reading unit 4 and the unread portion of the read right corresponding to the time that cannot be read due to the read inhibition. It is classified.

  The normal state is a state in which normal reading is performed with no unread portion. The suppression state is a state where there is an unread portion and reading suppression is performed. The suppression release state is a state where there is an unread portion and normal reading is being performed. The acceleration state is a state where there is an unread portion and acceleration reading is performed. Here, accelerated reading is a state in which more reading rights are given at one time, for example, giving twice the reading right than normal reading. Even during accelerated reading, since the bandwidth ratio to be distributed for each user does not change, the fairness of the bandwidth distribution for each user can be ensured.

  First, let the normal state be the initial state. In the normal state, when the frame accumulation amount becomes larger than the reading inhibition start threshold, the state is changed to the inhibition state. In the inhibited state, since reading is inhibited, the unread amount increases. When the frame accumulation amount becomes smaller than the reading inhibition end threshold value in the inhibition state, the state changes to the inhibition cancellation state. When the frame accumulation amount becomes larger than the reading inhibition start threshold value in the inhibition release state, the state changes to the inhibition state, and when the frame accumulation amount becomes smaller than the reading acceleration start threshold value, the state changes to the acceleration state. In the acceleration state, the read right to be given at the time of reading increases, and the read right increased compared with the normal state is subtracted from the unread portion, so the unread portion decreases. In the acceleration state, when the frame accumulation amount becomes larger than the reading acceleration end threshold value, the state shifts to the suppression release state, and when the unread portion becomes 0 regardless of the amount of the frame accumulation amount, the state transitions to the normal state. .

  In the third embodiment, even when the number of frames stored in the read reservation frame storage unit 10 having a finite length increases and exceeds the number of frames that can be stored, the effect of ensuring the fairness of the bandwidth allocated to each user In addition, the number of frames read from the inter-user low-priority class frame reading unit 6 is reduced, thereby preventing the occurrence of band breakage.

  (Fourth Embodiment) This embodiment is characterized in that dynamic bandwidth control of a high priority class is performed by call control.

  FIG. 17 is a diagram showing a fourth embodiment of the frame readout scheduling circuit according to the present invention. According to FIG. 17, the frame read scheduling circuit has a connection establishment control unit 13 in addition to the configuration of FIG. Based on the call control information from the outside, the connection establishment control unit 13 conducts up to a certain number of user frames, and other user frames are non-conducted, or suppressed to a certain band or less. Decide whether to implement control and instruct the traffic monitoring / shaping unit 2.

  By combining with an external authentication server, it is possible to perform control such as giving a band only to a user who has been authenticated by a specific condition, for example, a condition such as a configuration of a community that has purchased content. Further, instead of performing the control based on the call control information from the outside, the connection establishment control unit 13 can determine whether or not the user frame can be conducted from the state of the high priority class frame.

  In the fourth embodiment, it is possible to accommodate a large number of users while restricting the total bandwidth of the high priority class traffic to a certain level, and to ensure the quality when establishing communication and performing communication.

It is a figure which shows the frame read scheduling circuit by a prior art. It is a figure which shows 1st embodiment of the frame read scheduling circuit by this invention. It is a figure which shows the example of a bandwidth distribution in the case of carrying out fair distribution of the bandwidth of all the classes between users. It is a flowchart which shows the operation example of the low priority class flame | frame reading part 6 between users when all the zone | bands are equally distributed among users. It is a flowchart which shows the simplified operation example in the low priority class flame | frame reading part 6 between users when all the zone | bands are equally distributed among users. It is a figure which shows the example of band distribution in the case of carrying out fair distribution of the band of a low priority class between users. It is a flowchart which shows the example of a stop operation | movement of the low priority class flame | frame reading part 6 between users when the band of a low priority class is equally distributed between users. 10 is a flowchart showing an example of a simplified stop operation in the inter-user low priority class frame reading unit 6 when the low priority class bandwidth is fairly distributed among users. It is a flowchart which shows the operation example of the low priority class flame | frame reading part 6 between users in the case of carrying out fair distribution of the band of a low priority class between users. It is a figure which shows the example of a bandwidth distribution in the case of carrying out fair distribution of the bandwidth of high priority class # 2 and a low priority class between users. It is a figure which shows 2nd embodiment of the frame read scheduling circuit by this invention. It is a figure which shows the example of a scheduling in case there exists a frame only in a low priority class. It is a figure which shows the example of a scheduling when there exists a frame only in a high priority class. It is a figure which shows the example of a scheduling in case a frame exists in both a high priority class and a low priority class. It is a figure which shows 3rd embodiment of the frame read scheduling circuit by this invention. FIG. 6 is a state transition diagram of the inter-user low priority class frame reading unit 6 based on the state of the read reserved frame storage unit 10. It is a figure which shows 4th embodiment of the frame read scheduling circuit by this invention.

Explanation of symbols

1 Frame distribution part
2 Traffic monitoring / shaping unit 3 User individual frame accumulating unit 4 Intra-user low priority class frame reading unit 5 Shared frame accumulating unit 6 Inter-user low priority class frame reading unit 7 Virtual frame inserting unit 8 Interclass frame reading unit 9 Frame reading suppression Section 10 Reserved reserved frame storage section 11 Reserved reserved frame number monitoring section 12 Frame reading acceleration section 13 Connection establishment control section 14 Intra-user frame priority reading section 15 Inter-user frame fair reading section 16 Intra-user frame fair reading section

Claims (8)

To accommodate one or more users who use one or more connections, determine the bandwidth to be allocated to each connection based on the priority set for each connection and the bandwidth distribution ratio set for each user, and for frame reading In the method of scheduling,
A first step of allocating the input frame for each user and for each class with reference to information and / or attributes set in advance for the connection given to the input frame;
A second step of monitoring and / or shaping the frames allocated to the high priority class based on information set in advance and discarding frames exceeding the set allowable amount;
A third step of reading out frames for each user class allocated to the low priority class by a preset scheduling method;
A fourth step of reading out the low-priority class frame read in the third step and the frame allocated to the high-priority class by a preset scheduling method;
In the frame reading in the third step, the user reading of the read high priority class frame is stopped for a time corresponding to the high priority class frame length read in the fourth step, or the fourth step A frame reading scheduling method characterized in that, in the step, the entire third step is stopped for a time corresponding to the read high priority class frame length.   The frame read scheduling method according to claim 1, wherein the frame read in the third step is stored in a buffer.   3. The frame read scheduling method according to claim 2, wherein the read operation in the third step is suppressed or accelerated according to the number of frames accumulated in the buffer.   4. The frame read scheduling method according to claim 1, wherein in the second step, monitoring, shaping, and discard control of a high-priority class frame are performed by external call control. 5. To accommodate one or more users who use one or more connections, determine the bandwidth to be allocated to each connection based on the priority set for each connection and the bandwidth distribution ratio set for each user, and for frame reading In the circuit that performs scheduling,
Frame distribution means for distributing the input frame for each user and for each class with reference to information and / or a predetermined attribute assigned to the input frame and a connection;
The frame of the high priority class distributed by the frame distribution means is used as an input, and the input frame is monitored and / or shaped based on preset information, and frames exceeding the set allowable amount are discarded. , Traffic monitoring / shaping means for each user,
User-specific frame storage means for each user, which receives a frame of a low priority class distributed by the frame distribution means, and stores the input frame in a FIFO buffer prepared for each class;
In-user low priority class frame reading means for each user, which sequentially reads out the frames stored in the user individual frame storage means by a preset scheduling method;
Shared frame storage means for receiving a frame from the traffic monitoring / shaping means of each user as input and storing the input frame in a FIFO buffer prepared for each class;
Inter-user low priority class frame reading means for sequentially reading frames by a preset scheduling method from the intra-user low priority class frame reading means of each user;
The length of the frame passing through the shared frame storage unit is monitored for each user, a virtual frame for each user is generated from the acquired frame length information, and the generated virtual frame is transmitted to the low priority class frame reading unit between users. Virtual frame insertion means for inserting
Inter-class frame reading means for reading a frame by a preset scheduling method from the shared frame storing means and the inter-user low priority class frame reading means;
Frame reading inhibiting means for inhibiting the reading operation of the inter-user low priority class frame reading means based on traffic information acquired from the shared frame accumulating means and / or the interclass frame reading means. Frame readout scheduling circuit. Read reservation frame storage means for storing frames read by the low priority class frame reading means between users,
6. The frame reading scheduling circuit according to claim 5, wherein the interclass frame reading means reads a frame from the shared frame storing means and the read reserved frame storing means by a preset scheduling method. Read reserved frame number monitoring means for monitoring the number of frames stored in the read reserved frame storage means;
Frame reading acceleration means for accelerating the reading operation of the inter-user low priority class frame reading means in response to an instruction from the reading reserved frame number monitoring means;
7. The frame reading scheduling circuit according to claim 6, wherein the frame reading inhibiting means inhibits the reading operation of the inter-user low priority class frame reading means even in accordance with an instruction from the reading reserved frame number monitoring means. 8. A connection establishment control means for determining whether or not to execute control of a high priority class frame by external call control and instructing the traffic monitoring / shaping means. The frame readout scheduling circuit described.

Images