KR20160094564A - Method and apparatus for power management of cluster system - Google Patents

Abstract

The present invention relates to a power management method and apparatus for a cluster system. The power management method for the cluster system according to an embodiment of the present invention comprises the steps of: determining a reference request load which is a reference to compare power management policies; determining a target power saving rate and service performance based on the reference request load; determining an autonomous management policy according to relative importance between a power saving rate increasing rule and a service performance improving rule; simulating changes in the power saving rate and the service performance according to changes of parameters of the power saving rate increasing rule and the service performance improving rule; determining the parameters of the power saving rate increasing rule and the service performance improving rule according to a result of the simulation; and executing a power policy to which the determined parameters are applied.

Description

TECHNICAL FIELD [0001] The present invention relates to a power management method and apparatus for a cluster system,

The present invention relates to a power management method and apparatus of a cluster system, and more particularly, to a power management method and apparatus of a cluster system capable of designing an optimal autonomous policy.

A server cluster is a system that connects servers providing a specific service with one or a plurality of load distributors and distributes service requests from the outside to servers in a server cluster. In general, the number of servers that make up a server cluster is designed to meet the maximum demands. Therefore, there is a problem that the server waits for unnecessary active state for a request smaller than the maximum request and wastes power. In order to solve this problem, it is a power management policy of the server cluster system to keep only the server optimized for the request size active and to switch the remaining power to the low power state.

Power management policies for server cluster systems include time-based power management policies, load-based power management policies, and autonomous power management policies. The time-based power management policy is a method of scheduling the power state of each server node on a time basis, and the load-based power management policy adjusts the power state of each server node according to load conditions such as CPU, It is a method to control the power state by designing a rule that matches the situation and the power state. An autonomic power management policy is a method of determining the number of servers to remain active in accordance with the size of requests for the entire server cluster and switching the remainder to an inactive state, that is, a low power state.

Generally, the time-based and load-based power management policies can be created based on the experience of the server cluster manager, which is the power control rule of the server cluster manager, and the power management result with the policy can be predicted and satisfied. In contrast, the autonomic management policy operates autonomously based on its own rules, and the design parameters of the autonomous management policy are not intuitively related to the results, making it unsatisfactory to design satisfactorily. Furthermore, power savings and service performance are two opposing goals, one that increases with the other, which makes it more difficult to achieve a satisfactory design. Therefore, a method of satisfying both power saving and service quality is required in the autonomous power management policy.

It is an object of the present invention to provide a power management method and apparatus that satisfy both power saving and service quality.

A power management method of a cluster system according to an embodiment of the present invention includes the steps of: determining a reference request load as a reference for comparing power management policies; determining a target power reduction rate and a service performance based on the reference request load; A step of determining an autonomic management policy according to the relative importance of the power saving rate enhancement rule and the service performance enhancement rule, and the simulation of the power saving rate and the service performance change according to the parameters of the power saving rate enhancement rule and the service performance enhancement rule The power saving rate increase rule according to the simulation result. Determining a parameter of the service performance enhancement rule, and executing a power policy applying the determined parameter.

According to the present invention, it is possible to provide a power management method and apparatus satisfying both power saving and service quality.

1 is a diagram showing a server cluster system.
2 is a block diagram illustrating an internal configuration of a power management apparatus of a cluster system according to an embodiment of the present invention.
3 is a flowchart showing a power management method of a cluster system according to an embodiment of the present invention
FIG. 4 is a diagram illustrating a simulation result of a power saving rate and a service performance change according to a power management method of a cluster system according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. Note that, in the drawings, the same components are denoted by the same reference symbols as possible. Further, the detailed description of well-known functions and constructions that may obscure the gist of the present invention will be omitted.

Throughout the specification, when a part is referred to as being "connected" to another part, it includes not only "directly connected" but also "indirectly connected" between the devices in the middle. Throughout the specification, when an element is referred to as "comprising ", it means that it can include other elements as well, without excluding other elements unless specifically stated otherwise.

1 is a diagram showing a server cluster system.

1, a server cluster system 100 includes service clients 111, 113, ..., 11j, a network 120, a load distributor 130, active server nodes 141, 142, ..., A server node pool 151, 152, ..., 15m, and a power management server 160. [

The service clients 111, 113, ..., and 11j request services to a service server that provides necessary services through the network 120. [ At this time, the service clients 111, 113, ..., 11j request services to the service servers, but actually request services to the virtual servers of the load distributors.

The network 120 connects the service clients 111, 113, ..., 11j and the load distributor 130. [ Also, the network 120 may include all networks such as the Internet, an intranet, and the like.

The load distributor 130 receives a service request from the service clients 111, 113, ..., 11j via the network 120 and selects one of the server nodes included in the server node pool 151, 152, ..., And distributes the service request to the server nodes 141, 142, ..., and 14k. At this time, the load distributor 130 may receive the service request through the virtual server and distribute the request load to the actual server of the individual server node according to the load distribution algorithm. At least one load distributor 130 may be present.

Servers included in the server cluster system 100 constitute server node pools 151, 152, ..., and 15m and servers among the servers providing the actual services correspond to active server nodes 141, 142, ..., do. The active server nodes 141, 142, ..., and 14k distribute service requests from the load distributor 130 and provide services to the service clients 111, 113, ..., 11j.

The power management server 160 manages the power of the entire cluster system by controlling the power state of the server nodes. The power management server 160 can manage the power based on the autonomous power management policy and the power consumption model and the performance model of the server node.

2 is a block diagram illustrating an internal configuration of a power management apparatus of a cluster system according to an embodiment of the present invention.

2, the power management apparatus 200 includes a user interface 210, a communication unit 220, a storage unit 230, a simulation unit 240, a power policy design unit 250, a power policy execution unit 260, And a control unit 270.

The user interface 210 may provide an interface for a user to design a power policy. According to one embodiment of the present invention, a user may determine power policies and modify parameter values through the user interface 210. [ You can also select the baseline request load and see the results of the simulation.

The communication unit 220 connects the power management apparatus 200 and the server cluster system, and performs wired and / or wireless communication. According to an embodiment of the present invention, the communication unit 220 can send and receive messages to control power states with the active server nodes 141, 142, ..., and 14k and the server node pools 151, 152, ..., have.

The storage unit 230 stores programs and data necessary for the operation of the power management apparatus 200. [ The storage unit 230 may be a volatile storage medium or a nonvolatile storage medium and may be a combination of both storage media. The volatile storage medium may include a semiconductor memory such as a RAM, a DRAM, and a SRAM. The nonvolatile storage medium may include a hard disk and a flash NAND memory. According to one embodiment of the present invention, the storage unit 450 stores a reference request load 231, a model 232, and a power policy 235. The reference request load 321 is a request load that is a base for applying a policy, and refers to a CPU-oriented request load corresponding to a service provided by the cluster, a DISK-based request load, and the like. The model 232 includes a power consumption model 233 and a performance model 234, which can be used for the simulation process. The power policy 235 may include a plurality of designed power policies.

The simulation unit 240 can simulate the power saving rate and the service performance change according to the parameters of the power saving rate improvement rule and the service performance improvement rule by reflecting the reference request load, the power consumption model, the performance model, and the power policy.

The power policy design section 250 allows the user interface 210 to modify the parameters of the autonomic control policy with a module for designing the power policy. According to an embodiment of the present invention, the power policy designing unit 250 determines an autonomous management policy according to the relative importance of the power saving rate enhancement rule and the service performance enhancement rule, and increases the power saving rate according to the simulation result of the simulation unit 240 You can determine the parameters of rules and service performance improvement rules.

The power policy execution unit 260 executes the determined power policy to control the power state of the server nodes of the cluster. That is, the power policy execution unit 260 can execute the power policy applying the parameters determined by the power policy design unit 250. [

The control unit 270 is a component that controls the overall operation of the power management device 200. [ In the present invention, the controller 270 controls the overall operation of the power management apparatus 200 in designing and simulating the power policy and executing the power policy. In the above description, the simulation unit 240, the power policy design unit 250, the power policy execution unit 260, and the control unit 270 are configured as separate blocks, and each block performs a different function. However, For convenience, these functions are not necessarily distinguished. For example, the functions performed by the simulation unit 240, the power policy design unit 250, and the power policy execution unit 260 may be performed by the control unit 460 itself.

3 is a flowchart showing a power management method of a cluster system according to an embodiment of the present invention

First, in step 310, a reference request load is determined. The baseline request load is a reference load that is the basis for applying the policy, and is the basis for comparing power management policies. The base request load includes the CPU-oriented request load corresponding to the service provided by the cluster, and the DISK-oriented request load. According to one embodiment of the present invention, a user may enter a reference request load via the user interface 210. [

Then, in step 320, the target power saving rate and the service performance are determined based on the reference request load. And to determine the power savings and service performance to achieve through policy design. According to one embodiment of the present invention, a user may input a target power saving rate and service performance through the user interface 210. [

Further, in step 330, the autonomous management policy is determined according to the relative importance of the power saving rate improvement rule and the service performance improvement rule. In this case, the autonomic management policy can be determined by an optimization method that derives the satisfaction ratio (Pareto Optimal) of the weighted sum and min-max methods for the rules for improving the power saving rate and improving the service performance. At this time, according to the relative importance of the two goals, the autonomous management policy can be determined by selecting the most satisfactory satisfaction among the plurality of satisfaction waters.

At this time, the weighted sum method can be calculated by the following equation.

                                                   Equation (1)

Further, the min-max method can be calculated by the following equation.

                                              Equation (2)

Where N is the number of active server nodes, Npower is the number of active server nodes meeting the power consumption rate target, Nper is the number of active server nodes meeting the response time time objective, and w1 and w2 are the weight values .

This weighted sum and min-max method is an embodiment of the optimizing method for satisfying (Pareto Optimal), and it is also possible to use an optimizing method that induces satisfactory solutions in other methods.

Next, in step 340, the power saving rate and the service performance change are simulated according to the parameters of the power saving rate improvement rule and the service performance improvement rule. According to an embodiment of the present invention, a power policy is simulated using a reference request load and a power consumption model and a performance model of service nodes. It can calculate power savings and service performance of power management policies and show them to users. Simulation results can show the change of power saving rate and service performance according to changes of parameters of power saving rate improvement rule and service performance improvement rule.

Thereafter, in step 350, parameters of the power saving rate improvement rule and the service performance improvement rule are determined according to the simulation result. That is, the optimal power policy can be determined according to the simulation result. FIG. 4 shows an example of such a simulation result and a parameter determination process.

FIG. 4 is a diagram illustrating a simulation result of a power saving ratio and a service performance change according to a power management method of a cluster system according to an embodiment of the present invention.

Referring to FIG. 4, the X-axis of the graph represents the failure rate of the response time, which means the degree of deviation from the target response time. The Y-axis of the graph represents power consumption. The smaller the two goals, the better the goal. In the graph, A is a parameter of the service performance improvement rule, and the simulation result is satisfied by changing parameters of the power saving rate improvement rule when each value is fixed. Based on these results, if the satisfaction of the user is satisfied, the parameter can be used to confirm the autonomous power management policy. If you are not satisfied, you can repeat the design process. If there is no autonomous policy satisfied by the user, it is possible to select the autonomous policy that satisfies the most. In this case, the user may select it or use the distance from the user's target.

Returning back to FIG. 3, the power policy applying the parameters determined in step 360 is executed.

Through such power policy design process, it is possible to design a power policy that satisfies both power saving and service quality.

The embodiments of the invention thus far described may be embodied in instructions carried out by a processor and stored in a computer-readable storage medium. When these instructions are executed by a processor, they may generate means for implementing the functions / operations specified in the flowcharts and / or block diagrams described above. Each block in the flowcharts / block diagrams may represent hardware and / or software modules or logic that implement embodiments of the present invention. Further, the functions referred to in the block diagrams may occur outside the order mentioned in the drawings, or may occur simultaneously.

The computer-readable medium can include, for example, but is not limited to, a nonvolatile memory such as a floppy disk, ROM, flash memory, disk drive memory, CD-ROM, and other persistent storage, Available.

The embodiments of the present invention disclosed in the present specification and drawings are merely illustrative examples of the present invention and are not intended to limit the scope of the present invention in order to facilitate understanding of the present invention. It will be apparent to those skilled in the art that other modifications based on the technical idea of the present invention are possible in addition to the embodiments disclosed herein.

210: User interface
220:
230:
240: Simulation section
250: Power Policy Design Department
260: Power Policy Implementation Department
270:

Claims (1)

Determining a reference request load that is a basis for comparing power management policies;
Determining a target power saving rate and a service performance based on the reference request load;
Determining an autonomic management policy according to the relative importance of the power saving rate improvement rule and the service performance improvement rule;
Simulating a power saving rate and a service performance change according to the power saving rate improvement rule and the parameter change of the service performance improvement rule;
Determining parameters of the power saving rate enhancement rule and the service performance enhancement rule according to the simulation result; And
And executing a power policy applying the determined parameters. ≪ Desc / Clms Page number 22 >

Images