JP2011199468A - Transmission regulation method and system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To control the ratio of sample terminal MN(s) dynamically based on the load conditions of an end server ES in transmission regulation method and system which limits the load of the end server ES by estimating the total number of transmissions based on the number of transmissions of the sample terminal MN(s).SOLUTION: When the transmission operation is performed by a user terminal MN, only the sample terminal MN(s) classified according to the sample probability Rsample transmits to the end server ES, and transmission of other sample terminal MN(s) is pended. A control server CS determines the conditions of release transmission pended in each user terminal MN, in units of control slot, based on the access situation to the end server, and informs it to each user terminal MN. Furthermore, the control server CS determines the sample probability Rsample dynamically based on the access situation to the end server ES, and informs it to each user terminal MN. The user terminal MN satisfying the conditions of release pending transmits to the end server ES.

Description

  The present invention relates to an outgoing call control method and system for restricting access from a user terminal to an end server, and in particular, restricts outgoing calls from a user terminal to an end server to limit the number of accesses to the end server and the number of simultaneous connection sessions. The present invention relates to a transmission restriction method and system for restricting to a system capacity of a server or less.

  Download requests requesting distribution of popular songs immediately after the release of the hit chart on TV programs and radio programs, and upload requests for viewers to post their own messages in viewer-participation programs For example, request messages triggered by specific events tend to concentrate in a short time. However, in servers (end servers) that accept requests at these destinations, the number of requests from user terminals increases rapidly, and congestion occurs when the number of simultaneous connection sessions and the number of accesses exceed the end server processing capacity (for example, system capacity). The load state deteriorates rapidly, and in the worst case, the server operation is stopped.

  In response to these technical issues, an access path server dedicated to accepting requests is provided separately from the end server that accepts content distribution and posting requests, and this access path server accesses end servers in response to requests. Patent Document 1 discloses a technique for determining the order and access timing, and notifying each user terminal of this as an access path, and accessing the end server according to the access path in each user terminal.

  In addition, in Patent Document 2, when access from a user terminal is concentrated, transmission from all or a part of the user terminals is temporarily suspended on the user terminal side, and then to the user terminal using a broadcast wave or the like. Sequential reception-type outgoing control technology that reduces the load on the end server by sending control information describing the conditions for releasing the hold and releasing the outgoing hold so that access to the end server is less than the system capacity Is disclosed.

  In the techniques of Patent Documents 1 and 2 described above, the load on the end server is determined mainly based on the number of calls from the user terminal to the end server, and the load on the end server increases as the number of calls from the user terminal increases. Regarding transmission, restrictions are placed. However, the load on the end server is not only the number of calls from the user terminal, but also the time required for the end server to process the service request sent from the user terminal and the time required for data transfer (hereinafter referred to as service time). Depends on).

  For example, if the service request transmitted from the user terminal to the end server is content download, the larger the content data size, the longer the download time, that is, the processing time, and the load on the end server. However, since the server load is obtained based on the number of outgoing calls in the prior art, the server load is estimated without considering the data size (download time) of the content.

  In addition, even if the data size of the content is the same, the user terminal that requests it is connected to a high-speed network such as a broadband line using an optical fiber and to a low-speed network such as a cellular phone network. The server load differs greatly from case to case. Specifically, since the delivery time (communication session connection time) is longer in the delivery via the low-speed network than in the delivery via the high-speed network, the processing time is increased and the load is increased even in the end server. Therefore, in the prior art, the load is estimated to be smaller than actual for transmission via the low-speed network.

  As described above, the load on the end server is determined by the speed of the network to which the user terminal is connected, the size of transmitted / received data, the speed of the access line of the end server, and the time required for processing the service request transmitted from the user terminal. Dependent. However, in the above-described conventional technology, since transmission control is performed based only on the number of transmissions to the end server, depending on the access situation, for example, when a service with a long download time and a short service are mixed, It was difficult to accurately detect and control the load.

  These new technical issues include the measurement and monitoring of all transmissions from each user terminal to the end server, and rigorously analyzing them. However, if a large amount of access occurs, it is substantially difficult to analyze all the details of all outgoing calls and access lines and reflect them in the control in real time.

  Therefore, the inventors of the present invention control the transmission from the user terminal to the end server without strictly analyzing the transmission from each user terminal to the end server, and are established between the user terminal and the end server. Has invented and applied for a patent application (Patent Document 3).

JP 2009-43069 A JP 2008-2111730 A Japanese Patent Application No. 2008-232349

  In the prior art, each user terminal MN makes a call for requesting a desired service to the end server ES in response to the call request operation of the terminal user. However, in order to estimate the number of accesses to the end server ES, a part of the user terminal MN is classified into the sample terminal MN (s) with a predetermined sample probability Rsample, and only the sample terminal MN (s) is used for the call operation. While the transmission is permitted immediately, the other user terminals MN are temporarily suspended from the transmission. The number of transmissions to the end server ES by the sample terminal MN (s) is measured by the measurement server DS, and the total number of transmissions is calculated based on the number of accesses by the sample terminal MN (s) and the sample probability Rsample.

  However, since the sample probability Rsample is a fixed value in the prior art, if the number of transmitting user terminals MN is large, the traffic exceeds the system capacity of the end server ES only with access by the sample terminal MN (s), and overloading occurs. And congestion may occur.

  In addition, if such a situation is assumed and the sample probability Rsample is set to a low value in advance, then the accuracy of the total number of transmissions estimated based on the sample probability Rsample and the number of transmissions of the sample terminal MN (s) Therefore, there is a technical problem that the accuracy of access control based on the estimated value is reduced.

  The object of the present invention is to solve the above-mentioned problems of the prior art, hold the transmission of the user terminal except for the transmission of the sample terminal MN (s) which is a part thereof, and the sample terminal MN (s) is generated Dynamically control the ratio of sample terminals MN (s) based on the access status to the end server ES in a call control system that distributes the load on the end server by time by estimating the traffic volume based on the traffic volume It is to provide a method and system for restricting outgoing calls.

  In order to achieve the above object, the present invention provides a transmission regulation system that temporarily holds a transmission from a user terminal to an end server, and then sequentially releases the hold in units of control slots to limit the load on the end server. Each user terminal and the control server that cancels the suspension of outgoing calls are characterized in that they have the following configuration.

  (1) means for receiving a sample probability that the user terminal classifies a part of the user terminal as a sample terminal; means for classifying the own terminal into a sample terminal based on the sample probability; and the sample terminal in response to a call operation If it is a non-sample terminal, the means for holding the call, the means for receiving the hold release condition, and the call if the call on hold satisfies the received hold release condition. Means for continuing the suspension.

  A control server for measuring a load parameter value representative of the load status of the end server; a means for determining a hold release condition based on the load parameter value; and the sample probability based on the load parameter value. And means for notifying each user terminal of the sample probability and the hold release condition.

  (2) The means for measuring the load parameter value measures the amount of sample traffic for each control slot.

  (3) The means for measuring the load parameter value is configured to measure, for each control slot, the sum of the traffic volume of the sample terminal and the traffic volume due to the outgoing call released as the traffic volume arriving at the end server.

  (4) The means for measuring the load parameter value measures the total amount of traffic that is on hold for each control slot.

  (5) The means for measuring the load parameter value measures the maximum value of the hold time for each control slot for traffic that is on hold.

  According to the present invention, the following effects are achieved.

  (1) If the load on the end server is large, the percentage of sample accesses in the access to the end server will decrease, and the percentage of access due to release of the hold will increase. It is done. In addition, since the sample probability is set high if the load on the end server is small, the accuracy in estimating the total number of outgoing calls from the number of sample terminals can be improved.

  (2) If the means for measuring the load parameter value measures the sample traffic volume, the sample ratio can be corrected based on the sample traffic volume, so it is possible to directly avoid exceeding the system capacity of the end server due to the traffic generated by the sample terminal). it can. In addition, it is possible to directly improve the shortage of estimation accuracy of the number of sample terminals.

  (3) If the means for measuring the load parameter value measures the amount of traffic arriving at the end server, the sample ratio can be corrected based on the amount of traffic arriving at the end server, so sample terminals and non-sample terminals are generated. It is possible to directly avoid exceeding the end server system capacity due to traffic volume. In addition, it is possible to directly improve the shortage of estimation accuracy of the number of sample terminals.

  (4) If the means for measuring the load parameter value measures the total amount of traffic that is on hold, the sample ratio can be corrected based on the total amount of traffic that is on hold, thus maintaining necessary and sufficient control accuracy. However, by releasing a larger amount of pending traffic, it is possible to perform appropriate control that can reduce the pending continuation of pending traffic, that is, the waiting time.

  (5) If the means for measuring the load parameter value measures the maximum value of the hold time for traffic that is on hold, the sample ratio can be corrected based on the maximum hold time, so the waiting time for the pending traffic can be reduced. Appropriate control that can be shortened becomes possible.

It is the block diagram which showed the structure of the network to which the transmission control system of this invention is applied. 5 is a flowchart showing an operation of a user terminal MN. 5 is a flowchart showing the operation of a control server CS. It is a figure for demonstrating the determination method of the control slot which cancels | releases the holding | maintenance of transmission. It is the flowchart which showed the procedure of hold release determination. It is a figure for demonstrating the 1st control method of sample ratio Rsample. FIG. 10 is a diagram (No. 1) for describing a second control method of the sample ratio Rsample. FIG. 10 is a diagram (No. 2) for describing a second control method of the sample ratio Rsample. It is a figure for demonstrating the 3rd control method of sample ratio Rsample. It is a figure for demonstrating the 4th control method of sample ratio Rsample. It is the figure (the 1) explaining the effect of this invention compared with the prior art. It is FIG. (2) explaining the effect of this invention compared with the prior art. It is a figure for demonstrating the relationship between transmission timing and a destination address. It is a figure for demonstrating the relationship between the timing of hold cancellation | release, and a destination address. It is the figure which showed an example of the number of accesses for every destination address detected by the measurement server.

  Hereinafter, the best embodiment of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a network to which the outgoing call restriction system of the present invention is applied. Here, a user terminal MN accesses an end server ES (service providing server) to request a content distribution service. Then, in response to this, the transmission restriction when the end server ES delivers content to the user terminal will be described as an example.

  Each user terminal MN makes a call for an access requesting a desired service from the end server ES in response to a call request operation of the terminal user. However, in this embodiment, a part of the user terminal MN is classified into the sample terminal MN (s) with a predetermined sample probability Rsample, and only the sample terminal MN (s) is allowed to make a call in response to the call operation. On the other hand, the others are classified as non-sample terminals MN (n) and the call is temporarily suspended. Hereinafter, the sample terminal MN (s) and the non-sample terminal MN (n) may be collectively referred to as the user terminal MN.

  As will be described in detail later, the sample terminal MN (s) and the non-sample terminal MN (n) are assigned different addresses as destination addresses to the end server ES, and the measurement server DS transmits the sample terminal MN ( The traffic volume of s) is measured and reported to the control server CS. As will be described in detail later with reference to the flowchart, the control server CS establishes a communication session established between each user terminal MN and the end server ES at the timing of canceling the outgoing call held by each user terminal MN. , And the number of sessions that can be connected to the end server ES at the same time, and this is notified to the broadcast server BS as a condition for releasing the hold.

  The broadcast server BS transmits the control information in which the notified hold release condition is described by broadcast waves from the broadcast station BC. When each user terminal MN receives the control information through a broadcast wave, the user terminal MN releases the hold if the pending transmission request generation timing is earlier than the hold release timing, and if it is later than the hold release timing. Continue holding. Note that the method of notifying control information from the control server CS to each user terminal MN is not limited to broadcast waves, and may be notified by other appropriate notification means such as the Internet.

  In the user terminal MN released from the hold, a call for accessing the end server ES is executed, and a communication session is established between the user terminal MN and the end server ES. In the present embodiment, any of a plurality of IP addresses, port numbers, and URLs, or a combination (hereinafter collectively referred to as destination addresses AD1, AD2,...) Is virtually allocated to the end server ES, and the control information is included in the control information. Describes different destination addresses depending on whether a transmission request is generated (control slot) and also whether the user terminal MN is a sample terminal MN (s) or a non-sample terminal MN (n). Then, each non-sample terminal MN (n) whose hold has been released transmits to a unique destination address assigned to the generation timing (control slot) of the transmission request.

  That is, in the present embodiment, if the user terminal has the same generation request generation timing (control slot), the sample terminals MN (s) and non-sample terminals MN (n) have the same destination address. If the generation timing of the transmission request is different, the destination addresses to be assigned are different even between the sample terminals MN (s) or the non-sample terminals MN (n).

  If the service requested from each user terminal MN is a content distribution request, the end server ES distributes the content to the user terminal MN and shuts down the communication session after the completion. In this embodiment, the time required for the end server ES to provide a service in response to a request from each user terminal MN, in other words, the time during which a communication session is established is defined as “service time (TAT)”. .

  FIG. 2 is a flowchart showing the operation of each user terminal MN, and is autonomously and repeatedly executed in a predetermined cycle asynchronously with the control slot.

  In step S11, control information transmitted from the broadcasting station BC for each predetermined control slot, received at each user terminal MN, and stored in a memory or the like is read. As will be described in detail later, this control information includes a hold release condition in the next control slot τ1, and a destination address when the user terminal MN that has made a transmission operation in the next control slot τ1 accesses the end server ES. Various types of information are described, and the destination address of the end server ES includes a destination address ADs [τ1] assigned to the sample terminal MN (s) and a destination address ADn [τ1] assigned to the non-sample terminal MN (n). It is different.

  In step S12, it is determined whether or not there is a call that is on hold. At first, since it is determined that the call that is on hold is not registered, the process proceeds to step S19. In step S19, it is determined whether or not there is a new call operation, and if a user call operation is detected, the process proceeds to step S20. In step S20, a random number rand1 (0 <rand1 ≦ 1) is generated by applying a predetermined random number seed to the random number function. In step S21, the random number rand1 is compared with the sample probability Rsample [τ0]. The sample probability Rsample [τ0] is a probability that a part of the user terminal MN is classified as the sample terminal MN (s). As will be described in detail later, the sample terminal MN is provided for each control slot τ in the control server CS. It is dynamically controlled according to the traffic volume of (s), the traffic volume that is on hold, and the like, and is appropriately notified to all user terminals MN by a broadcast wave or the like.

  If rand1> Rsample [τ0], the process proceeds to step S22 to behave as a non-sample terminal MN (n). In step S22, the current call is put on hold, and the call request timing τ1 and the destination address ADn [τ1] are registered. On the other hand, if rand1 ≦ Rsample [τ0], the process proceeds to step S15 to behave as the sample terminal MN (s).

  In step S15, transmission of the sample terminal MN (s) is permitted, and transmission to the destination address ADs [τ1] for the sample terminal MN (s) acquired in step S11 is immediately executed. When a communication session is established with the end server ES in step S16, the process proceeds to step S17 to receive a content distribution service. In step S18, when the content distribution service is completed, the communication session established with the end server ES is blocked. In the measurement server DS, the amount of sample traffic from each sample terminal MN (s) to the end server ES is measured for each control slot and for each destination address. This measurement result is notified to the control server CS at a predetermined timing. The processing of steps S13 to S18 will be described later.

  FIG. 3 is a flowchart showing the operation of the control server CS. The main control unit including the CPU of the control server CS is volatile based on programs and control data stored in a nonvolatile storage device in advance. The storage area is executed as a work area. Whenever the current control slot [τ0] ends, the control server CS executes a hold release process for specifying a call to be released from the hold in the next control slot [τ1].

  In step S31, when the end timing of the current control slot τ0 is detected, in step S32, the traffic volume (sample traffic volume) of the service request transmitted from the sample terminal MN (s) to the end server ES in the current control slot τ0. ) Λsample [τ0] is acquired from the measurement server DS. In step S33, the sample traffic amount λsample [τ0] and the ratio Rsample [τ0] of the sample terminal MN (s) are applied to the following equation (1) to estimate the total traffic amount requested for transmission in the control slot τ0. The value λall [τ0] is calculated. The sample ratio Rsample [τ0] is notified from the control server CS for each control slot τ, and the initial value Rsample_init is also registered in the previous control server CS, and is notified to all user terminals MN by broadcast waves or the like. In step S34, the traffic amount (hold traffic amount) λwait [τ0] of the non-sample terminal MS (n) held in the control slot τ0 is estimated by the following equation (2).

  In step S35, the control slot τ0 is generated as one of the traffic amounts ψτx [τy] (x ≦ y) that occurs in the previous xth control slot τx and continues to be held at the end of the yth control slot τy. The amount of traffic ψτ0 [τ0] that is generated at the time and is kept suspended even at the end of the control slot τ0 is calculated. The initial values of the traffic amount ψτx [τy] to be kept on hold are all “0”, and transmission is performed at each user terminal MN, and the value gradually fills as the processing proceeds. In the present embodiment, all but the transmission of the sample terminal MN (s) are temporarily held, so that the traffic amount ψτ0 [τ0] to be held continuously is calculated by the following equation (3).

  In step S36, it is determined whether the sample ratio Rsample needs to be reviewed. In this embodiment, as will be described in detail later, (1) the amount of sample traffic, (2) the amount of traffic arriving at the end server ES (the sum of the amount of sample traffic in each control slot and the amount of traffic released from hold), (3 Load parameter values that represent the load status of the end server ES, such as the total traffic volume that is held on hold or (4) the maximum hold duration, are compared with the system capacity Cmax of the end server ES and the set threshold target. Is done. If the load parameter value has a predetermined relationship with the system capacity Cmax and the set threshold value target, the sample ratio Rsample is corrected to increase (decrease) in step S37, and otherwise, the current state is maintained.

  In step S38, the margin capacity φ [τ1] to be secured in the end server ES in preparation for the transmission by the sample terminal newly generated in the next control slot τ1 is the margin ratio Rmargin to be secured for the sample in the next control slot τ1. Is calculated by the following equation (4). The margin ratio Rmargin is set in advance to an appropriate value of, for example, “twice”.

  In step S39, the free capacity C [τ1] of the end server ES in the next control slot τ1 is processed by the end server ES in order to process the maximum capacity Cmax of the end server ES and the newly generated sample terminal MN (s). The margin capacity φ [τ1] required for ES is calculated by applying the following equation (5).

  In step S40, among the transmissions that occur in each control slot τk (k ≦ 0) before control slot τ0 and continue to be held after the end of control slot τ0, the suspension is released on a control slot basis from the oldest generation time. When this is done, the maximum value n of slot numbers that can release all pending transmissions in the next control slot τ1 is calculated as a condition for releasing the hold.

  Here, among the control slots that continue to hold at least one outgoing call, the maximum value n of the control slot numbers that can release the hold in the next control slot τ1 is the total traffic amount held. When the transmission operation is accumulated in chronological order in units of control slots, the sum up to the nth control slot is less than the system free capacity C [τ1] in the control slot τ1, but the (n + 1) th control slot The control slot number n is such that if the total amount of traffic that continues to be held after the end of the control slot τ0 is further accumulated, the system free capacity C [τ1] will be exceeded.

  That is, as shown in an example in FIG. 4, the traffic amount reserved in the control slot τp (p + 2 ≦ k) is “20”, the traffic amount in the control slot τp + 1 is “15”, and the control slot τp + If it is set to “30” in 2 and the system free capacity C [τ1] in the control slot τ1 is set to “50”, the suspension is started from the control slot τp and all the suspensions of the next control slot τp + 1 are released. Even so, the total capacity is “35” and does not reach the free capacity C [τ1]. However, if all the hold of the next control slot τp + 2 is canceled, the total capacity becomes “65” and exceeds the free capacity C [τ1]. In such a case, [p +1] is the maximum value n.

  Therefore, if the traffic amount generated in the control slot τx and kept suspended after the end of the control slot τ0 is ψτx [τ0], the maximum n (n ≦ 0) satisfying the following equation (6) is It becomes the maximum value of the slot number that can cancel all outgoing calls. If the maximum value n is obtained, all the pending transmissions can be canceled up to the nth control slot, and all the transmissions are kept pending from the (n + 2) th control slot.

  On the other hand, for the (n + 1) th control slot, if the total amount of traffic released from the hold until control slot number n is less than the free capacity C [τ1], the hold is released only for some outgoing calls. become. It should be noted that the traffic amount ψτx [τ0] during suspension that is the subject of suspension cancellation is obtained by equations (7.1), (7.2), and (7.3) in step S41 described later in the previous control cycle (control slot). It is done.

  Returning to FIG. 3, in step S41, with respect to the (n + 1) th control slot in which only a part of the outgoing calls is released, in order to calculate the ratio Rr [τ1] of outgoing calls that can be released in the control slot τ1, The amount of traffic ψτm [τ1] (k ≦ m ≦ 0) that continues to be held at the end of the control slot τ1 without satisfying the hold release condition, and the hold release in the control slot τ1 that satisfies the hold release condition The traffic amount ρτm [τ1] is calculated and stored based on the following equations (7) and (8), respectively. Here, ρτx [τy] is the amount of traffic that occurs in the control slot x and is released from the hold in the control slot y. The initial value is all “0”, and the value increases as the processing proceeds. Gradually fill up.

  The above formulas (7.1) and (8.1) are applied to all control slots with slot number m of m ≦ n, all pending transmissions are released at control slot τ1, and continue to be held at the end of control slot τ1. Traffic amount ψτm [τ1] to be zero [(7.1)], and the traffic amount ψτm [τ0] kept suspended at the end of the control slot τ0 and the traffic amount ρτm [τ1] released from the control slot τ1 Equal to [(8.1)].

  Equations (7.3) and (8.3) apply to all control slots with slot number m of m> n + 1, and all pending transmissions are not released in control slot τ1 [(8.3)] At the end of τ1, the same number of outgoing calls as that at the end of control slot τ0 is held on hold [(7.3)].

  The above formula (7.2) is applied to the control slot with slot number m = n + 1, and is free from the sum of the transmissions up to slot number n + 1 among all the transmissions that are held on hold at the end of control slot τ0. The amount of traffic reduced by the capacity C [τ1], that is, the traffic amount ψτn + 1 [τ0] generated at the control slot number τn + 1 and held at the end of the control slot τ0 is released from the control slot τ1. The number of transmissions with a reduced traffic volume is held on hold even at the end of the control slot τ1.

  The above equation (8.2) is also applied to the control slot with the slot number m = n + 1 for which transmission is suspended, and is suspended from the free capacity C [τ1] at the end of the control slot τ0 in the control slot τ1. Out of all traffic, the number of outgoing calls obtained by subtracting the total traffic amount up to slot number n is released.

In step S42, of the traffic amount ψτn + 1 [τ0] generated in the (n + 1) th control slot and continuing to be held at the end of the control slot τ0, the amount of traffic released from the control slot [τ1] ρτn + 1 ratio of [τ1] Rr [τ1] is obtained by the following equation (9).

  In step S43, a destination address to be assigned as an access destination for a call operation generated at each user terminal MN in the control slot τ1 is appropriately determined. In the present embodiment, the destination address ADs [τ1] assigned to the sample terminal MN (s) and the destination address ADn [τ1] assigned to the non-sample terminal MN (n) are determined.

  In step S44, the pair (τn + 1, Rr [τ1]) of the control slot number τn + 1 and the traffic volume ratio Rr [τ1] released from the hold and the destination addresses ADs [τ1], ADn [τ1 ] Is sent to the broadcast server BS as control information and transmitted from the broadcast station BC to the user terminal MN. In step S45, based on the following equation (10), out of the number of sessions (necessary capacity) established by the outgoing call released from the control slot τ1, the number of sessions φ [τ1 in which the service time continues after the control slot τ1 + i] is calculated.

  In other words, for example, when the number of sessions newly established by a call released from the control slot τ1 ends in the control slot τ1, the margin capacity φ [τ1] is a call released from the hold by the control slot τ0. The number of sessions established by the above is the number of sessions in which the service time continues even in the control slot τ1, plus the number of sessions established by the outgoing call released in the control slot τ1. The initial values of the number of sessions φ [τx] are all “0”, and the values are gradually filled with the progress of processing.

  The TAT is the service time (seconds), and can be obtained by quoting past performance data, or by separately providing a server for observing the TAT and measuring the service time of some or all sessions. Δτ is a time width (second) of the control cycle (control slot τ). Note that TAT / Δτ is a rounded-up integer value when it includes 0 or a decimal point.

  For example, if the service time TAT is 10 (seconds) and the control cycle Δτ is 5 (seconds), the end server provides services such as content distribution (downloading) in response to a call released from the control slot τ1. Since the time (service time) corresponds to “2” in the number of control slots, the service time started in the control slot τ1 is continued in the control slots τ1 and τ2. Accordingly, here, the margin capacities φ [τ1] and φ [τ2] are obtained and temporarily stored, and are used as the margin capacities φ [τ2] in step S39 in the next control cycle τ2.

  Returning to FIG. 2, if it is determined in step S12 that the user terminal MN has an outgoing call operation that is on hold, the process proceeds to step S13. In step S13, it is determined whether or not to cancel the hold for each outgoing call that is on hold.

  FIG. 5 is a flowchart showing the procedure for determining the hold release. In step S51, it is determined whether or not a call that has occurred before the hold release timing τn + 1 registered in the control information is on hold. Is determined. If it is on hold, the process proceeds to step S52, and all the hold is released. In step S53, it is determined whether or not the transmission generated at the hold release timing τn + 1 registered in the control information is on hold. If it is on hold, the process proceeds to step S54.

  In step S54, a predetermined random number seed is applied to the random number function to generate a random number rand2 (0 <rand2 ≦ 1). In step S55, the random number rand2 is compared with the hold release rate Rr [τ1] registered in the control information. If it is determined that rand2 ≦ Rr [τ1], the process proceeds to step S56 and the hold is released. If it is determined in step S55 that rand2> Rr [τ1], the hold release is postponed together with the transmission that occurs after τn + 1, and the hold is continued.

  Returning to FIG. 2, in step S <b> 14, it is determined whether or not the hold release condition is satisfied. If satisfied, the process proceeds to step S <b> 15. In step S15, the call released from the hold is executed. When a communication session is established with the end server ES in step S16, the process proceeds to step S17 to receive the requested content distribution service. If the completion of the service is detected in step S18, the communication session established with the end server ES is blocked.

  According to the present embodiment, if the load on the end server ES is large, the ratio of sample access to the access to the end server ES decreases and the ratio of access due to hold release increases. The fairness that can be increased. In addition, since the sample ratio Rsample is set high if the load on the end server ES is small, it is possible to improve the accuracy in estimating the total traffic volume from the traffic volume of the sample terminal.

  Next, a method for dynamically controlling the sample ratio Rsample in steps S36 and S37 based on load parameter values that can represent the load status and congestion status of the end server ES will be described in more detail.

[First control method]
FIG. 6 is a diagram for explaining the first control method. The load parameter setting threshold value target and / or the system capacity Cmax are defined in advance, and the load parameter value is equal to the setting threshold value target as in the control slot τi + 1. If it exceeds, the sample ratio Rsample is decreased by a predetermined ratio. On the other hand, if the load parameter value is below the set threshold value target as in the control slots τi, τi + 2, the current sample ratio Rsample is maintained. Note that the ratio Rsample may be increased when the load parameter value is significantly below the set threshold value target as in the control slot τi + 2.

[Second control method]
7 and 8 are diagrams for explaining the second control method. As shown in FIG. 7, the setting threshold value target of the load parameter value and its upper limit value HWM and lower limit value LWM are defined in advance, and the control slot If the load parameter value is between the upper limit value HWM and the lower limit value LWM as in τi, the sample ratio Rsample is maintained. On the other hand, when the load parameter value exceeds the upper limit value HWM as in the control slot τi + 1, the sample ratio Rsample is decreased by a predetermined ratio, while when the load parameter value falls below the lower limit value LWM as in the control slot τi + 2, Raise Rsample.

  FIG. 8 is a diagram showing an example of a method for setting the upper limit value HWM and the lower limit value LWM. The upper limit value HWM is set to k (for example, k = 2) times the set threshold value target, and the lower limit value LWM is set to the set threshold value. 1 / k times the target.

[Third control method]
FIG. 9 is a diagram for explaining the third control method. In the above two control examples, the instantaneous value for each control slot is used as the load parameter value. However, here, the moving window of the load parameter value, the moving square average, and the exponential smoothing are set by appropriately setting the time window. The averaged moving average or its standard deviation is obtained, and these are compared with the corresponding set threshold value and its upper and lower limit values HWM and LWM.

[Fourth control method]
FIG. 10 is a diagram for explaining the fourth control method. In the above three control examples, the sample ratio Rsample is controlled based on the measured value of the load parameter value, but here the sample ratio Rsample is controlled based on the predicted value of the load parameter value. As prediction methods, (a) autoregressive model (AR), (b) autoregressive moving average model (ARMA) model, (c) neural network or chaotic time series, etc. may be used. it can.

  (a) The AR model represents a system in which the output at a certain time is obtained as a linear combination of past outputs, and is expressed by the following equation (11)

  Here, y (n) represents the output of the system at time n in discrete time. a (k) is an arbitrary constant and is called an AR parameter because it is a parameter that determines the properties of the AR model. In the above equation (11), since the past p outputs are used, this is called a p-order AR model (or AR (p) model). ε (n) represents an element in which the output of the system cannot be expressed by a linear combination of past outputs, and is called a prediction error. This is an error that occurs because the output of the system is actually a linear combination of the past m (m> p) outputs, whereas the previous p linear combinations are used when modeling this. Another way of looking at ε (n) is also possible. That is, ε (n) is considered as an input to the system, and the output of the system at time n can be regarded as being obtained as a linear combination of the past p outputs and the input 2 (n) at time n.

  (b) The ARMA model represents a system in which the output at a certain time is obtained as the sum of the linear combination of the past output and the current linear combination, and is expressed by the following equation (12). However, the constants a (k) and b (k) are called an AR parameter and an MA parameter, respectively, and these are collectively called an ARMA parameter.

  In the above description, (1) sample traffic volume λsample [τ0], (2) arrival traffic volume to end server ES (sum of sample traffic volume and traffic volume generated by release of hold), (3) hold The total amount of ongoing traffic Σψτi [τ0], and (4) the maximum value of the hold duration, as described in “Load Parameter Value”, as described in detail below. May be different.

[Adopting sample traffic as load parameter value]
When the sample traffic amount λsample [τ0] in the immediately preceding control slot τ0 exceeds the system capacity Cmax and the set threshold value target, in step S37, the sample ratio Rsample [τ1] is calculated based on the following equation (13) as Rsample [τ0]. Is reduced (for example, Rsample [τ0] is multiplied by 0.5). Alternatively, the ratio of the excess amount is decreased.

  In this way, if the sample ratio Rsample is corrected based on the sample traffic amount λsample, it is possible to directly avoid the end server ES from exceeding the system capacity due to the traffic generated by the sample terminal MN (s). If the sample traffic amount λsample [τ0] falls below the system capacity Cmax or the set threshold value target by a certain amount or more, the sample in the control slot τ1 can be corrected by increasing the sample ratio Rsample [τ1] by comparison with Rsample [τ0]. By increasing the traffic amount λsample [τ1], it is possible to directly improve the lack of estimation accuracy of the number of sample terminals.

[Adopting traffic volume to end server ES as load parameter value]
When the traffic amount λ [τ0] arriving at the end server ES in the immediately preceding control slot τ0 falls below the system capacity Cmax or the set threshold value target, in step S37, the sample ratio Rsample [τ1] is calculated based on the following equation (14). Is increased by comparison with Rsample [τ0] (for example, Rsample [τ0] is multiplied by 1.1). Alternatively, the ratio of the amount below is increased.

  As described above, if the sample ratio Rsample is corrected based on the total number of accesses to the end server ES, the sample traffic amount λsample [τ1] in the control slot τ1 increases, resulting in insufficient estimation accuracy of the number of sample terminals. Can be improved directly. Further, if the arrival traffic amount λ [τ0] to the end server ES exceeds the system capacity Cmax or the set threshold value target, the sample terminal MN can be corrected by reducing the sample ratio Rsample [τ1] by comparison with Rsample [τ0]. (s) and the system capacity of the end server ES due to the traffic generated by the non-sample terminal MN can be directly avoided.

[Total traffic volume on hold as the load parameter value (Part 1)]
When the total traffic volume Σiψτi [τ0] during pending suspension falls below the margin capacity φ calculated using the sample capacity equivalent to the system capacity Cmax, the set threshold value target, or the immediately preceding control slot [τ0], that is, pending suspension If there is a remainder in the system capacity even after all the outgoing calls are released, in step S37, the sample ratio Rsample [τ1] is increased and compared with Rsample [τ0] based on the following equation (15). . Alternatively, the ratio of the excess amount is decreased.

  In this way, if the sample ratio Rsample is corrected based on the total number of outgoing calls that are on hold, it is possible to release a larger amount of reserved traffic while maintaining necessary and sufficient estimation accuracy, that is, control accuracy. Appropriate control that can reduce the waiting time of the middle traffic, that is, the waiting time can be realized.

[The total traffic volume on hold is used as the load parameter value (Part 2)]
If the total Σiψτi [τ0] of the pending ongoing traffic amount satisfies the following conditions, the sample ratio Rsample [ Reduce τ1] by comparison with Rsample [τ0].

  Condition 1: If Σiψτi [τ0] exceeds N times the system capacity Cmax (N is a set value), that is, if it takes N slots at the earliest to release the traffic volume currently reserved, Equivalent to

  Condition 2: When Σiψτi [τ0] exceeds the set threshold over_target, the same as above

  Condition 3: If Σiψτi [τ0] assumes that the same amount of sample traffic as that of the previous slot will occur in the future, it will take M slots at the earliest to release the currently reserved traffic amount. Same as above

  Furthermore, the following condition 4 may be further added from the viewpoint of referring to the traffic that is on hold as a load parameter.

  Condition 4: Among the traffic held by Σiψτi [τ0], when the maximum waiting time and the number of waiting slots, that is, the elapsed waiting time and the number of waiting slots after being held and exceeding the allowable maximum waiting time W, In other words, if you wait too long for the amount of pending traffic

  When any of the above four conditions or a single condition defined in advance is satisfied, the sample ratio Rsample [τ1] is reduced and corrected based on the following equation (16). Alternatively, it may be set to the minimum sample ratio Rsample [min] necessary for accuracy and control accuracy when estimating the total traffic amount from the sample traffic amount.

  Note that the minimum sample ratio rs_min necessary for accuracy is determined not by the ratio but by the amount of sample traffic. This depends on the sample accuracy that can be obtained from the law of large numbers in statistics. For example, if you want to maintain an error of 5% as accuracy, a sample amount of 278 is required if the total amount is 1000, 370 is required if it is 10,000, and if a sample amount of 384 is obtained, even if the population is infinite The error is within 5%. From this, it is only necessary to estimate the amount of generated traffic from the past DB and trends, and set rs_min based on the result, or estimate rs_min from the sample traffic amount obtained immediately before.

  In FIG. 11, the number of accesses to the end server ES is appropriately controlled by dynamically controlling the sample ratio Rsample based on the load parameter value of the end server ES (here, the amount of sample traffic) as described above. It is the figure which showed the mode compared with the prior art.

  The sample traffic amount λsample [τ0], which was less than the set capacity up to the control slot τi (for example, set to the same value as half the system capacity Cmax), starts to increase, and the set capacity is set in the next control slot τi + 1. Since the sample ratio Rsample is fixed in the prior art, the sample traffic amount λsample [τ0] continues to increase and exceeds the set capacity in the control slot τi + 1 as shown in FIG. End up.

  On the other hand, in the present invention, when the sample traffic amount λsample [τ0] exceeds the set capacity in the control slot τi + 1, the sample ratio Rsample is decreased, and this is repeated until it falls below the set capacity. As shown in (b), the sample traffic amount λsample [τ0] falls below the set capacity at or after the control slot τi + 2. As a result, not only can congestion caused by the transmission of the sample terminal MN (s) be avoided, but also the sample access can be reduced, and it is generated and held earlier than the current transmission of the sample terminal MN (s). Since more outgoing calls can be processed, more fair access control is possible.

  FIG. 12 is a diagram showing how the number of accesses to the end server ES can be accurately predicted by dynamically controlling the sample ratio Rsample based on the amount of sample traffic in comparison with the prior art. .

  When the sample traffic amount λsample [τ0] that has exceeded the lower limit LWM until the control slot τi starts to fall and falls below the lower limit LWM in the next control slot τi + 1, the sample ratio Rsample is fixed in the prior art. As shown in FIG. 5A, the sample traffic amount λsample [τ0] continues to decrease thereafter. As a result, the sample traffic amount λsample [τ0] falls below the lower limit LWM in the control slot τi + 1, and the accuracy of the total traffic amount λall [τ0] estimated based on the sample traffic amount λsample [τ0] decreases. Resulting in.

  On the other hand, in the present invention, when the sample traffic amount λsample [τ0] falls below the lower limit value LWM in the control slot τi + 1, the sample ratio Rsample is increased. Therefore, as shown in FIG. After +2, the number of transmissions λsample [τ0] exceeds the lower limit LWM. As a result, a sufficient number of sample terminals MN (s) can be secured for estimating the total traffic volume [τ0], and the total traffic volume [τ0] can be estimated with high accuracy.

  The above effect is an effect when the sample ratio Rsample is dynamically controlled by adopting the sample traffic amount λsample [τ0] as a load parameter value representative of the load state and congestion state of the end server ES. The same and unique effects can be obtained when the total number Σψτi [τ0] of transmissions that are kept on hold is used as the load parameter value to dynamically control the sample ratio Rsample.

  That is, if the sample rate Rsample is reduced when the total number of outgoing calls that are on hold exceeds the corresponding set threshold value target, the sample terminal MN (s) occupies the total number of accesses to the end server ES in each control slot τ. ) Access number is reduced, and the access number by canceling the hold of outgoing calls that are kept on hold can be relatively increased. As a result, since it becomes possible to process the outgoing calls that are generated and put on hold earlier than the current transmission of the sample terminal MN (s) earlier and more, fair access control becomes possible.

  Further, when the total number of outgoing calls on hold falls below the lower limit LWM, the sample ratio Rsample is increased, and a sufficient number of sample terminals MN (s) can be secured for estimating the total outgoing number λall [τ0]. Therefore, it is possible to estimate the total number of outgoing calls λall [τ0] with high accuracy while minimizing the unfairness that the outgoing call of the sample terminal MN (s) is processed before the outgoing call that is kept on hold.

  If the total number of outgoing calls Σψτi [τ0] during suspension is adopted as the load parameter value, (1) X times the system capacity Cmax is adopted as the set capacity. The number of control slots) can be shortened. Further, even if (2) the holding duration is adopted, the holding duration can be shortened.

  By the way, in the above formulas (1) to (10), the traffic volume generated and held in each control slot τm is managed by the control server CS, and the hold release rate Rr [ If τ1] is notified, it is assumed that the transmission according to the hold release rate R [τ1] is surely performed in the next control slot. However, in reality, not all user terminals MN whose transmission is put on hold can receive the broadcast wave, and not all user terminals MN on which the hold is released can send, and further the transmission timing. There may be a delay.

  In addition, since the control information notified for hold release is the hold release ratio, a statistical error may occur in the hold release determination. Further, since the release ratio is calculated based on the result of sample estimation in calculating the control information, an error may occur between the calculated value of the traffic volume to be released from the hold and the traffic volume arriving at the end server ES. is there.

  Therefore, there may be a discrepancy between the expected value ρest and the actual value ρreal of the traffic ρ determined from the traffic volume generated and held in each control slot τm and the hold release rate Rr [τ1]. The end server ES access control may be upset.

  On the other hand, in the present invention, a plurality of addresses ADx are virtually allocated to the end server ES, and the hold release timing is described in the control information and broadcasted to each user terminal MN as detailed below. When notifying, the address of the end server ES notified as the destination is different for each control slot, and regardless of the actual transmission timing, the transmission to the address ADr assigned to the control slot where the transmission has occurred is performed. To be forced. Then, the measurement server DS accurately monitors the actual value ρreal of the number of calls for each control slot by monitoring the destination address of each call, and based on this, the number of calls that are kept pending ψτm can be accurately grasped for each control slot τm.

  FIG. 13 is a diagram schematically illustrating a state in which different destination addresses are assigned to each user terminal MN depending on the timing of a call request.

  In the control information broadcast from the broadcasting station BS, ADs [τi-3] (for sample terminals) and ADn [τi-3] (for non-sample terminals) are used as the address of the end server ES in the next control slot τi-3. When notified, the user terminal MN that has made a call request in this control slot τi-3 is assigned the address ADs [τi-3] if it is a sample terminal MN (s), and the non-sample terminal MN (n) If so, the address ADn [τi-3] is assigned. Therefore, the sample terminal MN1 (s) immediately transmits to the address ADs [τi-3].

  Further, although the non-sample terminal MN2 (n) is put on hold in this control slot τi-3, if the hold is released by the next control information, the address ADn [τi newly notified by the control information -2], not to ADn [τi-3].

  Similarly, ADs [τi-2] and ADn [τi-2] are assigned as addresses of the end server ES to the user terminal MN3 that has made a transmission request in the control slot τi-2, so that it is reserved by the next control information. When canceled, the message is transmitted to the ADn [τi-2], not the destination address ADn [τi-1] newly notified by the control information.

  FIG. 14 is a diagram schematically showing the operation of the present embodiment. Here, in order to make the explanation easier to understand, the destination address assigned to the user terminal MN that has issued the transmission request in the control slot τi is represented as a sample terminal. Regardless of whether it is a MN (s) or a non-sampled terminal MN (n), all are expressed as ADs [τi].

  Among the transmissions generated in the control slot τi-3, the sample terminal MN (s) classified by the sample ratio Rsample immediately transmits to the destination address AD [τi-3] in the control slot τi-3 without being suspended. To do. On the other hand, all the remaining non-sample terminals MN are temporarily suspended for transmission, part of which is released from the control slot τi-2 and transmitted, and the other part is released from the control slot τi-1. And the rest is released in the control slot τi and transmitted. At this time, since the transmission requests of all the non-sample terminals MN are generated in the control slot τi-3, the destination address is all AD [τi-3] regardless of the hold release timing.

  FIG. 15 is a diagram showing an example of the monitoring result by the measurement server DS. The number of outgoing calls to the engine server ES can be detected for each destination address, whereby the number of outgoing calls in each control slot can be detected.

  More specifically, the estimated value ρτm_est [τ0] of the number of transmissions ρ that occurs in the control slot τm and is released from the hold in the control slot τ0 is obtained by the following equation (17). Equation (17.1) is applied to all control slots where the slot number m is m ≦ n (n: the maximum value of the control slot number that can be released), and the control slot τ-1 immediately before the control slot τ0 All transmissions ψτm_after [τ-1] that are on hold at the end are released. Note that ψτm_after [τ0] is a value obtained by correcting the number of outgoing calls kept on hold even after the end of the control slot τ0 based on the actual value of the number of outgoing calls released in the control slot τ0. ).

  Equation (17.3) is applied to all control slots whose slot number m is m> n + 1, and all outgoing calls that are on hold are not released in the control slot τ0, and even in the control slot τ0, The same number of outgoing calls will be held on hold. Equation (17.2) is applied to the control slot of slot number m = n + 1, and among all outgoing calls that are held on hold at the end of control slot τ-1, up to slot number n (= n−1) The number of outgoing calls obtained by subtracting the sum of the number of continuations on hold from the free capacity C [τ0] is released.

  In the present embodiment, the access to the end server ES is observed by the measurement server DS, and the destination address to the end server ES differs depending on the timing (control slot) at which the transmission request is generated. In DS, by identifying the destination address of each access, it is possible to obtain the actual value ρτm_real [τ0] of the number of outgoing calls generated by releasing the hold for each control slot. Therefore, if the number of transmissions estimated to be held until just before the control slot τ0 is ψτm_before [τ0], the number of transmissions that are held continuously after the end of the control slot τ0 and released from the hold in the control slot τ0. The number of transmissions ψτm_after [τ0] corrected based on the value is obtained by the following equation (18).

  It should be noted that the correction processes of the above equations (13) and (14) are performed for each control slot τm in which the transmission released from the hold in the control slot τ0 occurs. That is, if there are a plurality of control slots that are released from hold, the correction process is performed for each control slot.

  Then, by applying the hold continuation number ψτm_after [τ0] obtained in the above equation (14) to ψτk [τ0] in the above equation (6), the maximum value n of the control slot numbers that can be released from the hold can be more accurately determined. As a result, the number of communication sessions established between the user terminal MN and the end server ES can be accurately regulated below the system capacity of the end server ES in units of control slots.

  In the above embodiment, the destination address of each call has been described as being different for each timing when the call request is generated, regardless of the timing at which the hold is released, but the present invention is not limited to this. It may be made different for each combination of the generation timing of the outgoing call request and the timing of release of the hold.

  In this way, by calculating the total number of outgoing calls released from the hold, the number of outgoing calls released in the current control slot before and after applying the correction processing of the above formulas (17) and (18) and before that It becomes possible to grasp the difference in the number of outgoing calls released from the hold. As a result, for example, it is possible to detect a deviation from a control slot that should originally be released from hold, such as a delay in transmission timing due to a temporary disconnection of a wireless communication line.

  Further, in the above-described embodiment, the transmission of each user terminal MN has been described as being controlled so that the number of simultaneous connection sessions to the end server ES is equal to or less than the system capacity, but the present invention is limited to this. However, the present invention is equally applicable to the case where the number of accesses per unit time to the end server ES is regulated to be equal to or less than the system capacity.

  MN ... User terminal, ES ... End server, DS ... Measurement server, CS ... Control server, BS ... Broadcast server, BC ... Broadcast station, NW ... Network

Claims (12)

In the outgoing call regulation system that limits the load on the end server by temporarily holding the outgoing call from the user terminal to the end server and then releasing the hold sequentially in units of control slots.
A control server for canceling the hold of outgoing calls at each user terminal;
The user terminal is
Means for receiving a sample probability of classifying a portion of the user terminal into a sample terminal;
Means for classifying the terminal as a sample terminal with the sample probability;
In response to the call operation, a means for making a call if it is a sample terminal and a means for holding off the call if it is a non-sample terminal,
Means for receiving a hold release condition;
Means for making a call if the call on hold satisfies the received hold release condition, and continuing the hold if not satisfied,
The control server
Means for measuring a load parameter value representative of the load status of the end server;
Means for determining a condition for releasing the hold based on the load parameter value;
Means for determining the sample probability based on the load parameter value;
A transmission restriction system comprising: means for notifying each user terminal of the sample probability and the hold release condition.   The transmission restriction system according to claim 1, wherein the means for measuring the load parameter value measures a sample traffic amount for each control slot.   3. The transmission restriction system according to claim 2, wherein the means for determining the sample probability reduces and corrects the sample probability when the amount of sample traffic exceeds a predetermined threshold value.   The means for measuring the load parameter value measures, for each control slot, the sum of the traffic volume of the sample terminal and the traffic volume of the outgoing call that has been released as the traffic volume arriving at the end server. 1. The outgoing call regulation system according to 1.   5. The outgoing call control system according to claim 4, wherein the means for determining the sample probability corrects the sample probability to be increased when the amount of traffic arrived at the end server falls below a predetermined threshold value.   2. The transmission restriction system according to claim 1, wherein the means for measuring the load parameter value measures the total amount of traffic during suspension for each control slot.   7. The means for determining the sample probability based on a load parameter value reduces the sample probability when the sum of the traffic volume on hold exceeds a system capacity of the end server. The outgoing call regulation system described.   The means for determining the sample probability based on a load parameter value decrements the sample probability when the sum of the amount of traffic during suspension exceeds a predetermined threshold value. Outgoing restriction system.   The means for determining the sample probability based on the load parameter value finishes releasing all holdings assuming that the total amount of traffic that has been held continues that the sample traffic generated in the previous control slot will continue in the future. 7. The outgoing call regulation system according to claim 6, wherein the sampling probability is reduced and corrected when a control slot period exceeding a predetermined number is required.   2. The transmission restriction system according to claim 1, wherein the means for measuring the load parameter value measures the maximum value of the hold time for each control slot for traffic that is being kept on hold.   11. The transmission restriction system according to claim 10, wherein the means for determining the sample probability based on a load parameter value reduces and corrects the sample probability when the maximum value of the holding time exceeds a predetermined threshold value. . In the outgoing call control method for temporarily limiting the outgoing call from the user terminal to the end server, and then releasing the hold in order in units of control slots to limit the load on the end server
A control server for canceling the hold of outgoing calls at each user terminal;
The user terminal is
Receiving a sample probability of classifying a part of the user terminal into a sample terminal;
A procedure for classifying the terminal as a sample terminal with the sample probability;
In response to a call operation, a procedure to place a call if it is a sample terminal and hold a call if it is a non-sample terminal,
A procedure to receive the hold release condition,
A call that is put on hold if the received hold release condition is satisfied, and a call is kept if it is not satisfied,
The control server
A procedure for measuring load parameter values representative of the load status of the end server;
A procedure for determining a hold release condition based on the load parameter value;
Determining the sample probability based on the load parameter value;
And a procedure for notifying each user terminal of the sample probability and the hold release condition.

Images