JP2017037471A - Quality controller, and quality control method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To smoothly carry out a series of expansion processing on a server which corresponds to the traffic volume.SOLUTION: A management server 3 controls: to calculate a period of time required for expansion from a start point to a termination point of the expansion on a distribution server 1; to predict a transition of the traffic volume in the calculated expansion period of time; to calculate a traffic volume exceeding a processing capacity before the expansion until the expansion completes in the prediction result; and to reduce the traffic volume equivalent to the exceeding traffic volume from the traffic volume which is processed by the distribution server 1 up to point when the expansion is terminated.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a quality control apparatus and a quality control method.

  In a disaster or event, the amount of communication traffic may increase so rapidly that it exceeds the processing capacity of the system. If the processing capacity of the system is exceeded, the entire system becomes unstable, and existing connections are forcibly disconnected or new connection requests are rejected.

  Therefore, in order to protect the processing capacity of the system, communication control is performed to restrict new connection requests from being accepted from the excess. For example, Non-Patent Document 1 describes that a part of outgoing calls from general telephones that are not prioritized is regulated as “priority communication during a disaster”.

Ministry of Internal Affairs and Communications, "Priority communication during disasters", [online], 2009, [Search August 4, 2015], Internet <URL: http://www.soumu.go.jp/menu_seisaku/ictseisaku/net_anzen/ hijyo / yusen.html>

  The communication restriction that rejects a new connection request only protects the system as a temporary measure, and cannot provide a service to the user who makes the connection request, resulting in a decrease in long-term user satisfaction. Therefore, we will consider the case of dealing with the long-term increase in traffic volume by increasing the processing capacity of the system such as adding servers.

  An expansion period from the start of expansion to the completion of expansion occurs when adding a server. First, if the extension period is delayed, a situation may occur in which the traffic volume that has increased rapidly is not processed before the processing capacity of the system is increased. On the other hand, if the expansion period is set early, the number of servers that are idle due to excessive expansion increases, resulting in a decrease in server resource utilization and an extra cost increase.

  Thus, the main object of the present invention is to smoothly realize the server expansion process corresponding to the traffic volume.

In order to solve the above problems, the quality control apparatus of the present invention provides:
A communication status collection unit that collects changes in the amount of traffic processed by the traffic processing system;
An extension that determines that the processing capacity of the traffic processing system is to be increased when the traffic volume collected by the communication status collection unit exceeds the expansion threshold, and obtains an expansion period from the start of the expansion to the end of the expansion A processing unit;
The extension processing unit predicts the change in traffic volume during the extension period, finds the traffic volume that exceeds the processing capacity before the extension by the time the extension is completed, and suppresses the traffic corresponding to the excess traffic volume. And a suppression processing unit that controls the traffic volume to be suppressed from the traffic volume processed by the traffic processing system by the time when the addition is completed.

  As a result, by estimating in advance the amount of traffic that will be exceeded during the expansion period and suppressing it by that amount of traffic, it is possible to smoothly realize the server expansion processing that matches the traffic volume.

  According to the present invention, when the current traffic volume collected by the communication status collection unit exceeds a suppression threshold based on the processing capacity before the addition, the suppression processing unit separates from the traffic volume to be suppressed by a predetermined amount. It is characterized by suppressing the amount of traffic.

  As a result, the traffic processing system can be protected against a sudden increase in traffic volume that falls outside the range of traffic volume estimated in advance.

  The present invention is characterized in that the suppression processing unit determines a suppression period and a non-suppression period so that the buffer amount of the traffic volume that has already been transmitted falls within a predetermined range for the traffic volume to be suppressed. And

  As a result, the buffer amount is consumed even while the traffic volume is being suppressed, so that the user experience does not need to be felt.

In the present invention, the traffic processing system processes traffic volumes of a plurality of users,
The suppression processing unit equally divides the amount of traffic to be suppressed into a plurality of users and obtains the traffic amount of each user to be suppressed.

  Thereby, it can prevent that dissatisfaction concentrates on a specific user by suppressing traffic volume equally with respect to a some user.

In the present invention, the traffic processing system processes traffic volumes of a plurality of users,
The suppression processing unit obtains the traffic volume of each user to be suppressed in proportion to the traffic volume of each user when the traffic volume to be suppressed is allocated to a plurality of users.

  Accordingly, by suppressing the traffic volume of each user according to the traffic volume of the service currently used, any user can receive a minimum service after the suppression.

  According to the present invention, it is possible to smoothly implement the server expansion process corresponding to the traffic volume.

It is a block diagram of the quality control system concerning this embodiment. It is a block diagram of the management server concerning this embodiment. It is a flowchart of the quality control process concerning this embodiment. FIG. 4A is a graph showing a situation where the number of users being added is exceeded. FIG. 4B is a graph showing a result of predicting the number of users at the expansion start time. FIG. 5A is a graph showing the excessive user restriction process in the conventional technique. FIG.5 (b) is a graph which shows the advance process of an expansion period in the prior art. FIG. 6A is a graph showing a prediction-based quality control process. FIG. 6B is a graph showing an actual measurement-based quality control process. It is a graph which shows the quality control process which paid its attention to the buffer amount of the viewing terminal concerning this embodiment.

  Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a configuration diagram of a quality control system. The quality control system is configured by connecting a distribution server 1, a viewing terminal 2, and a management server 3 via a network.
Each device of the quality control system is configured as a computer having a CPU (Central Processing Unit), a memory, storage means (storage unit) such as a hard disk, and a network interface.
In this computer, the CPU executes a program (also referred to as an application or its abbreviated application) read on the memory, thereby operating a control unit (control means) configured by each processing unit described later.
In the following, a content distribution system that distributes content is illustrated as an example of a quality control system, but the present invention can be applied to any traffic processing system in which the traffic volume varies with time, such as a transaction processing system of a bank.

  Each distribution server 1 distributes viewing data traffic to each viewing terminal 2. For example, in the live streaming distribution, the viewing request from each viewing terminal 2 and the traffic accompanying the request increase only in a specific time period of the distribution program. The specific time zone is, for example, a program start time or an appearance scene time of an artist who is a main feature. In addition, since live streaming distribution is valuable for “watching now”, it is difficult to distribute the load by greatly shifting the distribution time of each user as in the case of recording distribution.

  Therefore, the management server (quality control device) 3 manages each distribution server 1 so that the system size can be flexibly expanded (addition of the distribution server 1) in response to insufficient system performance due to an increase in traffic. The addition of the distribution server 1 may be a physical extension in which a worker newly installs and connects a chassis, or a logic that makes it possible to newly use virtualized computer resources by using cloud or virtualization technology. It may be an additional number.

However, even if it is a physical expansion or a logical expansion, it takes a time in minutes even if the expansion period of the distribution server 1 is the shortest, so after observing that the traffic volume of the distribution server 1 has increased. Even if the start of expansion is instructed, the system performance may be insufficient before the expansion period is completed.
Therefore, the management server 3 uses both a long-term means of adding the distribution server 1 and a short-term means of controlling the traffic volume during the extension period as means for solving the system performance shortage with respect to the increasing traffic volume. .

  The management server 3 reduces the quality of the service so as to suppress the traffic amount already provided to the distribution server 1 in the traffic amount suppression control. Here, since it is sufficient to extend only the expansion period in which the system performance is insufficient, the management server 3 obtains an excess amount of the traffic amount due to the insufficient system performance in the expansion period, and only the excess amount (only the necessary amount) is obtained. By limiting the quality degradation, it is possible to protect the system operation of the distribution server 1 and make the user of the viewing terminal 2 feel less degradation in service quality.

FIG. 2 is a configuration diagram of the management server 3.
The management server 3 includes a communication status collection unit 31, a suppression determination unit 32, a suppression instruction unit 33, an expansion determination unit 34, and an expansion instruction unit 35.
The suppression processing unit includes a suppression determination unit 32 and a suppression instruction unit 33.
The expansion processing unit includes an expansion determination unit 34 and an expansion instruction unit 35.

  The communication status collection unit 31 collects information on traffic distributed by the distribution server 1 to the viewing terminal 2 (S11). The collected traffic information includes, for example, traffic such as the number of users of the viewing terminal 2 that receives the traffic, the CPU usage rate, the memory usage rate, and the network interface usage rate in addition to the transition of the traffic amount flowing to the viewing terminal 2 Resource utilization rate used for transmission of

  When each parameter value of the traffic information collected by the communication status collection unit 31 exceeds a predetermined threshold (suppression threshold) indicating a level at which there is a high possibility that a traffic drop will occur, the suppression determination unit 32 A suppression mode for instructing the control of the traffic amount to 33 is turned on (S12). Thereafter, when each parameter value of the collected traffic information falls below a predetermined threshold (suppression threshold), the suppression mode is returned to the suppression instruction unit 33.

The suppression instruction unit 33 instructs the distribution server 1 to suppress the traffic volume, triggered by the suppression determination unit 32 turning on the suppression mode (S13). The trigger for instructing S13 may be “when the suppression mode is on and during the expansion period” or “when it is during the expansion period”.
Therefore, the suppression instruction unit 33 calculates the specific instruction content of S13 for which user when and how much traffic volume is suppressed.

The extension determination unit 34 exceeds the predetermined threshold value (addition threshold value) indicating that each parameter value of the traffic information collected by the communication status collection unit 31 is likely to cause a shortage of system performance due to a future increase in traffic volume. If so, the extension instruction unit 35 is instructed to add the distribution server 1 (S14).
The expansion instruction unit 35 newly adds the distribution server 1 according to the instruction of S14 (S15). Here, the extension instruction unit 35 notifies the suppression instruction unit 33 of information for specifying the extension period such as the start point of expansion or the end point of the extension (S16), so that the suppression instruction unit 33 supports the instruction of S13. To do.

FIG. 3 is a flowchart of the quality control process. 3 will be described with reference to S11 to S16 of FIG. 2 as appropriate.
As S101, the suppression determination unit 32 turns on the suppression mode in S12.
As S102, the suppression instruction unit 33 prepares in advance the suppression content (suppression method) specified in S13. Details of this advance preparation will be made clear in the explanation after FIG.
As S111, the extension instruction unit 35 determines whether or not the present time is during the extension period according to the instruction of S16. If Yes in S111, the process proceeds to S121, and if No, the process proceeds to S112.

As S112, the suppression determination unit 32 determines whether or not the communication status is improved based on the current traffic amount collected by the communication status collection unit 31 before the start of the addition of S114. When the traffic volume falls below the suppression threshold value in S12 (S112, Yes), the suppression mode is turned off. On the other hand, if No in S112, the process proceeds to S113.
In S113, the extension determination unit 34 determines whether or not the extension instruction in S14 is necessary. If Yes in S113, the process proceeds to S114, and if No, the process proceeds to S121.
As S114, the expansion instruction unit 35 instructs the expansion of the distribution server 1 (S15), and starts the expansion. At the same time, the extension instruction unit 35 notifies the suppression instruction unit 33 of information for specifying the extension period (S16).

As S <b> 121, as described with reference to FIG. 1, the suppression instruction unit 33 starts suppression by distributing an excess amount of traffic volume due to insufficient system performance (suppressed traffic volume) among suppression target users.
The suppression target users are, for example, existing users who have already been connected to the distribution server 1, new users who are going to connect to the distribution server 1, or both existing users and new users accommodated by the distribution server 1 (all users). There are various types of services such as a user accommodated in the distribution server 1 in charge of a predetermined area (location), a paying member user, and a free member user.
The traffic of the suppression target user may be, for example, a call connected by the user or a user transaction.

In addition, the traffic volume for the suppression per one suppression target user can be calculated | required by either of the calculation formulas illustrated below, for example.
・ (Restricted amount per user) = (Fixed value not depending on the traffic amount of suppressed amount)
・ (Restricted amount per user) = (Restricted traffic volume) ÷ (Number of users targeted for suppression)
(Suppressed amount per user) = (Suppressed traffic amount) ÷ (Number of users to be suppressed) × (Multiplier based on ratio of distribution rate per user). This multiplier is proportional to the distribution rate for each user to be suppressed, for example. That is, as the user has a larger traffic volume, a larger traffic volume is suppressed.

  In addition, for example, when the traffic volume suppression control from the distribution server 1 to the viewing terminal 2 based on the suppression traffic volume is to suppress the video data traffic volume, the distribution server 1 reduces the distribution rate. In this process, the encoded video is encoded to the viewing terminal 2 after being encoded at the compression rate.

As S122, the communication status collection unit 31 determines whether or not the current communication status improvement has been improved, similar to S112. If Yes in S122, the suppression of S122 is terminated (S123) and the suppression mode is turned off. On the other hand, if No in S122, the process proceeds to S131.
As S131, the extension instruction unit 35 confirms the completion of the extension started in S114. As S132, the suppression instruction unit 33 confirms the end of suppression started in S121, and returns the process to S111.

  Hereinafter, with reference to FIGS. 4 to 7, specific instruction contents of the suppression instruction unit 33 in S <b> 13 as to when and how much traffic volume is suppressed for which user will be described.

FIG. 4A is a graph showing a situation where the number of users being added is exceeded.
The communication status collection unit 31 collects the actual user number transition N [t]. The suppression determination unit 32 predicts the expected number of users transition Nz [t] based on N [t]. Both Nz [t] and N [t] exceed the user number increase threshold Nth at t11. Therefore, the extension determination unit 34 determines the extension start time t11 and sets the period from the t11 to the extension end time t12 as the extension period.

Here, the system processing capacity of all the distribution servers 1 is the number Nca of users that can be processed before expansion until t12, and increases from the number of users Ncb that can be processed after expansion from t12. As indicated by Nz [t], when the number of users changes so as not to exceed the system processing capacity, there is no problem if only the number of users is increased.
However, as indicated by N [t], when the number of users exceeding the system processing capacity occurs during the expansion period (time t2 to t12), there is a possibility that the traffic corresponding to the excess number of users from Nca may not be processed. (In the figure, the number of unprocessed users is indicated by area Ne).

FIG. 4B is a graph showing a result of predicting the number of users at the expansion start time.
The suppression determination unit 32 predicts a change in the number of users during the expansion period at time t11. The prediction result is, for example, a half line starting from the graph point P1 (t11, Nth), and becomes N1 [t], N2 [t], and N3 [t] in descending order of the gradient. The slope of the graph indicating the increase rate of the user can be obtained from the transition of N [t] (difference in time series) before the prediction time t11.

Here, as described below, the suppression determination unit 32 predicts an excess from the processing capacity based on the processing capacity of the traffic volume before and after the suppression period and the traffic volume transition in the suppression period. The point P2 (t12, Nca) through which N2 [t] passes is a point for determining whether or not the number of users exceeds the processing capacity during the expansion period. That is, it is determined that N1 [t] whose slope is larger than N2 [t] (increase in the number of users is large) exceeds the processing capacity. On the other hand, N3 [t] whose inclination is smaller than N2 [t] is determined not to exceed the processing capability.
For N1 [t] determined to exceed the processing capacity, the area Ne indicating the number of excess users is approximated as an area of a triangle formed by three vertices (P2, P3, P4). Note that the point is P3 (t2, Nca) and the point P4 (t12, N1 [t12]). The point P5 (t12, Nth) is used in the description of FIG.

  FIG. 5A is a graph showing the excessive user restriction process in the conventional technique. The transition of the number of users before regulation is Na [t], and the transition of the number of users after regulation is Nb [t]. During the excess period (t2 to t12), the number of users exceeds the processing capacity (Na [t2] to Na [t12]> Nca). Therefore, the distribution server 1 restricts not to accept new connections during the excess period (t2 to t12) (Nb [t2] to Nb [t12] = Nca), thereby replenishing the processing capacity after the expansion. Avoid processing failures due to not being in time at the moment. However, the forced regulation of not accepting new connections loses the opportunity for new users.

FIG.5 (b) is a graph which shows the advance process of an expansion period in the prior art. Hereinafter, description will be made while comparing FIG. 4 (a) and FIG. 5 (b). In FIG. 5B, by setting the number-of-users increase threshold Nth lower than that in FIG. 4A, the extension determination unit 34 can detect the initial increase in traffic volume and advance the extension period (before t11). t31 before t31, t12).
Thereby, the expansion end time t32 can be changed before N [t] increases to a level exceeding the processing capacity. However, as a side effect of advancing the extension period, if it changes as Nz [t], an excessive margin of system processing capacity is created, and resource utilization efficiency decreases.

FIG. 6A is a graph showing a prediction-based quality control process. As a difference from FIG. 4 and FIG. 5, the Y axis is changed from the number of users to the traffic volume. The conversion from the number of users to the traffic volume can be obtained by, for example, the calculation formula “(traffic volume) = (number of users) × (average delivery rate per user)”.
The processing capacity of the management server 3 is the pre-expansion processable traffic volume Tca until the expansion end time t12, and increases from the expansion end time t12 to the post-expansion processable traffic volume Tcb. For example, Tca = Nca × (average delivery rate per user).
In the traffic volume transition Ta [t] before suppression, an excess (area of the triangle Te surrounded by the vertices P2, P3, and P4) occurs in the period (t2 to t12) in which the pre-addition processable traffic volume Tca is exceeded. End up.

The post-suppression traffic amount transition Tb [t] executes quality control processing so that a traffic amount lower than Ta [t] flows in the excess period (t2 to t12). Here, the amount of suppression of Tb [t] (the area of the triangle Tf surrounded by the vertices P1, P2, and P3) needs to be equal to or larger than the area of the triangle Te. In order to minimize the deterioration of the experience, it is desirable to set Tb [t] so that the area of the triangle Tf = the area of the triangle Te.
In this way, the excess amount Te during the expansion period is estimated in advance, and the suppression amount Tf that is the same amount as the excess amount Te is provided, so that the quality to be reduced is minimized by calculating backward from the estimated completion time of the expansion. It is done.

FIG. 6B is a graph showing an actual measurement-based quality control process. Even if the excess is suppressed in advance on the prediction basis of FIG. 6A, an unexpected unexpected increase in traffic volume may occur.
Therefore, in order to protect the system of the distribution server 1 from a sudden increase in traffic volume, a traffic volume suppression threshold value Ttb is provided between Tth and Tca to urgently suppress traffic in the range of Ttb to Tca. A “suppression zone” that is necessary is distinguished from a “normal zone” outside the suppression zone.
Then, when Ta [t] transitions within the “suppression zone”, the suppression instruction unit 33 does not exceed Tca, so that the predetermined amount of traffic is reduced to Tb [t]. The quality control process is forcibly executed separately from the suppression amount Tf obtained in (1). In other words, the suppression zone is an emergency area where it is necessary to forcibly reduce the traffic volume.

FIG. 7 is a graph showing quality control processing focusing on the buffer amount of the viewing terminal 2.
The suppression instruction unit 33 may dynamically change (the suppression amount per user) described in S121 according to the buffer amount of one user.
The buffer amount transition B [t] is the amount of data (buffer amount) stored in a given terminal among the viewing terminals 2 when attention is paid to that terminal. Note that the buffer maximum capacity Bc, which is the maximum value of B [t], is determined in advance depending on the size of the memory of the viewing terminal 2 and the like.
As the time elapses, the viewing terminal 2 digests the received traffic by reading (reading) the buffer amount and reproduces the result on the viewing screen. That is, B [t] decreases every time it is read for reproduction, but increases every time it is received from the distribution server 1.

The suppression instruction unit 33 determines the current traffic amount while referring to the buffer amount according to the following rules.
When the buffer amount is large (B [t]> buffer amount upper limit threshold Btb), the traffic amount is decreased (a suppression process is started at time t41).
When the buffer amount is small (B [t] <buffer amount lower limit threshold Bta), the traffic amount is increased (for example, the suppression process is interrupted at time t42).
Reduce the traffic volume so that the total sum of all users for the amount of traffic reduction (suppression amount) is equal to or greater than the excess amount Te in FIG.
That is, the suppression instruction unit 33 determines a period during which the traffic volume is suppressed and a period during which the traffic volume is not suppressed so that the buffer amount B [t] falls within a predetermined range (buffer amount upper limit threshold Btb to buffer amount lower limit threshold Bta). .

  As a result, even when the traffic volume is being suppressed, the buffer volume is digested and played back, so that the user's viewing experience does not deteriorate. Furthermore, the quality suppression periods exemplified in t41 to t42 may be provided so as to overlap each other by a plurality of users (parallel processing), or may be provided so as not to overlap (so as to be regulated in order of users). (Serial processing).

In the present embodiment described above, even if the number of viewing terminals 2 suddenly increases, the suppression instruction unit 33 protects the distribution server 1 and the service of the viewing terminal 2 during the period until the distribution server 1 is added. An appropriate suppression amount Tf is calculated so that the deterioration in quality is as small as possible (see FIG. 6A).
As a result, in the streaming distribution system, although the distribution rate of the entire system is slightly reduced, it is possible to overcome the unstable period until the expansion period is completed while suppressing the influence of the decrease in the distribution rate.

1 Distribution server 2 Viewing terminal 3 Management server (quality control device)
31 Communication Status Collection Unit 32 Suppression Determination Unit 33 Suppression Instruction Unit 34 Extension Determination Unit 35 Extension Instruction Unit

Claims (8)

A communication status collection unit that collects changes in the amount of traffic processed by the traffic processing system;
An extension that determines that the processing capacity of the traffic processing system is to be increased when the traffic volume collected by the communication status collection unit exceeds the expansion threshold, and obtains an expansion period from the start of the expansion to the end of the expansion A processing unit;
The extension processing unit predicts the change in traffic volume during the extension period, finds the traffic volume that exceeds the processing capacity before the extension by the time the extension is completed, and suppresses the traffic corresponding to the excess traffic volume. And a suppression processing unit that controls the traffic volume to be controlled from the traffic volume processed by the traffic processing system by the time when the addition is completed. The suppression processing unit, when the current traffic amount collected by the communication status collection unit exceeds a suppression threshold based on the processing capacity before the addition, a predetermined amount of traffic separately from the suppression traffic amount The quality control device according to claim 1, wherein the quality control device is suppressed. The suppression processing unit determines a suppression period and a non-suppression period so that the buffer amount of the traffic volume that has already been transmitted falls within a predetermined range for the traffic volume to be suppressed. The quality control apparatus according to 1. The traffic processing system processes the traffic volume of a plurality of users,
The said suppression process part calculates | requires the traffic amount of each user who divides | segments into the some user the part of the said traffic amount to suppress, and is characterized by the above-mentioned. The quality control device described. The traffic processing system processes the traffic volume of a plurality of users,
The said suppression process part calculates | requires the traffic volume of each user to suppress in proportion to the traffic volume of each user, when allocating the part of the said traffic volume to a some user. The quality control apparatus of any one of Claim 3. The quality control device has a communication status collection unit, an expansion processing unit, and a suppression processing unit,
The communication status collection unit collects changes in traffic volume processed by the traffic processing system,
The extension processing unit determines that the processing capacity of the traffic processing system is to be increased when the traffic amount collected by the communication status collection unit exceeds the expansion threshold, and from the time when the expansion starts to the time when the expansion ends. Seeking the extension period of
The suppression processing unit predicts a traffic volume transition during the expansion period, and obtains a traffic amount that exceeds the processing capacity before the expansion by the time when the expansion is completed in the prediction result. A quality control method characterized by controlling the amount of traffic to be suppressed corresponding to the amount of traffic to be suppressed from the amount of traffic processed by the traffic processing system by the time when the addition is completed. The suppression processing unit, when the current traffic amount collected by the communication status collection unit exceeds a suppression threshold based on the processing capacity before the addition, a predetermined amount of traffic separately from the suppression traffic amount The quality control method according to claim 6, wherein the quality control method is suppressed. The suppression processing unit determines a suppression period and a non-suppression period so that the buffer amount of the traffic volume that has already been transmitted falls within a predetermined range for the traffic volume to be suppressed. 6. The quality control method according to 6.

Images