KR101366750B1 - Method for managing power of the server cluster and the server cluster using the same - Google Patents

Abstract

The present invention relates to a power management method of a server cluster and a server cluster using the method, the power management method of a server cluster according to an embodiment of the present invention, monitoring the load of a plurality of servers or switches, and Determining whether to shrink the movable server, and stopping the operation of the server according to the reduction decision. Thereby, power can be efficiently managed.

Server cluster, power, management

Description

Method for managing power of the server cluster and the server cluster using the same}

The present invention relates to a power management method of a server cluster and a server cluster using the method. More particularly, the present invention relates to a power management method of a server cluster capable of efficiently managing power and a server cluster using the method.

The present invention is derived from the research conducted as part of the IT growth engine technology development project of the Ministry of Knowledge Economy and the Ministry of Information and Communication Research and Development. [Task Management No. 2008-S-010-02, Title: Development of Multi-layer Optical Network Control Platform Technology] ].

In general, a web server or a firewall does not have sufficient processing performance compared to a communication bandwidth. Thus, a server cluster is formed using a plurality of server systems to perform a service by using a load balancing function of an L4 switch.

The load balancing capabilities of existing L4 switches are focused on solving the problem of load concentrating on one server by distributing the load evenly across all servers, and considering weights in case of different server performance. There is also. In a server cluster, the number of servers is usually calculated and designed based on peak load time, and the service is operated as the load is distributed to these servers.

The problem with such a server cluster is that all servers run at all times regardless of the load by running servers based on peak load. Typically, peak hours are only part of the day, and at other times all servers are forced to record low utilization rates. In general, however, server systems are characterized by a modest change in power consumption compared to load increases and decreases. Therefore, the power consumption of the server cluster is always consumed similarly based on the peak time regardless of the load, and also causes the same power consumption required for cooling and heating the installed IDC. As a result, the existing server cluster structure generally has a problem of using power inefficiently.

An object of the present invention is to provide a power management method of a server cluster capable of efficiently managing power, and a server cluster using the method, in order to solve the problems of the prior art described above.

According to an embodiment of the present invention for achieving the above object and other objects, the power management method of the server cluster, the step of monitoring the load of a plurality of servers or switches, and determining whether to reduce the operation server of the plurality of servers And stopping the operation of the server according to the reduction decision.

In addition, the apparatus for managing power in a server cluster according to an embodiment of the present invention for achieving the above-described object and the above-mentioned object and other objects includes a plurality of servers, a switch for distributing loads of the plurality of servers, and a plurality of servers. Or a control unit for monitoring the load of the switch, determining whether to reduce the operation of the plurality of servers, and controlling to stop the operation of the server according to the reduction decision.

According to the embodiment of the present invention, the power management method of the server cluster and the server cluster using the method can efficiently use power by reducing the movable server according to load or traffic. This can significantly reduce power usage by restricting unnecessary server operations.

In addition, power can be managed efficiently by expanding the movable server according to load or traffic. In particular, when the load server or the traffic increases, the server can be expanded immediately in a state where the movable server is reduced, so that the power usage can be efficiently managed.

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

1 is a diagram illustrating a server cluster according to an exemplary embodiment of the present invention.

Referring to the drawings, a server cluster according to an embodiment of the present invention includes a switch 110, a controller 130, and a plurality of servers 150.

The switch 110 may be, for example, an L4 switch. The L4 switch 110 distributes the load from the Internet network 105. For example, a server performing the same role among the plurality of servers 150 may be managed through a virtual IP (VIP), and the load or traffic directed to the server may be received by the L4 switch 110 having the VIP. Can be distributed (load balanced). As the load balancing method, various methods such as round robin, hashing, and list connection may be used.

In the meantime, in accordance with an embodiment of the present invention, the switch 110 may block or allocate a new request according to the control server 130 or the server enlargement decision to be described later.

The controller 130 monitors the traffic load of the L4 switch 110, controls the load balancing function, monitors the load of the plurality of servers 150, and controls the stopping and restarting of the server.

The controller 130 may be located independently of the L4 switch 110 or the plurality of servers 150 as shown in the drawings, but is not limited thereto and may be located in the L4 switch 110.

The controller 130 monitors the traffic load and the server load, determines whether the plurality of servers 150 are reduced or expanded, and stops or restarts the server according to a predetermined policy. At this time, in order to prevent or minimize service interruption, the load balancing function or the routing function of the L4 switch 110 is controlled to stop the server after receiving a new request to the server to be stopped. When the server is restarted, the load balancing function is controlled after checking the server operation.

As such, as a method of not receiving a new request in a situation where the existing service continues, a weight reduction round (WRR) method may be used to reduce the weight online during the load balancing function of the L4 switch 110. Can be.

If there is no weighted round robin (WRR) function, the function of the L4 switch 110 may be extended to not only enable / disable (or on-line / off-line) connection status of each server. The L4 switch 110 may be modified and supplemented to operate by adding a state that does not receive a new service request (session), such as available / unavailable / new stop (on-line / off-line / stop).

Meanwhile, in order to emphasize the gist of the present invention, only the configuration of the single L4 switch 110 is illustrated, but is not limited thereto. That is, as in the actual application, the same can be applied to the duplication of the L4 switch 110, the duplication of the port, and the like.

As described above, by controlling the on / off of the server operation in accordance with the load, only an appropriate number of servers can be operated, and power can be saved.

FIG. 2 is a simplified internal block diagram of the controller of FIG. 1.

Referring to the drawings, the controller 130 may include a switch interface 210, a server interface 215, a measurement / analysis unit 220, a switch control unit 225, a server control unit 230, an upper interworking interface 235, and the like. And a policy decision unit 240. In addition, the controller 130 may perform a function such as a database or a log.

The switch interface 210 and the server interface 215 make and communicate with the L4 switch 110 and the plurality of servers 150 by making a communication protocol or a specific command.

The measurement / analysis unit 220 collects the server 150 and the traffic load periodically, analyzes the result, and processes the result into a form necessary for policy decision and provides the result.

The switch control unit 225 controls the load balancing function and the routing function of the L4 switch 110 in accordance with the policy decision by the policy decision unit 240.

The server controller 230 performs on / off control for the server. According to the reduction or enlargement decision of the server determined by the policy determination unit 240, the server is controlled to stop or restart.

In this case, a method of stopping or restarting the server may be implemented in various forms. For example, it is possible to forcibly enter the power saving mode through the management function of the servers 150-1, 150-2, ... 150-n. In addition, a method of cutting off power supplied to the server through a remotely controllable power supply device may be used. In addition, when there is no request for a certain time, the server 150-1, 150-2, ... 150-n is set to automatically save / suspend, interrupt the communication to the server to induce server interruption and boot through the network (restart You can also use In the present invention, various types of server on / off methods may be used.

The higher interworking interface 235 performs interworking with a higher management system.

The policy determination unit 240 determines a server to stop or restart the operation among the plurality of servers 150 based on the measurement and analysis results of the measurement / analysis unit 220. At this time, it is possible to set various criteria.

For example, for a CPU utilization rate of the servers 150-1, 150-2, ... 150-n, a first threshold for reducing the server and a second threshold for expanding the server are set, and the server reduction is lower than the first threshold. If the determination is equal to or greater than the second threshold, the server enlargement decision may be made.

In addition, with respect to the number of service requests per server 150-1, 150-2,...

In addition, the server zoom decision may be made based on the estimated hourly load according to the usage history.

Moreover, the above-mentioned method can be mixed and used. In the present invention, the reduction / expansion determination methods of the movable server are not limited.

3 is a flowchart illustrating a power management method of a server cluster according to an exemplary embodiment of the present invention.

Referring to the drawings, the server and the traffic load is monitored (S310). Monitoring of such load may be performed by the control unit 130, in particular the measurement / analysis unit 220 in the control unit 130.

Next, it is determined whether to reduce the operation server (S315). The control unit 130, in particular, the policy determination unit 240 in the control unit 130 determines whether to reduce the operation server based on at least one of the CPU usage rate of the server, the number of service requests, and the estimated hourly load amount. That is, it is determined whether the preset threshold has been reached.

Next, an interruption server is selected (S320). The control unit 130, in particular, the policy determination unit 240 in the control unit 130 stops the operation among the plurality of servers 150 based on at least one of the CPU usage rate of the server, the number of service requests, and the estimated hourly load. Select a server to run. For example, the second server 150-2 of the plurality of servers 150 shown in FIG. 1 may be determined to stop operation.

Next, the server new request allocation is stopped (S325). The switch 110 is controlled to stop allocating a new request to the selected server. Such an operation may be performed by the controller 130, particularly the switch controller 225 in the controller 130.

Next, it is determined whether the service termination wait (S330), and all the services of the server is terminated (S335), and if so, the operation of the server is stopped (S340). Even if it is determined that the server is to be stopped, it is preferable not to stop the operation of the server immediately, but to wait until the service already in progress is terminated and to stop the operation of the selected server. The server operation may be stopped by the controller 130, in particular, by the server controller 230 in the controller 130.

On the other hand, after step 315 (S315), it is also possible to determine whether or not to expand the operation server (S350). The control unit 130, in particular, the policy determination unit 240 in the control unit 130 determines whether to expand the operation server based on at least one of the CPU utilization rate of the server, the number of service requests, and the estimated hourly load amount. That is, it is determined whether the preset threshold has been reached.

Next, a restart server is selected (S360). The control unit 130, in particular, the policy determination unit 240 in the control unit 130 restarts the operation among the plurality of servers 150 based on at least one of the CPU usage rate of the server, the number of service requests, and the estimated hourly load amount. Select a server to run. For example, the second server 150-2 that has already been suspended among the plurality of servers 150 shown in FIG. 1 may be determined to restart operation.

Next, request to start the server (S365). Such an operation may be performed by the controller 130, in particular, the server controller 230 within the controller 130.

Next, it is determined whether the service is on standby (S370) and whether the corresponding server is up (S375), and if so, the switch is controlled to start a new request allocation to the server to be restarted (S380).

On the other hand, whether to reduce or restart the above-mentioned operation server can be determined periodically, and accordingly, if none of the step 315 (S315) and 350 (S350), performing the check cycle waiting step (S355) It is possible to.

On the other hand, the drawing is shown to be terminated after the step 340 (S340) and step 380 (S380), but is not limited to this, it is possible to continue the above-described check cycle waiting step (S355). That is, the server and the traffic load is monitored (S310), and then steps S315 to S380 may be repeated periodically.

On the other hand, the configuration of the present invention is not limited to the application to the server cluster environment, but can be applied to the firewall, IDS / IPS situation that deploys the equipment of the same function through the L4 switch, it is built in any cluster environment for load balancing The present invention can also be embodied as computer readable codes on a computer readable recording medium. A computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like, and may also be implemented in the form of a carrier wave (for example, transmission over the Internet). Include. The computer readable recording medium may also be distributed over a networked computer system so that computer readable code can be stored and executed in a distributed manner.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is clearly understood that the same is by way of illustration and example only and is not to be construed as limiting the scope of the invention as defined by the appended claims. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention.

1 is a diagram illustrating a server cluster according to an exemplary embodiment of the present invention.

FIG. 2 is a simplified internal block diagram of the controller of FIG. 1.

3 is a flowchart illustrating a power management method of a server cluster according to an exemplary embodiment of the present invention.

Claims (14)

Monitoring the load of the plurality of servers or switches; Determining whether to reduce a number of movable servers of the plurality of servers; Stopping the operation of the server according to the reduction determination; Determining whether to expand the plurality of operation servers of the plurality of servers; And Starting the server according to the enlargement decision; The determining whether or not to shrink, Determining whether to reduce the movable server based on the utilization rate of the server and the number of service requests; The determining whether to enlarge, And determining whether to expand the movable server based on the utilization rate of the server, the number of service requests, and the estimated hourly load amount. delete The method of claim 1, And stopping allocating a new request to the corresponding server according to the shrinking decision. The method of claim 1, The stopping step, And forcibly transitioning the corresponding server to a power saving mode, cutting off the power of the corresponding server, or blocking communication to the corresponding server. delete delete The method of claim 1, And allocating a new request to the corresponding server according to the enlargement decision. A plurality of servers; A switch for distributing the load of the plurality of servers; And Monitoring the load of the plurality of servers or switches, determining whether to reduce operation of the plurality of servers, controlling to stop the operation of the server according to the reduction determination, And a controller configured to determine whether to expand the server cluster of the server cluster and to control the server to be operated according to the expansion decision. Wherein, Determining whether to reduce the movable server based on the CPU utilization rate and the number of service requests of the server; And determining whether to expand the movable server based on the CPU utilization rate of the server, the number of service requests, and the estimated hourly load amount. delete 9. The method of claim 8, The switch, And allocating a new request to the server according to the reduction decision. 9. The method of claim 8, Wherein, And controlling the server to forcibly transfer the server to a power saving mode, cut off power of the server, or block communication with the server when the server stops operating. delete delete 9. The method of claim 8, The switch, And assigning a new request to the server according to the expansion decision.

Images