WO2016075771A1 - Computer system and autoscaling method for computer system - Google Patents

Abstract

A computer system having a varying load, wherein computer resources are appropriately allocated according to the load even when load variations are difficult to predict and even when different computer resource allocation methods are effective for different application conditions. In order to accomplish this feature, the computer system is at least provided with: a physical server which has computer resources including a processor, a storage device, and a communication interface; a virtualization device which virtualizes computer resources and allocates virtual computer resources to a plurality of virtual environments; and a management device which manages the plurality of virtual environments. The management device comprises: a monitoring unit which monitors loads on the plurality of virtual environments; a control unit which determines, based on the monitored loads, whether it is necessary to add another virtual environment; and a configuration changing unit which requests addition of a virtual environment in accordance with the determination made by the control unit, removes usage restrictions on computer resources that constitute existing virtual environments, and, upon completion of the addition of a virtual environment, restores the usage restrictions on the existing virtual environments.

Description

Computer system and autoscale method in computer system

The present invention relates to a computer system that changes allocation of computer resources according to a load, and an autoscaling method in the computer system.

In a computer system used in a company, the load varies depending on time and time. In order to maintain a service level such as system response time even when the load fluctuates, a technique for dynamically controlling the allocation of resources (computer resources) has been proposed.
For example, in Patent Document 1, in a virtual environment, a future load fluctuation is predicted for a business system configured by a plurality of business clusters, and a plurality of hardware resources are allocated to the predicted result. A technique for allocating resources necessary to follow load fluctuations by combining means for changing is disclosed.
Autoscale automatically determines the number of virtual machines under the load balancer (load balancer) and the amount of resources allocated to the virtual machine according to the load on the virtual machine, such as CPU usage and network traffic. It is a function to increase or decrease.

Japanese Patent No. 5332065

However, it may be difficult to predict the load fluctuation in the computer system. For example, in the case of an event ticket online sales system or the like, a great deal of access that exceeds a prior prediction may occur and concentrate. In the prior art, when a resource addition method such as adding a VM is scheduled based on the prediction of load increase, and there is a load exceeding the prediction, there is a risk that the addition of the VM will not be in time and service performance will be degraded.

Also, the effective resource addition method may differ depending on the application status in the computer system. For example, in a system consisting of one interconnected Web server and one DB server, if the allowable number of simultaneous connections in the DB server is 100 and the number of simultaneous connections for the DB server in the Web server is 100, add one Web server Even so, since the total number of simultaneous connections for the DB server exceeds the allowable number of simultaneous connections in the DB server, the service performance may be deteriorated.

In order to solve these problems, a first object of the present invention is to auto-allocate computer resources dynamically and maintain the service level of business even when it is difficult to predict load fluctuations. The purpose is to provide technology related to scale.
A second object of the present invention is to provide a technology relating to autoscaling for appropriately allocating computer resources appropriately according to the state of an application in a computer system and maintaining a service level of business. It is in.

A computer system according to the present invention includes a physical server having computer resources including a processor, a storage device, and a communication interface, a virtualization device that virtualizes computer resources and assigns them to a plurality of virtual environments, and a management that manages the plurality of virtual environments. A monitoring unit that monitors the loads of a plurality of virtual environments, a control unit that determines whether or not a virtual environment needs to be added according to the monitored loads, and a virtual unit that is based on the determination of the control unit A configuration that instructs the virtualization device to add an environment, cancels the use restrictions on the computer resources that make up the existing virtual environment, and resets the use restrictions that have been released after the addition of the virtual environment is complete to the existing virtual environment And a change unit.

According to the present invention, even when there is no load fluctuation prediction information, and similarly, depending on the state of the application in the computer system, the dynamic allocation of the computer resources is effectively performed, and the service level of the work is reduced. Autoscale to maintain can be realized.

It is a block diagram which shows the structure of the computer system which concerns on Example 1 of this invention. It is a figure which shows the logical connection relationship and role of a virtual server. It is a figure which shows system construction request information. It is a figure which shows system configuration information. It is a figure which shows physical server performance information. It is a figure which shows service performance information. It is a figure which shows system performance model information. It is a figure which shows scaling rule information. It is a figure which shows the monitoring flowchart which concerns on Example 1 of this invention. It is a figure which shows the scale-out flowchart which concerns on Example 1 of this invention. It is a figure which shows connection information. It is a figure which shows the parameter determination flowchart which concerns on Example 2 of this invention.

Hereinafter, as an embodiment of the present invention, Example 1 and Example 2 will be described with reference to the drawings.

FIG. 1 is a block diagram illustrating a configuration of a computer system according to the first embodiment of the present invention.
The computer system includes a management unit 11, a business unit 13, and a network 16.

The management unit 11 includes a control unit 18 and a management information group 19. The control unit 18 includes a system monitoring unit 181, a system configuration calculation unit 182, a system configuration change unit 183, and a management software control unit 184. The management information group 19 includes system construction request information 191, system configuration information 192, physical server performance information 194, service performance information 195, system performance model information 196, and scaling rule information 197.

The management unit 11 is executed by the physical server 12. The physical server 12 includes a CPU 121, a storage device 122, and a communication IF 123. Each function is realized by the CPU 121 executing the control unit 18 stored in the storage device 122. The storage device 122 also stores the management information group 19. The physical server 12 is connected to the network 16 via the communication IF 123.

The business unit 13 is one or more in the computer system, and includes one or more virtual servers (hereinafter may be referred to as “VM”). FIG. 1 illustrates an example in which three virtual servers of the virtual server 131, the virtual server 132, and the virtual server 133 operate as one business unit. The virtual server plays various roles depending on the work executed in the computer system. For example, the virtual server 131 operates as an LB (load balancer, load balancer), the virtual server 132 operates as a Web server, and the virtual server 133 operates as a DB server.

The hypervisor 15 virtualizes the computer resources of the physical server 14 and plays a role of providing a virtual server in the business unit 13. The physical server 14 includes a CPU 141, a storage device 142, and a communication IF 143. The business unit 13 is stored in the storage device 142 and executed by the CPU 141. The business unit 13 is connected to the network 16 via the communication IF 143.

FIG. 2 is an explanatory diagram illustrating an example of a logical connection relationship and roles of virtual servers in the business unit 13. The LB 21 operates as a load balancer and distributes the received request message to any one of the Web 22, the Web 23, and the Web 24. Web22, Web23, and Web24 are all Web servers. The DB 25 is a DB server and accepts access from the Web 22, Web 23, and Web 24.
The Webs 22 to 24 are set as the initial number of Webs (initial number) of Webs that receive the request messages distributed from the LB 21 and are increased or decreased as necessary (the dotted line portion in FIG. 2 indicates an increase). Show time).

FIG. 3 is an explanatory diagram showing an example of the system construction request information 191. As shown in FIG. The system construction request information 191 includes a role 302 indicating the role of the virtual server (VM) deployed in the business unit 13, a CPU core allocation number 303 indicating the CPU core allocation number for each role, and a CPU operating frequency per core. It includes an initial number 305 indicating the core frequency 304 and the number of initially deployed units for each role. These values are set by the administrator.

The system configuration changing unit 183 deploys the VM in the business unit 13 in accordance with the system construction request information 191 before step S1001 (FIG. 9) of the monitoring flowchart described later starts.
However, in the case of a VM having the role shown in the capping target 902 of scaling rule information (FIG. 8) to be described later, the CPU core allocation number and the capped frequency are calculated and set according to the following formula. Here, capping means limiting (using an upper limit) the usage rate (usage amount) of computer resources (resources).
CPU core allocation number to be set = CPU core allocation number 303 / (1-capping ratio 903)
Capped frequency to be set = CPU core allocation number to be set × 1 core frequency 304 × capping ratio 903

On the other hand, if the VM does not have the role indicated by the capping target 902, the CPU core allocation number and the capped frequency are calculated according to the following formula, and the values are set.
CPU core allocation number to be set = CPU core allocation number 303
Set capped frequency = 0

FIG. 4 is an explanatory diagram showing an example of the system configuration information 192. The system configuration information 192 includes a server name 401 that is an identifier for identifying each VM, a role 402 that indicates the role of the VM, a CPU core allocation number 403 that indicates the number of CPU cores allocated to the VM, and a CPU operation per core. 1 core frequency 404 indicating the frequency, capped frequency 405 indicating the capped CPU operating frequency, and operation destination server 407 which is an identifier of the physical server 12 on which the VM operates. The system configuration information 192 is set by the control unit 18 in accordance with the contents of the system configuration change.

FIG. 5 is an explanatory diagram showing an example of physical server performance information. The physical server performance information 194 includes a date and time 601 indicating the date and time when performance is monitored, a server name 602 that is an identifier of the monitored physical server, a CPU usage rate 603 that indicates the CPU usage rate of the monitored physical server, and a memory that indicates the memory usage rate. It includes a usage rate 604, a network flow 605 indicating the communication flow of the network, and a disk flow 606 indicating the input / output amount of the disk. The system monitoring unit 181 monitors the performance of the physical server 14 and records the monitored performance information in the physical server performance information 194.

FIG. 6 is an explanatory diagram showing an example of service performance information. The service performance information 195 includes a date and time 701 indicating the date and time when performance is monitored, a request number 702 indicating the number of requests received by the business unit 13 per unit time, and a response time 703 indicating a response time (response time) to the request. . The system monitoring unit 181 monitors the performance of the business unit 13 and records the monitored performance information in the service performance information 195.

FIG. 7 is an explanatory diagram showing an example of system performance model information. The system performance model information 196 includes a total resource amount 801 indicating the total number of CPU cores of the system and a performance 802 indicating the number of requests that can be processed by the system when the total resource amount 801 is present. These values are set by the administrator. Before starting step S1001 (FIG. 9) of the monitoring flowchart to be described later, the administrator determines a value shown in the system performance model information 196 by a method such as a verification experiment or literature reference, and stores it in the system performance model information 196. Set its value.

FIG. 8 is an explanatory diagram showing an example of scaling rule information. The scaling rule information 197 determines a capping target 902 indicating the role of the VM to be CPU capped, a capping ratio 903 indicating the ratio of capping set in the VM, and whether to execute request allocation to the newly added VM. Load increase monitoring period 904 indicating a monitoring period to perform, trigger coefficient 906 related to a trigger condition for determining whether to start addition processing of VM, normal time monitoring interval 907 which is a normal time monitoring interval, and high It includes a high load monitoring interval 908 that is a monitoring interval at the time of loading. These values are set by the administrator.

FIG. 9 is a flowchart (monitoring flowchart) showing a monitoring process for the business unit 13 executed by the control unit 18.
In step S1001, the system monitoring unit 181 monitors the business unit 13 at the interval set as the normal monitoring interval 907 (FIG. 8), and records the monitored performance information result in the service performance information 195 (FIG. 6). . Specifically, the monitored date and time 701, the number of requests 702 received by the business unit 13 per unit time, and the response time 703 for the request are recorded. Further, the system monitoring unit 181 monitors the physical server 14 and records the result of the monitored performance information in the physical server performance information 194 (FIG. 5). Specifically, the monitored date / time 601, server name 602, CPU usage rate 603, memory usage rate 604, network flow rate 605, and disk flow rate 606 are recorded.

Next, in step S1002, the management software control unit 184 determines whether the scale-out trigger condition is satisfied according to the following equation.
Number of Requests> Trigger Factor × System Capacity Here, the number of requests is the latest number of requests 702 recorded in the service performance information 195 (FIG. 6). The trigger coefficient is the trigger coefficient 906 of the scaling rule information 197 (FIG. 8). The system capacity searches the system performance model information 196 (FIG. 7) using the total number of CPU core allocations of the system as a key, and uses the value of the corresponding performance 802 (the number of requests that can be processed by the system). The total CPU core allocation number of the system is the total CPU core allocation number 403 of the record whose role 402 is “Web” in the system configuration information 192 (FIG. 4).

If the management software control unit 184 determines that the trigger condition is satisfied in step 1002 (YES), the management software control unit 184 executes the processing after step S1003 as the scale-out processing. When it determines with not satisfy | filling (NO), the process from step S1001 is repeated.

FIG. 10 is a flowchart (scale-out flowchart) showing processing executed by the control unit 18 as scale-out processing after step S1003 in FIG.
In step S1101, the system configuration changing unit 183 starts processing for adding a virtual server (VM) to the business unit 13 in accordance with information recorded in the system construction request information 191 (FIG. 3). Specifically, in the system construction request information 191 (FIG. 3), the value of the CPU core allocation number 303 and the 1 core frequency 304 of the record whose role 302 is “Web” is used, and the virtual server on the hypervisor 15 is used. The process of deploying (making it available) (VM) is started. However, for the purpose of leaving a surplus capacity when the load increases, the CPU core allocation number set in the virtual server (VM) is changed from 1 to the CPU core allocation number 303 from 1 to the capping ratio 903 (scaling rule information 197 in FIG. 8). ) Minus the value used. For example, when the CPU core allocation number 303 is 4 and the capping ratio 903 is 50%, the CPU core allocation number set in the virtual server (VM) is 8 (= 4 / (1-0.5) = 4. /0.5). Further, the capping rate set for the virtual server (VM) to be added is the capping rate 903 corresponding to the role of the VM to be added (capping target 902 in FIG. 8) from the scaling rule information 197 (FIG. 8) (for example, (If the role is “Web”, the capping rate is 50%).

In step S1102, the system configuration change unit 183 executes a process of canceling CPU capping of the operated virtual server. The operated virtual server is found by searching for a record whose role 402 is “Web” in the system configuration information 192 (FIG. 4). The capping amount to be released is a value obtained by multiplying the CPU allocation number 303 of the newly added virtual server by the 1-core frequency 304. For example, when the capping amount to be released is 12000 MHz, the Web server recorded in the system configuration information 192 (FIG. 4) is VM1, and the capped frequency 405 of VM1 is 12000 MHz, the following is performed. That is, a value (0) obtained by subtracting the capping amount (12000 MHz = 3000 MHz × 4) to be released from the capped frequency 405 (12000 MHz) of the VM1 is newly set to the VM1 as the capped frequency of the VM1, and the VM1 is capped. Record the value at frequency 405.

In step S1103, the system monitoring unit 181 changes the monitoring interval from the normal monitoring interval 907 to the high load monitoring interval 908, and monitors the business unit 13 at the set interval. And the same information as step S1001 (FIG. 9) is recorded on the service performance information 195 (FIG. 6). Similarly, the physical server 14 is monitored at the interval set as the high load monitoring interval 908, and the monitored information is recorded in the physical server performance information 194 (FIG. 5).

In step S1104, the management software control unit 184 determines whether a high load continues for a predetermined period. Here, the value of the load increase monitoring period 904 of the scaling rule information 197 (FIG. 8) is used as the predetermined period. Whether or not the high load continues is determined based on whether or not the trigger condition shown in step S1002 (FIG. 9) is satisfied.

If the management software control unit 184 determines that the high load continues (YES), in step S1105, the management software control unit 184 changes the setting of the LB (load distribution device) and starts request allocation to the newly added VM. However, if the VM addition started in step S1101 has not yet been completed, the management software control unit 184 changes the setting of the LB (load distribution device) after waiting for the completion of the addition.

When the management software control unit 184 determines that the high load is not continued (NO), in step S1109, the system configuration change unit 183 deletes the VM added in step S1101. However, if the VM addition started in step S1101 has not yet been completed, the system configuration changing unit 183 stops the addition.

A supplementary description will be given of the mechanism of the processing flow of steps S1101 to S1105 and S1109. As the scale-out process, the system configuration change unit 183 first starts the virtual server (VM) addition process (step S1101), and subsequently releases the CPU capping (step S1102). CPU capping is faster than adding and deleting VMs, and is suitable as a response to sudden load increase / decrease. Therefore, by canceling CPU capping in parallel with VM addition processing, The close state is reproduced in advance to cope with the load increase (fluctuation). The system monitoring unit 181 monitors this at a high load monitoring interval (step S1103). If the load increase (fluctuation) is temporary or if the influence of the load increase (fluctuation) can be absorbed by canceling CPU capping, the high load state does not continue (step S1104). The process is canceled without adding to (step S1109). However, if the high load state continues even after the CPU capping is released (step S1104), the management software control unit 184 responds by assigning a load to the added VM (step S1105). .

In step S1106, the system configuration changing unit 183 resets (returns) capping to the VM for which capping was canceled in step S1102. The target VM for capping is the same as the target VM in step S1102. Further, the amount of capping is the same as the amount of capping released in step S1102.

In step S1107, the system monitoring unit 181 returns the monitoring interval to the normal monitoring interval 907 and monitors the business unit 13 at the set interval. And the system monitoring part 181 records the same information as step S1001 (FIG. 9) in the service performance information 195 (FIG. 6).

In step S1108, the system configuration calculation unit 182 adds the VM information that has been added in step S1101 to the system configuration information 192 (FIG. 4). However, if step S1109 is executed, the information is not added and the process is terminated.

In the first embodiment, the computer resource to be capped has been described as a CPU. However, other computer resources (memory, disk, network, etc.) may be targeted.

As described above, according to the first embodiment, even when there is no load fluctuation prediction information, dynamic allocation of computer resources is effectively performed, and auto-scaling for maintaining the service level of business is realized. it can.

Next, a second embodiment of the present invention will be described with reference to FIGS.
In the first embodiment, it is determined whether or not the scale-out trigger condition is satisfied (step S1002 in FIG. 9). If the condition is satisfied (YES), VM addition processing is started (step S1101 in FIG. 10). . However, depending on the state of the application, the addition of the VM may not be effective. For example, in a system consisting of one interconnected Web server and one DB server, if the allowable number of simultaneous connections in the DB server is 100 and the number of simultaneous connections for the DB server in the Web server is 100, add one Web server Even so, the service performance is not improved because the total number of simultaneous connections for the DB server exceeds the allowable number of simultaneous connections in the DB server. In view of such a case, in the second embodiment, a resource addition method at the time of load increase is selected using information (connection information) indicating the state of the application.

In the second embodiment, the connection information shown in FIG. 11 is used in addition to the system configuration shown in FIG. In place of the flowchart of the first embodiment shown in FIGS. 9 and 10, the flowchart shown in FIG. 12 is used.

FIG. 11 is an explanatory diagram showing an example of connection information. The connection information 198 includes a role 1201 indicating the role of the VM and a connection number 1202 indicating the number of communication connections set in the VM to which the role is assigned. This connection information 198 is set by the administrator. The connection information 198 is stored in the management information group 19 (indicated by a dotted frame in FIG. 1).

FIG. 12 is a flowchart (parameter determination flowchart) showing the parameter determination process executed by the control unit 18.
Steps S1001 and S1002 are the same as those in the first embodiment. However, if it is determined in step S1002 that the scale-out trigger condition is satisfied (YES), the management software control unit 184 executes step S1303 instead of step S1003. In step S1303, the management software control unit 184 determines whether or not the number of connections in the Web server group falls within the number of connections in the DB server group even when one Web server that is a virtual server is added.

Specifically, the determination is made based on the following equation.
Number of connections set in Web server × (Number of operated Web servers + 1) <Number of connections set in DB server × Number of operated DB servers Here, the number of connections set in Web server The number of connections 1202 of the record whose role 1201 is “Web”. Similarly, the number of connections set in the DB server is the number of connections 1202 of the record whose role 1201 is “DB” in the connection information 198. The number of operated Web servers is the number of records whose role 402 is “Web” in the system configuration information 192 (FIG. 4). Similarly, the number of operated DB servers is the number of records whose role 402 is “DB” in the system configuration information 192 (FIG. 4).

If it is determined that the number of connections is not sufficient for the addition of the virtual server (Web server) (does not satisfy the above judgment formula) (NO), there is a possibility that the system performance will not be improved even if the Web server is added. Therefore, in that case, in step S1102, the system configuration change unit 183 executes the capping release processing of the existing virtual server, and thereafter ends the processing.

When it is determined that the number of connections is sufficient (addition of the above determination formula) to the addition of the virtual server (Web server) (YES), step S1003 shown in FIG. (Scale-out processing) is executed.

As described above, according to the second embodiment, dynamic allocation of computer resources is effectively performed according to the application state in the computer system, and auto-scaling for maintaining the service level of business is realized. it can.

11 management unit, 12 physical server, 13 business unit, 14 physical server, 15 hypervisor, 16 communication network, 18 management unit, 19 management information group, 121, 141 CPU, 122, 142 storage device, 123, 143 communication IF, 131 to 133 VM, 181 system monitoring unit, 182 system configuration calculation unit, 183 system configuration change unit, 184 management software control unit, 191 system construction request information, 192 system configuration information, 194 physical server performance information, 195 service performance information, 196 System performance model information, 197 Scaling rule information, 198 Connection information

Claims (12)

A physical server having computer resources including a processor, a storage device and a communication interface;
A virtualization device that virtualizes the computer resources and assigns them to a plurality of virtual environments;
A management device for managing the plurality of virtual environments;
A computer system comprising at least
The management device is
A monitoring unit that monitors loads of the plurality of virtual environments;
A control unit that determines whether or not a virtual environment needs to be added according to the monitored load;
Instructs the virtualization apparatus to add a virtual environment according to the determination of the control unit, releases the use restriction of the computer resources constituting the existing virtual environment, and releases the virtual environment after the addition of the virtual environment is completed. And a configuration change unit that resets the use restrictions to the existing virtual environment. The computer system according to claim 1,
The computer system according to claim 1, wherein the computer resource for which the use restriction is removed is a processor. The computer system according to claim 1 or 2,
The management device has a storage unit that holds a use restriction ratio of the computer resource,
The configuration change unit sets a use restriction amount of a computer resource of the virtual environment to be added based on the use restriction ratio. It is a computer system as described in any one of Claims 1-3, Comprising:
After the instruction to add the virtual environment,
The monitoring unit changes the monitoring time interval to a short time,
The configuration changing unit completes the addition of the virtual environment when the high load state is continued during the changed time interval, and cancels the addition when the high load state is not continued. A computer system characterized by deleting a virtual environment that has already been added. A computer system according to any one of claims 1 to 4, wherein
In addition to determining whether or not to add a virtual environment according to the monitored load, the control unit determines whether or not to add the virtual environment according to the number of application communication connections processed by the computer system. A computer system characterized by that. A computer system according to claim 5, wherein
When the control unit determines that the number of communication connections is insufficient for the addition of the virtual environment,
The computer system according to claim 1, wherein the configuration change unit only cancels the use restriction of the computer resources constituting the existing virtual environment and does not instruct the addition of the virtual environment. A physical server having computer resources including a processor, a storage device and a communication interface;
A virtualization device that virtualizes the computer resources and assigns them to a plurality of virtual environments;
An autoscaling method in a computer system comprising at least a management device for managing the plurality of virtual environments,
The management device
A first step of monitoring a load of the plurality of virtual environments;
A second step of determining whether or not a virtual environment needs to be added according to the monitored load;
A third step of instructing the virtualization unit to add a virtual environment according to the determination;
A fourth step of releasing the restriction on the use of computer resources constituting the existing virtual environment;
And a fifth step of resetting the released use restriction to the existing virtual environment after the addition of the virtual environment is completed. The auto-scaling method according to claim 7,
The computer scale resource in the fourth step is a processor. The autoscale method according to claim 7 or claim 8, wherein
The management device holds a use restriction ratio of the computer resource,
The third step further includes a step of setting a use restriction amount of the computer resource of the virtual environment to be added based on the use restriction ratio. The autoscale method according to any one of claims 7 to 9,
The fifth step includes a step of changing the monitoring interval to be monitored to a short time, and a step of completing the addition of the virtual environment when a high load state is continued during the changed monitoring interval; And a step of canceling the addition when the high load state is not continued and deleting the already added virtual environment. The autoscale method according to any one of claims 7 to 10, wherein
The management device executes a step of determining whether or not to add the virtual environment according to the number of communication connections of the application processed by the computer system between the second step and the third step. A featured autoscale method. The auto-scaling method according to claim 11,
When the number of communication connections is insufficient for the addition of the virtual environment, the management device instructs to add the virtual environment only by releasing the use restriction of the computer resources constituting the existing virtual environment. Auto-scaling method characterized by not.

Images