JP6186287B2 - System management server and control method - Google Patents

Description

  The present invention relates to an apparatus and a method for monitoring and controlling a system constructed across data centers.

  Scale-out type server systems are popular, especially in large-scale Web systems. In a scale-out type server system, the load on the system is monitored, and when the load on the system increases, a server can be added according to a configuration change instruction and the load can be distributed to eliminate the bottleneck.

  However, in such a scale-out type server system, even if a server whose load is directly increased is scaled out, the second part becomes a bottleneck, and the performance of the server system is not improved. was there.

  Therefore, Patent Document 1 discloses a technique for predicting a secondary bottleneck based on the monitoring result of the server load.

JP2010-237901

  However, in Patent Document 1, the location where a secondary bottleneck is predicted by the monitoring control server is limited to the server. In a server system constructed by using components such as resources and software provided by a plurality of data centers, the network between the data centers may become a secondary bottleneck. Not. In addition, there is no mention of means for solving the predicted secondary bottleneck.

  An object of the present invention is to provide means for predicting and eliminating a network bottleneck associated with server scale-out in a server system constructed across data centers.

  In order to achieve the above object, in the present invention, a server system management server constructed across data centers monitors the amount of network usage between data centers used by the server system, and scales the server based on the monitoring results. Calculate the predicted value of network usage after execution of out. Then, the network setting between the data centers is changed according to the predicted value. When there is sufficient free bandwidth in the network used by the server system that has executed the scale-out, the bandwidth allocated for the communication of the server system is increased. If the available bandwidth is insufficient, the path is switched so that communication of the server system is executed on another network.

  In addition, the management server calculates a predicted value of the network usage after the server scale-in is executed, and changes the network settings between the data centers according to the predicted value. Reduce the network bandwidth allocated to the communication of the server system that executed the scale-in, or switch the network to be used.

  According to the present invention, it is possible to eliminate a network bottleneck associated with server scale-out.

It is a component diagram showing the schematic structure of a system. It is a figure of the computer with which integrated operation management operates. It is a component figure of an integrated operation management part. It is an example of a configuration management table. It is an example of a performance management table. It is an example of the connection relation management table between data centers. It is an example of a structure change means management table. It is a flowchart which shows the structure change procedure at the time of scale-out. It is a flowchart which shows the structure change procedure at the time of scale-in.

  Embodiments of a system configuration change method according to the present invention will be described below with reference to the drawings.

  FIG. 1 is a schematic configuration diagram showing an embodiment of a system in the present embodiment. In this embodiment, a system is constructed using resources and applications of the data center (DC) A 102 and the data center (DC) B 103 that provide cloud services.

  The data center A 102 and the data center B 103 have an operation management unit (122, 124) that monitors and controls the hardware and software in the data center, and these are integrated operation management of the integrated management server 125 via the network 123. Controlled by the unit 101. FIG. 1 shows a configuration in which the integrated operation management unit 101 is deployed in the data center A 102 and the integrated management server 125 outside the data center B 103. However, the integrated operation management unit 101 may be deployed in the data center A 101 or data It may be deployed in the center B 102.

  The data center A 102 and the data center B 103 are connected by two physical networks (117, 118). These networks are physical lines that connect the network device 116 and the network device 121. The data center B 121 has bandwidth control devices 119 and 120 for each network in order to control the bandwidths of the networks 117 and 112. These bandwidth control devices 119 and 120 are controlled by the operation management unit 124, and the usage amount of the network is also monitored by the operation management unit 124. The virtual router deployed for each system determines which of the network 117 and the network 118 is used for communication.

  Physical server devices 113 and 115 are connected to the network devices 116 and 121, respectively. The virtualization apparatuses 112 and 114 provide a virtual environment using the resources of the server apparatuses 113 and 115. The system 104 is an example of a system constructed on these virtual environments. In this system, a load balancer 107 and Web servers 106a to 106c are deployed as components on the virtual environment of the data center B 103. The Web servers 106a to 106c are connected to a database server 105 that is arranged as a component on the virtual environment of the data center A 102. In addition, the data center B 103 is provided with a virtual router 111 that controls a network path in the data center B 103 of the system 104.

  Similarly, on the virtual environment in the data center A 102 of the system 104, a virtual router 109 that controls the network path in the data center A 102 is provided as a component. In the router 121 and the router 116, the network is set in advance so that the route to the network A 117 or the network B 118 can be switched by the route information change setting in the virtual router 111 and the virtual router 109. ing.

  FIG. 2 is a diagram showing a configuration of a computer system in which the integrated management server 125 and the operation management unit 122 124 operate. In this computer system, a CPU 201, a memory 202, a communication interface 204, and an input / output device 205 are connected through a bus 209. A program that realizes the integrated operation management unit 101 and the operation management units 122 and 124 is stored in the external storage device 207, read into the memory 202 via the bus 209 by the input / output device 205, and executed by the CPU 201. Communication between the integrated operation management unit 101 and the operation management unit 122 is realized by the network 123 using the communication interface 204.

  FIG. 3 is a component diagram of the integrated operation management unit 101 and the operation management unit 103 controlled by the integrated operation management unit. The integrated operation management unit 101 manages the configuration of the system to be monitored using the configuration management table 304. The monitoring unit 301 monitors the management target system based on the information in the configuration management table 304, and records the monitored information in the performance management table 305.

  The bottleneck determination unit 302 is configured to manage the performance data recorded in the performance management table 305, the inter-data center connection relation management table 306 in which the network information for connecting the data centers is recorded, and the selectable configuration change means. Based on the management table 307, a required configuration change method is selected. The configuration change unit 303 transmits the configuration change selected by the bottleneck determination unit 302 to each data center, and implements the configuration change.

  The monitoring unit 301 and the configuration changing unit 303 in the integrated operation management unit 101 monitor and control each data center, and the monitoring unit 310 and the configuration of the operation management unit 103 arranged in each data center. This is realized via the changing unit 311. In the operation management unit 103, the configuration change according to the server load is realized by the server configuration change unit 312. The server configuration changing unit 312 increases or decreases the number of registered servers based on a preset threshold within a preset range according to the load on the server. A technique related to such server scale-out and scale-in is disclosed in, for example, Japanese Patent Application Laid-Open No. 2010-237901.

  FIG. 4 is an example of the configuration management table 304. This table includes a “system name” column 401 indicating a system name for identifying a system, a “DC A” column 403 and a “DC B” column 404 indicating data center A and data center B components constituting the system, The “NW” column 405 indicating the identifier of the network to which the network is connected and the “allocated bandwidth” column 406 indicating the bandwidth allocated to the communication of this system.

  FIG. 5 is an example of the performance management table 305. This table includes a “system name” column 502 indicating a system name for identifying a system, an “NW” column 503 indicating an identifier of a network used in the system, and an “average” indicating an average throughput obtained by monitoring the network. A “value” column 504, a “maximum value” column 505 indicating the maximum value of the throughput, and the like.

  FIG. 6 is an example of the inter-data center connection relationship management table 306. This table includes a “NW” column 602 indicating a network identifier, a “data center” column 603 indicating a data center to which the network is connected, a “maximum bandwidth” column 604 indicating the maximum bandwidth of the network, an unallocated network An “unallocated standby” column 605 indicating a bandwidth, a “delay” column 606 indicating a normal delay time, and a “cost” column 607 indicating a network usage cost are configured.

  FIG. 7 is an example of the configuration change unit management table 307. This table includes a “configuration change” column 703 indicating possible configuration change means for the bottleneck occurrence location indicated in the “bottleneck location” column 702.

  FIG. 8 is a flowchart showing a processing procedure when the scale-out of the Web server 106 is performed. This flowchart is notified to the integrated operation management unit 101 through the communication unit 309 and executed by the bottleneck determination unit 302 when server scale-out is realized by the server configuration change unit 312.

  In step 801, the bottleneck determination unit 302 estimates the network usage after the Web server 106 is scaled out. This estimate increases, for example, by acquiring the average usage amount (column 504) of the currently used network from the performance management table 305 and referring to the components of the data center A and data center B in the configuration management table 304. It is obtained by increasing the average usage by the proportion of the number of Web servers 106.

Next, in step 802, it is determined whether the network usage obtained in step 801 falls within the assumed usage of the currently used network. This is obtained by comparing the estimated value of the network usage obtained in step 801 with the value in the “allocated bandwidth” column 406 of the configuration management table 304. If the estimated value of the network usage is equal to or less than the value in the “allocated bandwidth” column 406, the bandwidth is sufficient in the current network, and thus this flow ends.
If the estimated value of the network usage is larger than the value in the “allocated bandwidth” column 406, the bandwidth of the network will be insufficient. Therefore, the process proceeds to step 803 to determine whether the bandwidth can be increased in the current network. This is because the value in the “allocated bandwidth” column 406 of the configuration management table 304 representing the current bandwidth for the network being used and the connection relation management table 306 between data centers indicating the maximum expandable range for the network being utilized. This can be determined by comparing the value in the “unallocated bandwidth” column 605.

  If the value in the “unallocated bandwidth” column 605 is equal to or greater than the value in the “allocated bandwidth” column 406, the bandwidth can be increased, and the process proceeds to step 805. In step 805, the configuration changing unit 303 increases the bandwidth of the network allocated to the communication of the system 104 by a predetermined amount according to the estimated value obtained in step 801. The bottleneck determination unit updates the “allocated bandwidth” column 406 of the configuration management table 304 with the increased bandwidth value.

  If the value in the “unallocated bandwidth” column 605 is smaller than the value in the “allocated bandwidth” column 406 in step 803, the bandwidth cannot be increased, and the process proceeds to step 804. In step 804, it is determined whether the bandwidth can be increased by switching the network to be used to another network. In this determination, a network that can be used by system A is searched from the inter-data center connection relationship table 306, and the value in the “unallocated bandwidth” column 605 that represents the unallocated bandwidth of the network is the estimated value of network usage. It is possible to switch when there are more. If it is determined that switching is possible, the process proceeds to step 806, and the configuration change unit 303 switches the route to be used by changing the route settings of the virtual routers 109 and 110 deployed in the respective data centers. To implement.

  In step 804, it is determined whether or not switching is possible only with the network bandwidth. Here, in the case where a delay time required from the user or a cost is incurred in using the network, the maximum allowable delay is determined. It is also possible to select a network that receives values such as time and cost and satisfies these multiple parameters.

  With the above configuration and processing, it is possible to eliminate a network bottleneck associated with server scale-out.

  FIG. 9 is a flowchart showing a processing procedure when the server is scaled in. This flowchart is notified to the integrated operation management unit 101 through the communication unit 309 when the server configuration change unit 312 executes a scale-in that reduces the number of Web servers 106 that are operating, and the bottleneck determination unit 302 Executed.

  In step 901, the bottleneck determination unit 302 estimates the amount of network usage after the scale-in of the Web server 106. This estimate is obtained by reducing the average usage amount by the proportion of the number of Web servers 106 to be reduced, similarly to Step 801 in the flowchart of FIG.

  In step 902, the bottleneck determination unit 302 determines whether there is a network set as a priority network among a plurality of networks connecting the data center A and the data center B. When the scale-in is executed, the amount of network usage decreases, so that communication can be performed by concentrating on a part of a plurality of networks connecting the data centers. In such a case, the priority network is set as the aggregation destination network. It is assumed that which network is the priority network is set in advance.

  In step 902, when there is a network set as a priority network, the bottleneck determination unit 302 determines whether or not the free bandwidth of the priority network is sufficient. Specifically, referring to the “allocated bandwidth” column 406 of the configuration management table 304, the total value of the bandwidths allocated to networks not set as the priority network is calculated. Further, referring to the “unallocated bandwidth” column 605 of the inter-data center connection relation table 306, the value of the unallocated bandwidth of the priority network is acquired.

  If the value of the unassigned bandwidth is equal to or greater than the total value assigned to networks other than the priority network, it is determined that the free bandwidth of the priority network is sufficient, and the process proceeds to step 905. In step 905, the configuration changing unit 303 switches the route to be used by changing the route setting of the virtual routers 109 and 110 provided in the respective data centers. Here, the path is switched so that the communication of the system that is being executed in a network other than the priority network is executed in the priority network.

  If the value of the unallocated bandwidth is smaller than the total value allocated to networks other than the priority network, it is determined that the free bandwidth of the priority network is not sufficient, and the process proceeds to step 904. In step 904, the bottleneck determination unit 302 determines whether the network bandwidth used by the reduced system 104 can be reduced. Here, it is assumed that the minimum value of the bandwidth allocated to the system 104 is set in advance.

  If the estimated value of the network usage obtained in step 901 is equal to or greater than the minimum value, the bottleneck determination unit 302 determines that the network bandwidth used by the system 104 can be reduced, and proceeds to step 905. In step 905, the configuration changing unit 303 reduces the network bandwidth allocated to the communication of the system 104 by a predetermined amount according to the estimated value. Then, the bottleneck determination unit 302 updates the “assigned bandwidth” column 406 of the configuration management table 304 with the decreased bandwidth value.

  If the estimated network usage value obtained in step 901 is smaller than the minimum value, the bottleneck determination unit 302 determines that the network bandwidth used by the system 104 cannot be reduced, and ends the process. To do.

  With the above configuration and processing, it is possible to prevent excessive use of network resources after the scale-in of the server.

Claims (8)

A management server that monitors and controls a system constructed using a plurality of virtual servers provided by the first and second data centers;
A predetermined bandwidth of the network between the first and second data centers is assigned to communication associated with the processing of the system,
The management server
Monitor the usage when using the network in communication of the system,
When a scale-out of the virtual server is detected, a predicted value of the usage to be used for communication of the system after the scale-out is calculated based on the number of virtual servers increased by the scale-out and the usage used for monitoring. And
When the predicted value exceeds the allocated bandwidth, the management server instructs the first or second data center to increase the allocated bandwidth. The management server according to claim 1,
The management server
Based on the predicted value and the unassigned bandwidth of the network, it is determined whether there is an unassigned bandwidth sufficient to increase the assigned bandwidth, and when it is determined that there is no unassigned bandwidth, A management server that instructs the first or second data center to execute communication associated with system processing in a network different from the network. A management server according to claim 2,
The management server
When a scale-in of the virtual server is detected, a predicted value of the amount used for communication of the system after the scale-in is calculated based on the number of virtual servers decreased due to the scale-in and the monitored usage. And
Instructing the first or second data center to switch the communication path so as to reduce the bandwidth allocated to the communication associated with the processing of the system based on the predicted value after the scale-in. Management server characterized by A management server according to claim 3,
The management server
Based on the predicted value after scale-in and the unallocated bandwidth of the predetermined network different from the network, when it is determined that the communication being performed in the network can be performed in the predetermined network,
A management server that instructs the first or second data center to switch a communication path so that communication executed in the network is executed in the predetermined network. A control method for a system constructed using a plurality of virtual servers provided by first and second data centers,
A predetermined bandwidth of the network between the first and second data centers is assigned to communication associated with the processing of the system,
Monitor the usage when using the network in communication of the system,
When a scale-out of the virtual server is detected, a predicted value of the usage to be used for communication of the system after the scale-out is calculated based on the number of virtual servers increased by the scale-out and the usage used for monitoring. And
When the predicted value exceeds the allocated bandwidth, the system control method is configured to instruct the first or second data center to increase the allocated bandwidth. A system control method according to claim 5, comprising:
Based on the predicted value and the unassigned bandwidth of the network, it is determined whether there is an unassigned bandwidth sufficient to increase the assigned bandwidth, and when it is determined that there is no unassigned bandwidth, A method for controlling a system, characterized by instructing the first or second data center to execute communication associated with system processing in a network different from the network. A system control method according to claim 6, comprising:
When a scale-in of the virtual server is detected, a predicted value of the amount used for communication of the system after the scale-in is calculated based on the number of virtual servers decreased due to the scale-in and the monitored usage. And
Instructing the first or second data center to switch the communication path so as to reduce the bandwidth allocated to the communication associated with the processing of the system based on the predicted value after the scale-in. A system control method characterized by the above. A system control method according to claim 7, comprising:
Based on the predicted value after scale-in and the unallocated bandwidth of the predetermined network different from the network, when it is determined that the communication being performed in the network can be performed in the predetermined network,
A method for controlling a system, comprising: instructing the first or second data center to switch a communication path so that communication performed on the network is performed on the predetermined network.

Images