JP4868262B2 - Distributed processing system, job distribution method and program - Google Patents

Description

  The present invention relates to a distributed processing system, a job distribution method and program thereof, and more particularly to a distributed processing system suitable for power saving, and a job distribution method and program thereof.

  In a conventional distributed processing system, the processing capability of each information processing apparatus that distributes jobs according to the operation status of the information processing apparatus that is the job distribution destination for jobs (JOBs) that are units of work to be performed by the information processing apparatuses Management is performed in consideration of the processing performance of the entire system, depending on the amount of space and availability.

  As a technique related to this, for example, Patent Document 1 describes a method of distributing batch jobs in a multi-server environment. This method predicts the time it takes to execute the batch job from the batch job characteristics and the number of input data, predicts the load status of each execution server over the entire time range, and executes the batch job based on it Select the server. Further, every time a batch job is executed, the load generated by the execution of the batch job is measured, and the batch job characteristics are updated based on the actually measured value. According to this, it is said that the accuracy of batch job characteristics and the accuracy of selection of execution servers can be improved.

  However, such a distributed processing method does not take into account the characteristics of each power supply that supplies power to each information processing apparatus, so that the system is operating without optimizing the power consumption of the entire system. was there.

  In connection with this, a distributed processing method considering power conversion efficiency is also known. For example, in Patent Document 2, a method of distributing processing to a plurality of processing units in a plurality of processing units whose power consumption changes according to the amount of allocated processing, each processing unit having a power supply module. Is described. This method measures the input current of the power supply module, measures the change in the input current of the power supply module of the one processing unit when the amount of processing assigned to one of the plurality of processing units is changed, and Based on the measured change in input current, a processing unit to which a newly generated process is assigned is determined. According to this, it is said that the processing can be distributed in consideration of the voltage conversion efficiency of the power source.

JP 2007-249491 A JP 2009-060171 A

  As described above, in the conventional distributed processing technique, the entire system is determined based on the processing capability and availability of each information processing apparatus that distributes jobs according to the operation status of the information processing apparatus that is the job distribution destination. Distributed processing has been performed to improve performance. Such a method does not consider the characteristics of each power source that supplies power to each information processing apparatus, and has a problem that the system is operated without optimizing the power consumption of the entire system.

  In addition, as a distributed processing method considering power conversion efficiency, a method of transferring to another distribution destination when the power conversion efficiency of the distribution destination deteriorates as in Patent Document 2 is also known. When considering the various operating states of the entire system, there is a problem in that the power conversion efficiency of the entire system is not necessarily maximized and the power consumption is not minimized at all timings for distributing jobs.

  The object of the present invention is to solve the above-described problems and to always minimize the power consumption of the entire system at all timings for distributing jobs, regardless of the power supply type / configuration, and always consumes the system. It is an object to provide a distributed processing system capable of minimizing power, a job distribution method and a program thereof.

  In order to achieve the above object, a distributed processing system according to the present invention includes a plurality of power supply units and a plurality of information processing units connected to any one of the power supply units and supplied with power from each power supply unit, A system common processing unit commonly connected between the client and each of the information processing units, the system common processing unit including a storage unit that stores power-conversion efficiency characteristics of each of the power supply units; A power consumption monitoring unit that monitors power consumption of each information processing unit, and a power consumption of each information processing unit monitored by the power consumption monitoring unit when a job from the client is distributed to each information processing unit; Job distribution for determining an information processing unit as a job distribution destination so that the power conversion efficiency of the entire system is maximized from the information on the power-conversion efficiency characteristics of each power unit stored in the storage unit The previous determination unit, and having a job control unit of the determined job distribution destination to the information processing unit dispatches jobs from the client.

  A job distribution method of a distributed processing system according to the present invention includes a plurality of power supply units, a plurality of information processing units connected to any of the power supply units, and supplied with power from each power supply unit, a client, A job distribution method for a distributed processing system having a system common processing unit commonly connected to each information processing unit, wherein the system common processing unit stores the power-conversion efficiency characteristics of each power supply unit The power consumption of each information processing unit monitored by the power consumption monitoring unit when the power consumption of each information processing unit is monitored and a job from the client is distributed to each information processing unit. From the information on the power-conversion efficiency characteristics of each power supply unit stored in the unit, the information processing unit as a job distribution destination is determined so that the power conversion efficiency of the entire system is maximized. Of the determined job distribution destination to the information processing unit, characterized in that it dispatches jobs from the client.

  A job distribution program for a distributed processing system according to the present invention includes a plurality of power supply units, a plurality of information processing units connected to any of the power supply units, and supplied with power from each power supply unit, a client, A job distribution program for a distributed processing system having a system common processing unit commonly connected to each of the information processing units, the computer constituting the system common processing unit being a power of each power supply unit A storage unit that stores conversion efficiency characteristics, a power consumption monitoring unit that monitors power consumption of each information processing unit, and a power consumption monitoring unit that distributes jobs from the clients to the information processing units. From the information on the power consumption of each information processing unit to be monitored and the power-conversion efficiency characteristics of each power source unit stored in the storage unit, Functions as a job distribution destination determination unit that determines an information processing unit as a job distribution destination so that the source conversion efficiency is maximized, and a job control unit that distributes jobs from the client to the information processing unit as the determined job distribution destination Is for.

  According to the present invention, it is possible to minimize the power consumption of the entire system at all times by maximizing the power conversion efficiency of the entire system at all timings for distributing jobs, and the power-conversion of each power supply. By having the efficiency characteristic information, the system power consumption can always be minimized regardless of the power source type / configuration.

It is a block diagram which shows the whole structure of the distributed processing system which concerns on embodiment of this invention. It is a sequence diagram explaining operation | movement of the distributed processing system shown in FIG. It is a graph explaining the power-conversion efficiency characteristic of each power supply part memorize | stored in the memory | storage part shown in FIG.

  Next, embodiments of a distributed processing system, method, and program according to the present invention will be described in detail with reference to the drawings.

  FIG. 1 explains the overall basic configuration of the distributed processing system according to the present embodiment. The distributed processing system shown in FIG. 1 includes a plurality of power supply units (power supply devices) 1 to n, a plurality of information processing units (information processing devices) 11 to nm connected to any one of the power supply units 1 to n, The system common processing unit 100 is a common control unit commonly connected to the plurality of information processing units 11 to nm, and a plurality of clients 1000-1 to 1000-k are connected to the system common processing unit 100. . In this distributed processing system, job processing requests are distributed according to requests from the clients 1000-1 to 1000-k.

  The information processing units 11 to nm are supplied with power from the respective power supply units 1 to n. In the example of FIG. 1, the plurality of information processing units 11 to nm include information processing units 11 and 12 connected under the power supply unit 1 and information processing units 21, 22 and 23 connected under the power supply unit 2. In the same manner, the information processing units n1,..., Nm connected under the power source unit n are configured. The number of information processing units connected to each of the power supply units 1 to n is not limited to the example in FIG.

  The system common processing unit 100 includes a job (JOB) control unit 101, a power calculation / comparison unit (job distribution destination determination unit) 102, and a power consumption monitoring unit 103.

  Upon receiving a job processing request from the clients 1000-1 to 1000-k, the job control unit 101 requests the power calculation / comparison unit 102 to specify an information processing unit as a job distribution destination. Also, by receiving job distribution destination information from the power calculation / comparison unit 102, the job is distributed to the designated information processing units 11 to nm.

  The power calculation / comparison unit 102 has a storage unit 104 that stores the power supply conversion efficiencies of a plurality of power supply units 1 to n having different power-conversion efficiency characteristics. When the power calculation / comparison unit 102 is connected to the job control unit 101 and the power consumption monitoring unit 103 and receives a designation request from the job control unit 101 for the information processing units 1 to 1 nm as a job distribution destination, the power calculation / comparison unit 102 monitors the power consumption. The information of the power consumption of the information processing units 11 to nm is requested to the unit 103, and the power consumption of each of the information processing units 1 to nm is acquired. The power calculation / comparison unit 102 that has acquired the information uses the information in the storage unit 104 that stores the power conversion efficiency of each of the power units 1 to n and the information on the power consumption of each of the information processing units 11 to nm that has been acquired. Information processing units 11 to nm that are job distribution destinations that maximize the power conversion efficiency of the system, that is, the power consumption of the entire system is minimized, and the information processing unit 11 that is the job distribution destination to the job control unit 101 Specify ~ nm.

  The power consumption monitoring unit 103 monitors the power consumption of the information processing units 11 to nm, receives a request from the power calculation / comparison unit 102, and sends the power consumption of the information processing units 11 to nm to the power calculation / comparison unit 102. Send information.

  Next, the operation of the power calculation / comparison unit 102 will be described with reference to FIGS. 2 and 3 in the operation of each component configuration of FIG.

  First, when the job control unit 101 receives a job processing request from the clients 1000-1 to 1000-k (step S0), the job control unit 101 sends a designation request for the information processing units 11 to nm to be distributed to the power calculation / comparison unit 102. Request (step S1).

  Next, the power calculation / comparison unit 102 that has received the request requests the power consumption monitoring unit 103 for information on the power consumption of the information processing units 11 to nm (step S2), and the power consumption W11 to 11 of the information processing units 11 to nm. Wnm information is acquired (step S3).

  Next, the powers Wo1 to Won output from the power supply units 1 to n are obtained from the information on the power consumption W11 to Wnm acquired in step S3 (step S4). That is, the output powers Wo1 to Won of the power supply units 1 to n are equal to the sum of the power consumptions W11 to Wnm of the information processing units 11 to nm that are supplied with power from the power supply units 1 to n.

となる。

It becomes.

  Next, based on the output power Wo1 to Won of the power supply units 1 to n obtained in step S4 and the information on the power-conversion efficiency characteristics (FIG. 3) of the power supply units 1 to n stored in the storage unit 104, the conversion efficiency E1 -En are indexed (step S5). The conversion efficiency characteristic of the power supply varies depending on the output power. By storing the power-conversion efficiency characteristic information as shown in FIG. 3 in the storage unit 104, the conversion efficiencies E1 to En can be indexed from the information of the output powers Wo1 to Won of the power supply units 1 to n. Become.

  Next, the input powers Wi1 to Win of the power supply units 1 to n are obtained from the conversion efficiencies E1 to En of the power supply units 1 to n as a result of step S5 (step S6). For example, the output voltage Won of the power supply unit n is equal to the product of the input power Win of the power supply unit n and its conversion efficiency En,

となる。この式を変形すると、電源部ｎの入力電力Ｗｉｎは、電源部ｎの出力電圧Ｗｏｎとその変換効率Ｅｎとの商に等しく、

It becomes. By transforming this equation, the input power Win of the power supply unit n is equal to the quotient of the output voltage Won of the power supply unit n and its conversion efficiency En,

となる。

It becomes.

  Next, when the job is distributed to the information processing unit 11 or 12 under the power supply unit 1, or when the job is distributed to the information processing unit 21, 22 or 23 under the power supply unit 2, the power supply units 1 to n When jobs are distributed to the subordinate information processing units 11 to nm, the expected output powers of the power supply units 1 to n (hereinafter, referred to as the power increases WA (which may be a fixed value) corresponding to the job type) (Estimated output power) WAo1 to WAon are obtained from the output powers Wo1 to Won of the power supply units 1 to n as the result in step S4 (step S7). For example, the expected output power WAon of the power supply unit n is equal to the sum of the output power Won and the power increase WA according to the job type,

となる。

It becomes.

  Next, the predicted output power WAo1 to WAon of the power supply units 1 to n obtained in step S7 and the information on the power-conversion efficiency characteristics (FIG. 3) of the power supply units 1 to n stored in the storage unit 104 are predicted. Conversion efficiency (hereinafter, expected conversion efficiency) EA1 to EAn are indexed (step S8).

  Next, predicted input powers (hereinafter, predicted input power) WAi1 to WAin of the power supply units 1 to n are obtained from the predicted conversion efficiencies EA1 to EAn that are the results in step S8 (S9). For example, the expected input power WAin of the power supply unit n is equal to the quotient of the output voltage Won of the power supply unit n and its expected conversion efficiency EAn,

となる。

It becomes.

  Next, the information processing unit 11 or 12 under the power supply unit 1 is determined from the input power Wi1 to Win of the power supply units 1 to n obtained in step S6 and the predicted input power WAi1 to WAin of the power supply units 1 to n obtained in step S9. When the job is distributed to the information processing units 11 to 22 under the power supply units 1 to n, such as when the job is distributed to the information processing units 21, 22 or 23 under the power supply unit 2 The input power of the entire system (hereinafter, “predicted input power”) T1 to Tn is calculated (step S10).

  For example, the expected input power T1 of the entire system when a job is distributed to the information processing unit 11 or 12 under the power supply unit 1 is the expected input power WAi1 of the power supply unit 1, the input power Wi2 of the power supply unit 2,. This is the sum of n input powers Win. Further, the expected input power T2 of the entire system when a job is distributed to the information processing unit 21, 22 or 23 under the power supply unit 2 is the input power Wi1 of the power supply unit 1, the expected input power WAi2 of the power supply unit 2,. This is the sum of the input power Win of the power supply unit n. Similarly, the predicted input power Tn of the entire system when jobs are distributed to the information processing units n1 to nm under the power supply unit n is the input power Wi1 of the power supply unit 1, the input power Wi2 of the power supply unit 2,. This is the sum of the expected input power WAin of the power supply unit n. That is, the expected input power T1, T2,.

となる。

It becomes.

  Next, the power supply units 1 to n that have the smallest values among the predicted input powers T1 to Tn of the entire system that are the results of step S10 are determined (step S11).

  Next, the information processing units 11 to nm having the minimum power consumption of the information processing units 11 to 11 to which power is supplied from the power supply units 1 to n determined in step S10 are determined (step S12). When determining the information processing unit 11-nm with the lowest power consumption, it is determined in a predetermined number (predetermined identification number) order, for example, in ascending order of number (young number), or randomly (arbitrarily) You may decide.

  Then, the power calculation / comparison unit 102 transmits the designation of the information processing units 11 to nm of the job distribution destination determined in step S12 to the job control unit 101 (step S13), and the job control unit 101 is designated. The job is distributed to the information processing units 11 to nm (step S14).

  As described above, in the present embodiment, the power consumption monitoring unit 103 monitors the power consumption of each of the information processing units 11 to nm and reports it to the power calculation / comparison unit. Then, the power calculation / comparison unit 102 having the storage unit 104 that stores the power-conversion efficiency characteristics of the power units 1 to n transfers the jobs from the clients 1000-1 to 1000-k to the information processing units 11 to nm. In response to a request from the job control unit 101 that distributes, information processing units 11 to nm that distribute jobs are determined in consideration of power consumption and power conversion efficiency at that time.

  That is, in the present embodiment, in a distributed processing system composed of a plurality of power supply units 1 to n and information processing units 11 to nm, the power consumption of each information processing unit 11 to nm and the power supply units 1 to n are Based on the information on the power-conversion efficiency characteristics, the information processing units 11 to nm that distribute jobs that maximize the power conversion efficiency of the entire system, that is, minimize the power consumption of the entire system, are determined.

  In other words, in the present embodiment, the power-conversion efficiency characteristics of each of the power supply units 1 to n are provided, and the power consumption of the entire system is minimized based on the information about the information and the power consumption of each of the information processing units 11 to nm. It is possible to determine the information processing units 11 to nm for distributing such jobs.

  Therefore, according to the present embodiment, it is possible to minimize the power consumption of the entire system at all times by maximizing the power conversion efficiency of the entire system at all timings for distributing jobs. By having the power-conversion efficiency characteristic information of the units 1 to n, the system power consumption can always be minimized without depending on the power source type / configuration.

  The above distributed processing system and its job distribution method can be realized by hardware, software, or a combination thereof.

  For example, the above distributed processing system can be realized by hardware, but can also be realized by a computer reading a program for causing the computer to function as the distributed processing system from a recording medium and executing the program. it can.

  Also, the job distribution method of the above distributed processing method can be realized by hardware, but the computer reads a program for causing the computer to execute the job distribution method from a computer-readable recording medium and executes it This can also be realized.

  Further, the hardware and software configurations described above are not particularly limited, and any hardware and software configurations can be applied as long as the functions of the respective units described above can be realized. For example, any of the above-described components may be individually configured with independent circuits, components, or the like, or may be configured integrally by incorporating a plurality of functions into one circuit, component, or the like.

  Although the present invention has been described with reference to the embodiments, the present invention is not limited to the above embodiments. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

  As described above, the present invention can be used for a distributed processing system suitable for power saving, a job distribution method, and a program thereof.

1 to n power supply units 11 to 2 m information processing unit 100 system common processing unit 101 job (JOB) control unit 102 power calculation / comparison unit (job distribution destination determination unit)
103 power consumption monitoring unit 104 storage unit 1000-1 to 1000-k client

Claims (6)

Multiple power supplies,
A plurality of information processing units connected to any of the subordinates of each of the power supply units and supplied with power from each of the power supply units;
A system common processing unit connected in common between the client and each of the information processing units;
The system common processing unit is
A storage unit for storing power-conversion efficiency characteristics of each power supply unit;
A power consumption monitoring unit that monitors power consumption of each of the information processing units;
When distributing a job from the client to each information processing unit, the power consumption of each information processing unit monitored by the power consumption monitoring unit and the power-conversion of each power source unit stored in the storage unit A job distribution destination determination unit that determines an information processing unit of a job distribution destination so that the power conversion efficiency of the entire system is maximized from the information with the efficiency characteristics;
A distributed processing system, comprising: a job control unit that distributes a job from the client to an information processing unit of a determined job distribution destination. The job distribution destination determination unit
From the information on the power consumption of each information processing unit, obtain the output power output from each power supply unit,
Indexing the conversion efficiency based on the output power of each power supply unit and information on the power-conversion efficiency characteristics of each power supply unit stored in the storage unit, and obtaining the input power of each power supply unit from the result,
When the job is distributed to each of the information processing units under each of the power supply units, the expected output power of each power supply unit in consideration of the power increase corresponding to the type of job is obtained,
Based on the information of the predicted output power of each power supply unit and the power-conversion efficiency characteristics of each power supply unit stored in the storage unit, the expected conversion efficiency is indexed, and the respective power supplies are indexed from the result. The expected input power of the
The expected input power of the entire system expected when the job is distributed to the information processing unit under each power supply unit from the input power of each power supply unit and the expected input power of each power supply unit for each power supply unit Seeking
Determining a power supply unit with a minimum expected input power value for the entire system;
The distributed processing system according to claim 1, wherein among the information processing units supplied with power from the determined power source unit, the one with the minimum power consumption is determined as an information processing unit as a job distribution destination. Multiple power supplies,
A plurality of information processing units connected to any of the subordinates of each of the power supply units and supplied with power from each of the power supply units;
A job distribution method for a distributed processing system having a system common processing unit commonly connected between a client and each of the information processing units,
The system common processing unit is
Storing the power-conversion efficiency characteristics of each power supply unit;
Monitoring the power consumption of each information processing unit,
When distributing a job from the client to each information processing unit, the power consumption of each information processing unit monitored by the power consumption monitoring unit and the power-conversion of each power source unit stored in the storage unit From the information on the efficiency characteristics, determine the information processing unit of the job distribution destination so that the power conversion efficiency of the entire system is maximized,
A job distribution method for a distributed processing system, wherein a job from the client is distributed to an information processing unit of a determined job distribution destination. When determining the job distribution destination information processing unit,
From the information on the power consumption of each information processing unit, obtain the output power output from each power supply unit,
Indexing the conversion efficiency based on the output power of each power supply unit and information on the power-conversion efficiency characteristics of each power supply unit stored in the storage unit, and obtaining the input power of each power supply unit from the result,
When the job is distributed to each of the information processing units under each of the power supply units, the expected output power of each power supply unit in consideration of the power increase corresponding to the type of job is obtained,
Based on the information of the predicted output power of each power supply unit and the power-conversion efficiency characteristics of each power supply unit stored in the storage unit, the expected conversion efficiency is indexed, and the respective power supplies are indexed from the result. The expected input power of the
The expected input power of the entire system expected when the job is distributed to the information processing unit under each power supply unit from the input power of each power supply unit and the expected input power of each power supply unit for each power supply unit Seeking
Determining a power supply unit with a minimum expected input power value for the entire system;
4. The job distribution method for a distributed processing system according to claim 3, wherein the information processing unit having the minimum power consumption among the information processing units supplied with power from the determined power source unit is determined as a job distribution destination information processing unit. Multiple power supplies,
A plurality of information processing units connected to any of the subordinates of each of the power supply units and supplied with power from each of the power supply units;
A job distribution program for a distributed processing system having a system common processing unit connected in common between a client and each of the information processing units,
A computer constituting the system common processing unit;
A storage unit for storing power-conversion efficiency characteristics of each power supply unit;
A power consumption monitoring unit that monitors power consumption of each of the information processing units;
When distributing a job from the client to each information processing unit, the power consumption of each information processing unit monitored by the power consumption monitoring unit and the power-conversion of each power source unit stored in the storage unit A job distribution destination determination unit that determines an information processing unit of a job distribution destination so that the power conversion efficiency of the entire system is maximized from the information with the efficiency characteristics;
A computer-readable program for causing a job control unit to distribute a job from the client to an information processing unit of a determined job distribution destination. The job distribution destination determination unit
From the information on the power consumption of each information processing unit, obtain the output power output from each power supply unit,
Indexing the conversion efficiency based on the output power of each power supply unit and information on the power-conversion efficiency characteristics of each power supply unit stored in the storage unit, and obtaining the input power of each power supply unit from the result,
When the job is distributed to each of the information processing units under each of the power supply units, the expected output power of each power supply unit in consideration of the power increase corresponding to the type of job is obtained,
Based on the information of the predicted output power of each power supply unit and the power-conversion efficiency characteristics of each power supply unit stored in the storage unit, the expected conversion efficiency is indexed, and the respective power supplies are indexed from the result. The expected input power of the
The expected input power of the entire system expected when the job is distributed to the information processing unit under each power supply unit from the input power of each power supply unit and the expected input power of each power supply unit for each power supply unit Seeking
Determining a power supply unit with a minimum expected input power value for the entire system;
6. The program according to claim 5, wherein the information processing unit having the minimum power consumption among the information processing units supplied with power from the determined power source unit is determined as an information processing unit as a job distribution destination.

Images