WO2012124295A1 - コンピュータシステム、制御システム、制御方法および制御プログラム - Google Patents
コンピュータシステム、制御システム、制御方法および制御プログラム Download PDFInfo
- Publication number
- WO2012124295A1 WO2012124295A1 PCT/JP2012/001563 JP2012001563W WO2012124295A1 WO 2012124295 A1 WO2012124295 A1 WO 2012124295A1 JP 2012001563 W JP2012001563 W JP 2012001563W WO 2012124295 A1 WO2012124295 A1 WO 2012124295A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- control
- data
- task
- computer system
- characteristic
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5083—Techniques for rebalancing the load in a distributed system
- G06F9/5088—Techniques for rebalancing the load in a distributed system involving task migration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
Definitions
- the present invention relates to a computer system, a control system, a control method, and a control program.
- Patent Document 1 describes a system in a distributed storage system that calculates a no-access probability based on access statistical information to a distributed storage node, and relocates data using the calculated no-access probability. ing.
- the system described in Patent Document 1 can lengthen the time during which each disk node is in the power saving mode by performing data arrangement so that the expected value of no-access time is high.
- Patent Document 2 describes a computer system that appropriately distributes an I / O (Input / Output) load.
- the computer system described in Patent Literature 2 calculates an average of I / O access frequencies for each integrated device as an expected value of I / O access frequencies.
- Non-Patent Document 1 describes a time-series analysis technique for capturing and expressing characteristics of data from time-series data such as temperature and earthquake.
- the most essential problem in realizing computer system control is that there is a lack of information for controlling jobs (applications) and data.
- the above-described technical approach has a problem that data access unknown to the computer system cannot be predicted.
- the above-described technical approach has a problem that it cannot cope with fluctuations in access load and the like that have no precursor information for a computer system.
- a computer system provides a service related to some real world. Therefore, it is common for humans to monitor and judge the situation in the real world to cope with predictable load fluctuations on the computer system.
- a simple example would be to increase server capacity or expand applications for the Christmas season at a shopping site. That is, it can be said that a change in the external environment of the computer system can suggest a change in usage pattern expected in the near future of the system.
- the current computer system performance problem is often caused by a bottleneck for data I / O (Input / Output). For this reason, if problems associated with data I / O can be avoided, it is considered that many performance problems can be solved and adaptable to changes. For example, it is conceivable to realize control in which data is distributed and stored in advance so that data I / O caused by the load is distributed as the load increases.
- the data structure has advantages and disadvantages depending on the pattern of access to the data, so the data structure can also be controlled. It is preferable. In addition, it is necessary to be able to control task placement that appropriately leveles the CPU load and does not cause a network bottleneck associated with data I / O. Furthermore, in order to continue operating the system with power according to the required performance, it is preferable that control over system components (such as stopping the power supply of the node) is also possible. Thus, in order to adapt flexibly to changes in the usage pattern of computer systems, it is necessary to be able to autonomously control not only data placement but also data structure, task placement, or system components. It is believed that there is.
- the problem with the computer system is that the computer system cannot flexibly adapt to changes in usage patterns and external environments.
- the reason is that the operating status of the system (computer load, the characteristics of the running application, the characteristics of the data being held and managed, etc.) is automatically changed according to changes in the usage form and changes in the external environment. This is because the control function for flexibly adapting the computer system to the usage form in response to changes in the external environment is not provided more than necessary (more than necessary and sufficient for the computer system).
- a computer system to flexibly adapt to changes in the system usage form and changes in the external environment. More specifically, a computer system, a control system, a control method, and a control program capable of performing control related to system components, task arrangement, data arrangement, or data structure of a computer system based on future access characteristics to data The purpose is to provide.
- the control system is a characteristic estimation means for estimating the characteristics of a task or data based on characteristic estimation source data related to a task operating on a computer system to be controlled or data held by the computer system. And one or more control execution means for performing stop / operation control on the system components of the computer system, changing the task placement, changing the data placement or changing the data structure according to the issued control command, and by the characteristic estimation means
- a control strategy to be executed based on the operation status of the computer system at a predetermined or specified time in the future derived from the estimated task or data characteristics stop / operation control and task for the system components of the computer system Change placement, change data placement, data structure Control strategy determining means for determining the control processing contents combining any one or more of the changes of the control, and a control instruction for issuing a control command to the control execution means in accordance with the control processing contents determined by the control strategy determining means And issuing means.
- the computer system is a characteristic estimation means for estimating a characteristic of a task or data based on characteristic estimation source data which is information related to a task operating on the computer system or data held by the computer system.
- one or more control execution means for performing stop / operation control on the system components of the computer system, changing task placement, changing data placement, or changing data structure in accordance with the issued control command, and characteristic estimation means
- control strategy to be executed based on the operation status of the computer system at a predetermined or specified time in the future derived from the task or data characteristics estimated by Change of arrangement, change of data arrangement, data structure
- Control strategy determining means for determining the control processing contents combining any one or more of the further, and control instruction issuing for issuing control instructions to the control execution means in accordance with the control processing contents determined by the control strategy determining means Means.
- control method provides a task or data characteristic based on characteristic estimation source data that is information related to a task operating on a computer system to be controlled or data held by the computer system.
- a control strategy to execute based on the computer system's operational status at a predetermined or specified time in the future derived from the estimated task or data characteristics -Determine the control processing content that combines any one or more of operation control, task allocation change, data allocation change, and data structure change, and in accordance with the determined control processing content, to the system components of the computer system Stop / operation control, task placement change, data placement change or For one or more control execution means for changing the data structure, and issues a control command.
- control program allows a computer to execute a task or a task based on characteristic estimation source data that is information related to a task operating on a computer system to be controlled or data held by the computer system.
- the system of the computer system as a control strategy to be executed based on the operation status of the computer system at a predetermined or specified time in the future derived from the process of estimating the characteristics of the data, the estimated task or the characteristics of the data Processing for determining control processing contents combining any one or more of stop / operation control for component, task placement change, data placement change, data structure change, and computer according to the determined control processing content Stop / operation control for system components Changing Task arrangement, for one or more control execution means for changing the change or the data structure of the data arrangement, characterized in that to execute a process for issuing a control command.
- the present invention estimates the characteristics of data and tasks using arbitrary characteristic estimation source data, and uses data or task characteristic models with similar characteristics when the characteristic estimation source data is insufficient. It is characterized by tentatively estimating characteristics of target data or tasks. Further, the present invention is characterized in that control execution is determined using the model information and information from an external information source, and a control command is issued to each component of the system.
- FIG. 1 is a block diagram illustrating a configuration example of a control system of a computer system according to an embodiment of this invention.
- the control system of the computer system shown in FIG. 1 includes four types of control strategy execution as a data / task characteristic estimation unit 1, a characteristic record management unit 2, a control strategy determination / command issue unit 3, and a control strategy execution unit group 4.
- Means that is, system component element control execution means 41, task arrangement strategy control execution means 42, data arrangement strategy control execution means 43, and data structure strategy control execution means 44).
- the data / task characteristic estimation means 1 estimates the data held and managed by the computer system to be controlled and the characteristics of the task processed on the computer system.
- the data / task characteristic estimation unit 1 includes a characteristic estimation source data acquisition / holding unit 11 and a characteristic estimation process execution unit 12. These may be implemented as software operating on a computer. Further, the data / task characteristic estimation means 1 may include a plurality of characteristic estimation source data acquisition / holding means 11 and a plurality of characteristic estimation processing execution means 12.
- a task refers to a processing execution unit of an application or service realized on a computer system to be controlled.
- data characteristics or task characteristics may include current or future usage or behavior of the data or task (e.g., when it will be executed, when it will be accessed), and the tendency of access to that data (e.g., Read excessive, Write excessive, sequential access, random access).
- a method, a calculation formula, and a calculation program for expressing these data and task characteristics are called a characteristic model.
- the characteristic estimation source data acquisition / holding means 11 acquires and stores (stores) data for performing characteristic estimation (characteristic estimation source data).
- the characteristic estimation source data acquisition / holding unit 11 is realized by, for example, software that acquires data and task history logs, software that records the history logs, and a recording medium.
- the characteristic estimation source data may be data indicating one end of a task or data characteristic such as information indicating an actual usage state, and no special requirement is provided.
- the characteristic estimation source data may be at least data related to the estimation target data or task.
- the characteristic estimation processing execution means 12 estimates the characteristic model representing the characteristics of the data or the task using the data (history log etc.) stored in the characteristic estimation source data income / holding means 11, and the estimation result Is stored in the characteristic record management unit 2. Note that only one or more characteristic models are held on the system. Note that the characteristic model to be an estimation candidate may be prepared in advance or may be generated by the model generation unit 123 described later during the execution of the estimation process.
- the characteristic estimation processing execution unit 12 uses, for example, the characteristic estimation source data obtained from the characteristic estimation source data acquisition / holding unit 11, and uses the most as one of one or more arbitrary characteristic models as future data and task characteristics. Choose a suitable model, or a combination of such characteristic models. As the selection criterion, the probability of the characteristic model, that is, the degree (likelihood) suitable for the model representing not only the present data but also the future characteristic of the data or task as the estimation target is used.
- the “likelihood” includes a likelihood (function) obtained by capturing a numerical value indicating the value as a function.
- the characteristic record management unit 2 holds and manages characteristic information that is information of the characteristic model estimated by the characteristic estimation processing execution means 12.
- the characteristic record management unit 2 is realized by, for example, database software, a computer that operates the database software, and a storage medium.
- FIG. 2 is a block diagram showing a more detailed configuration example of the characteristic estimation processing execution means 12.
- the characteristic estimation process execution unit 12 may include a matching model finding unit 121, a model set recording unit 122, and a model generation unit 123.
- the conforming model finding means 121 selects a characteristic model or combination of characteristic models most suitable for the target data or task from the characteristic model group stored in the model set recording unit 122 as the conforming model.
- the matching model finding unit 121 receives data from the property estimation source data acquisition / holding unit 11, performs likelihood calculation, and selects a characteristic model showing the best likelihood above a threshold as a matching model.
- the adaptive model selected in this way is stored in the characteristic record management unit 2 as a characteristic model estimation result.
- the model set recording unit 122 holds and manages information on characteristic models that are candidates for likelihood calculation, that is, characteristic models that are candidates for data or task characteristic models of the system to be controlled.
- the model set recording unit 122 is realized by, for example, database software, a computer that operates the database software, and a storage medium.
- the model generation unit 123 generates a new model and stores it in the model set recording unit 122 when there is no model suitable for the characteristic estimation source data in the existing characteristic model. Note that this system can also be configured by omitting the model generation means 123.
- FIG. 3 is a block diagram showing another configuration example of the characteristic estimation processing execution means 12.
- the characteristic estimation process execution unit 12 may include a model set recording unit 122, a class matching state model finding unit 124, and a data matching model finding unit 125.
- FIG. 2 the configuration example of FIG. 2 described above is based on the premise that there is a one-to-one correspondence between the data or task whose characteristics are to be estimated and the corresponding characteristic model. However, there are cases where it is better not to treat data or tasks and characteristic models as one-to-one, but treat data or tasks as a set. For example, when the access time zone for data is determined for each user, it is expected that a characteristic model should be applied to each metadata (for example, the owner user) attached to such data. .
- class such a set of data is referred to as a “class” in the present invention.
- the class include a user, a group, or a similar set clustered by content similarity (for example, a document describing a topic of XX).
- the class conformity state model finding means 124 selects an appropriate characteristic model or a combination of the characteristic models from the characteristic model group stored in the model set recording unit 122, similarly to the conforming model finding means 121 shown in FIG. To do.
- the class conformity state model finding unit 124 receives data from the property estimation source data obtaining / holding unit 11, performs likelihood calculation, and selects a characteristic model that shows the best likelihood above a threshold as an application model.
- the class conformity state model finding unit 124 stores the characteristic model thus selected in the characteristic record management unit 2 as an estimation result.
- the likelihood calculation of the adaptive characteristic model is not performed for each data, but using the characteristic estimation source data of the data set belonging to the class to which the data that is the target of characteristic estimation belongs, A matching characteristic model for the class is found (estimated) and stored in the characteristic record management unit 2.
- the data matching model finding unit 125 calculates a characteristic model for each data from the characteristic model found by the class matching state model finding unit 124. Data may belong to multiple classes. Therefore, the data matching model finding unit 125 estimates a characteristic model for each data using a method such as taking the sum of the matching characteristic models or selecting a model with the highest likelihood. The data matching model finding unit 125 may analyze individual characteristic estimation source data of the target data, and may estimate a data-specific characteristic model if it is not suitable for any of the class matching models.
- the characteristic record management unit 2 holds and manages characteristic information (in this embodiment, characteristic model) estimated by the characteristic estimation processing execution means 12.
- the characteristic record management unit 2 is realized by, for example, database software, a computer that operates the database software, and a storage medium.
- FIG. 4 is an explanatory diagram showing an example of a data holding format of the characteristic record management unit 2.
- the characteristic record management unit 2 may store an identifier (“Oid”) that specifies data and a characteristic model (“compatible state model”) that matches the identifier.
- Oid identifier
- compatible state model characteristic model
- characteristic models A and B are characteristic models expressing states 1 and 2, respectively.
- the characteristic model A (state 1) is a model like a linear equation indicating the characteristic that the access amount increases in proportion to time
- the characteristic model B (state 2) has many accesses during the day. It may be a model such as a trigonometric function showing characteristics such that there are few nights.
- the characteristic model information is not stored as it is, but may be stored as one-stage converted information (information on a control instruction that can be derived using the characteristic model).
- FIG. 5 is an explanatory diagram showing another example of the data holding format of the characteristic record management unit 2. For example, a control that estimates a time zone (Idle time zone) in which future data is not used, stores the unused time in a specific node, and stops the specific node in the time zone. When performing the above, information on the Idle time zone may be stored for each data identifier. Thereby, the calculation at the time of control becomes easy.
- Idle time zone time zone in which future data is not used
- Idle time zone information can be derived from the estimated characteristic model (in this example, “characteristic model A”). Accordingly, such calculation for derivation only needs to be performed before the storage in the characteristic record management unit 2.
- the control strategy determining / command issuing means 3 determines a control strategy that meets the conditions based on the data or task characteristic information estimated by the data / task characteristic estimating means 1 and performs control to execute the determined control strategy.
- a control command is issued to the strategy execution means group.
- the control strategy determination / command issuing means 3 includes a control strategy determination / command issuing means 31.
- the control strategy determination / command issuing means 31 is a control strategy execution means provided in the system for determining a control strategy using the characteristic information stored in the characteristic record management unit 2 and providing a control instruction based on the control strategy. It is appropriately issued to the system component control execution means 41, the task arrangement strategy control execution means 42, the data arrangement strategy control execution means 43, and the data structure strategy control execution means 44.
- the control strategy determination / instruction issue means 31 is realized by software operating on a computer, for example.
- the control strategy execution means group 4 is realized by, for example, software that runs on a computer and hardware such as a computer device that is controlled in conjunction with the software.
- the physical operation execution location of the control strategy execution means 41 to 44 varies depending on the application system.
- the control strategy execution means 41 to 44 need not necessarily be configured to operate on the same computer.
- a plurality of control strategy execution means 41 to 44 may exist, and in such a case, each may operate on a different computer.
- the control strategy execution means 41 to 44 may be any means that can actually control the system to be controlled, and any method for mounting the means may be used.
- a configuration provided in the control target system may be used, or a configuration provided in another system (such as a management system for managing the control target system) different from the control target system may be used.
- the other system refers to another system having a function such that the state of the control target system is changed by controlling the other system.
- the system component control execution means 41 controls the operating state of the elements constituting the system to be controlled (hereinafter referred to as system component elements) in accordance with the control command from the control strategy determination / command issuing means 31.
- system component elements include software that changes an operation mode defined in ACPI (Advanced Configuration and Power Interface) in a node computer constituting a distributed computer system.
- the system component to be controlled may be not only a computer but also a smaller unit such as a hard disk. For example, stopping the rotation of the platter by the driver software on the hard disk or stopping the power supply of the entire hard disk also corresponds to the control of the system components.
- changing the rotation frequency of the CPU or starting and stopping the GPU board also corresponds to the control of the system components.
- the task placement strategy control execution means 42 determines which of the system components to execute a task executed on the system, The method (determination destination) is controlled.
- the task allocation strategy control execution means 42 controls the task allocation method for the task allocation function unit (software) in the distributed computer system.
- the task placement strategy control execution means 42 sets an operation algorithm when a plurality of task assignment algorithms are implemented in software.
- the task allocation strategy control execution means 42 may be changed from a setting assigned by the round robin algorithm to a setting for changing the assignment ratio depending on the load state of each node being measured.
- a method of preferentially allocating a node storing data used by a task may be considered.
- task placement determination control it may be possible to enable control such as changing the placement node according to an identifier assigned to a task.
- a placement node for example, a node on which hardware (such as GPGPU or GPGA) capable of executing a specific process at high speed or a node on which software with high-speed processing is installed is installed.
- the identifier given to the task may be given by any job execution middleware, for example.
- the data arrangement strategy control execution means 43 determines in which system component the data is stored in the realized system according to the control command from the control strategy determination / command issuing means 31. It should be noted that the determination of the data storage destination preferably includes a function of moving the arranged data or increasing the number of replicas.
- the data placement strategy control execution means 43 is realized by software for placing data in a distributed storage system, for example. Further, for example, in the data arrangement algorithm of the distributed storage system, it can be realized by mounting a plurality of arrangement algorithms and changing which one is used.
- the data arrangement strategy control execution unit 43 receives a data movement command from the control strategy determination / command issuing unit 31 and executes various control processes such as a process of moving data.
- the data structure strategy control execution means 44 changes the method of determining the storage method for the means for determining which structure is used when data can be stored in a plurality of structures in the computer system. , Change the storage method of existing data.
- An instruction to change the storage data storage method is issued to the data structure strategy control execution means 44 by the control strategy determination / instruction issue means 31, and the data structure strategy control execution means 44 executes the data storage method in accordance with the instructions.
- the data structure strategy control execution means 44 executes processing such as storing data stored in the hard disk on the memory, or changing the stored database (data migration).
- FIG. 6 is a block diagram showing a more detailed configuration example of the control strategy determination / command issuing means 31.
- the control strategy determination / instruction issue means 31 may include an instruction issue means 311, a processing rule pair control process content storage unit 312, an information acquisition means 313, and a condition determination means 314. Good.
- the instruction issuing means 311 is a system component control execution means 41, a task placement strategy control execution means 42, a data placement strategy control execution means 43, a data structure strategy, which are four types of control strategy execution means provided in the system.
- a control command is actually issued to the control execution means 44.
- the instruction issuing unit 311 generally transmits an instruction to each of the strategy control execution units 41 to 44 operating on a computer different from the computer on which the instruction issuing unit 311 operates via a network interface.
- the control command to be issued is passed from the condition determining unit 314 to the command issuing unit 311.
- the processing rule pair control processing content storage unit 312 is a storage device that stores a condition (rule) for performing arbitrary control and information on control command content in association with each other. Note that a rule pair control processing content input / output unit 315 may be further provided as means for editing the contents of the processing rule pair control processing content storage unit 312.
- the information acquisition unit 313 acquires information necessary for the condition determination by the condition determination unit 314 from the characteristic record management unit 2 and the other information management unit 21. Information such as what information is acquired from where may be determined in advance by a setting file or the like. The information acquisition unit 313 acquires not only information indicating the status of the computer system to be controlled but also information indicating the real-world status that can be acquired via the other information management unit 21, for example.
- the other information management unit 21 is any information holding unit or information providing unit other than the characteristic record management unit 2.
- Examples of information provided by the other information management unit 21 include the following. -Weather forecast-Power information of the data center where the system is operating-Load information of the computer nodes that make up the system-Log information of any services that run on the computer nodes that make up the system-On the WWW (World Wide Web) Information (such as services called web pages, microblogs) ⁇ TV broadcast contents, TV broadcast caption information
- the system can be flexibly adapted to changes in the real world.
- the condition determination unit 314 uses the contents of the processing rule pair control process content storage unit 312 and the information obtained from the information acquisition unit 313 to store the characteristics of each task or each data stored in the characteristic storage management unit 2. Based on the model, it is determined whether or not the control processing content stored in the processing rule pair control processing content storage unit 312 should be executed as a control strategy. Specifically, the condition determination unit 314 stores the status indicated by the information obtained from the information acquisition unit 313 (including not only the system status but also the real world status) in the processing rule pair control processing content storage unit 312. A control strategy to be executed is determined by determining whether or not a condition (rule) for performing arbitrary control is satisfied. When the condition determining unit 314 determines that the condition is satisfied, the condition determining unit 314 issues an instruction to each of the strategy control executing units 41 to 44 so that the instruction issuing unit 311 executes the control content indicated by the control processing content. Ask.
- control conditions rules
- control processing contents are described with examples.
- a case where a recent node with a small access amount is stopped for power saving will be described as an example.
- a rule “when there is a node whose access amount in the most recent time is below a certain value” applies.
- the control instruction set (control processing content) is “stop nodes whose access amount is below a certain value”.
- the operation is provided as access log information for each node as the other information management unit 21, and the information acquisition unit 313 acquires the access log information. Then, the condition determination unit 314 detects a node whose access amount in the latest ⁇ time is equal to or less than a certain value from the access log information.
- the control instruction set is issued to the strategic control execution means that can control the target node.
- the instruction issuing unit 311 instructs the system component element control execution unit 41 (for example, a power control program) that can control the node No. 1 to stop the power supply. Issue.
- the determination as to whether or not the conditions for performing the control are met not only uses the information obtained from the information acquisition unit 313 directly, but also a certain time in the future (a time specified by a rule, a predetermined time, etc.) Is determined together with the situation of the control target system derived by fitting to the characteristic model. For example, a condition for performing control by assigning a variable name to data obtained by the information obtaining unit 313 in advance or time information (or elapsed time from the present time) indicating a certain date and time in the future to the characteristic model. By giving conditional expressions expressed using these variables, the condition determining means 314 can determine whether or not the control target system matches the conditions for controlling at that time by function calculation. It may be.
- condition for performing control in addition to the conditional expression, date / time information as a determination criterion, data to be used, information indicating a task identifier or range, and the like may be given. Further, not the conditional expression but the determination processing program itself may be given to the system. For example, an address where the determination processing program is actually stored may be given as rule information, and the condition determination means 314 may execute the determination processing program.
- the characteristic model of the data held by the node 1 is acquired from the characteristic record management unit 2, and the data identifier (the above data A) having a high data access probability is extracted from the acquired characteristic model. Then, a process for instructing the data placement strategy control means 34 to move the data A to another node may be added before node 1 stops.
- the command issuing means 311 may instruct the task placement strategy control executing means 42 to use a node other than node 1 as the task placement strategy related to the data A. Further, the command issuing means 311 does not necessarily move the data, but only READ as the characteristic model of the data A, that is, if it is known that the access method to the data A is limited to the READ, the data A An instruction may be issued to the data placement strategy control unit 34 and the task placement strategy control execution unit 42 so that the READ target node is a different node.
- the control strategy determination instruction issue means 31 is a system component control execution means 41, a task placement strategy control means 33, a data placement strategy control execution means 43, and a data structure strategy control execution.
- system control such as power saving can be performed.
- the processing rule pair control processing content storage unit 312 can hold a plurality of rules as described above. Each rule may be stored in association with a corresponding control instruction set.
- the rule there is system control such that when a product name handled in TV caption information is displayed in a shopping site of a certain product, resources of the shopping site are increased.
- the rule is “when a specific product name is detected in the external information recording unit that is subtitle information of TV”, and the control instruction set is “increase the number of computers for shopping sites (increase computer nodes, load Change the data placement algorithm to be distributed, distribute the shopping site data, create replicas for load balancing, and change the task placement algorithm for the shopping site to accommodate the increased number of computers) ” This is true.
- the rule is “the power consumption monitor value of the entire data center has been recorded above a certain level”
- the control instruction set is “degenerate the number of operation computers (task data supported by nodes with low load). It is also possible to consolidate to other nodes, that is, change the data arrangement and task arrangement to increase the number of nodes that can be stopped, and issue a node stop instruction). In this way, power saving can be achieved.
- FIG. 7 is a flowchart illustrating an example of a process flow of the characteristic estimation process of the present embodiment.
- the characteristic estimation process is executed periodically, for example.
- the execution cycle may be any time.
- step S101 data (characteristic estimation source data) used when performing characteristic estimation while an arbitrary service or computer operating on a computer system to be controlled by the control system of the present embodiment executes a predetermined function.
- Generate step S101 in FIG. 7.
- an access log to a storage node generated (updated) when a storage access occurs in a distributed storage system can be cited.
- Step S101 is regularly executed irrespective of the characteristic estimation process in the control system. Therefore, the characteristic estimation process starts substantially from step S102.
- the characteristic estimation source data acquisition / holding means 11 acquires the characteristic estimation source data generated in step S101 at a predetermined timing (such as regular or updated timing) (step S102).
- a predetermined timing such as regular or updated timing
- the software that implements the characteristic estimation source data acquisition / holding unit 11 performs the storage access log data reading process described above.
- the characteristic estimation source data is not limited to storage access log data, and is not limited to one type.
- the characteristic estimation process execution means 12 uses the characteristic estimation source data obtained in step S102 to perform tasks on the computer system to be controlled and data held by the system.
- a characteristic model expressing the characteristic is estimated (calculated) (step S103). Details of step S103 will be described later.
- the characteristic estimation processing execution unit 12 associates the task or data that is the target of characteristic estimation with characteristic model information that is information indicating the estimated characteristic estimation model (characteristic model). Then, it is recorded in the characteristic record management unit 2 (step S104).
- the characteristic estimation process execution unit 12 writes (updates) the content of the estimated characteristic model in, for example, a database that implements the characteristic record management unit 2.
- steps S102 to S104 are repeatedly executed.
- FIG. 8 is a flowchart showing an example of a more detailed processing flow of the feature model estimation processing (step S103 in FIG. 7).
- the matching model discovery unit 121 of the characteristic estimation process execution unit 12 first receives the characteristic estimation source data from the characteristic estimation source data acquisition / holding unit 11 (step S201), and the task or data to be estimated In step S202, it is determined whether or not there is characteristic estimation source data of a certain amount or more. Here, if there is characteristic estimation source data of a certain amount or more, the process proceeds to step S203, and if not, the process proceeds to step S209.
- the matching model finding unit 121 acquires information on candidate characteristic models from the model set recording unit 122 (step S203). Then, the matching model finding unit 121 performs likelihood calculation using the acquired characteristic estimation source data for each acquired model and each combination of models (step S204). Next, the matching model finding unit 121 determines whether any of the calculated likelihoods is equal to or greater than a predetermined threshold value (step S205). If there is one or more models with a likelihood equal to or greater than the threshold, the matching model finding unit 121 selects the model with the best likelihood among them as the characteristic model of the task or data that is the current estimation target (step S206). ). If there is no model with a likelihood equal to or greater than the threshold, the model generation unit 123 creates a new model from the characteristic estimation source data and records it in the model set recording unit 122 (step S207).
- step S208 the model information selected or estimated in steps S206, S207, and S209 is recorded in the characteristic record management unit 2 in association with the data or task information that is the estimation target. If there is no model generation unit 123, the characteristic may be unknown in step S207, and information indicating the characteristic unknown may be recorded in the characteristic record management unit 2. Further, for example, a characteristic model for characteristic unknown data / characteristic unknown task may be prepared and recorded.
- step S209 a process of temporarily applying some model is performed.
- the following are examples of the provisional model determination process.
- a model for unknown characteristics is prepared, which is provisionally applied. This may be the same as the case where there is no model generation unit 123 described above.
- ⁇ ⁇ ⁇ Others are methods that apply a conforming model of other data or tasks from the data given to the estimation target or metadata attached to the tasks. For example, any suitable model of data or task to which the same metadata is assigned may be selected. Further, for example, in the case of a task exclusively for data or data access, the same conformity model as that of data having similar contents may be applied by comparing the content similarity of data.
- FIG. 9 is a flowchart illustrating an example of a processing flow of control strategy determination / execution processing according to the present embodiment.
- the information acquisition unit 313 acquires information from the characteristic record management unit 2 and other information management unit 21 and passes the information to the condition determination unit 314.
- the information acquisition unit 313 acquires various pieces of information from the characteristic record management unit 2 and other information management units 21 according to a predetermined method, for example.
- the condition determination unit 314 may request the information acquisition unit 313 to acquire necessary information based on the condition (processing rule) stored in the processing rule pair control processing content storage unit 312.
- condition determination unit 314 passes information for specifying necessary information described in the rule to the information acquisition unit 313, and the information acquisition unit 313 performs the characteristic record management unit 2 based on the information.
- information may be acquired from other information management unit 21.
- the condition determination means 314 determines whether the passed information or the computer system or other situation indicated by the information is suitable for any of the conditions stored in the processing rule vs. control processing content storage unit 312. Determination is made (step S302). If the condition is met (Yes in step S303), the condition determining unit 314 determines the control processing content with the highest priority among the matched conditions as a control strategy to be executed, and controls the instruction issuing unit 311 to control it. Request to issue an order. In response to the request from the condition determining unit 314, the command issuing unit 311 issues necessary control commands to the strategy control executing units 41 to 44 in accordance with the control processing content determined as the control strategy to be executed (step S304). ).
- step S303 if there is no matching condition (No in step S303), the process is terminated without doing anything.
- FIG. 10 is a block diagram illustrating a configuration example of the present embodiment.
- the distributed storage system control system shown in FIG. 10 includes a network 500, data node computers 51 (51-1 to 51-3), a characteristic record DB node computer 54, a control command node computer 55, and a job scheduler computer 56. And a meta database node 57.
- the network 500 is a network that connects elements (computers) constituting the system, and is realized by Ethernet (registered trademark) or the like. All other computers are IA servers equipped with a CPU and a storage device.
- a program or software is shown as one of the components of the present embodiment, but the program or software more specifically refers to a CPU that operates according to the program. To do. Therefore, when the expression “XX program does” is used, it means that the CPU provided in the computer executes processing related to the operation according to the program.
- the data node computer 51 is a computer that performs the function of holding the data of the distributed storage system.
- the data node computer 51-1 includes a distributed storage server realization program 511, an access log acquisition unit 512, an access log recording unit 513, a data recording unit 514, a data structure control unit 515, a data movement unit 516, a data Node computer power supply control means 517.
- the other data node computers 51 (51-2 and 51-3 in this example) have basically the same configuration as the data node computer 51-1, but may have different configurations. .
- the distributed storage server realization program 511 is a program for realizing a distributed storage node in the data node computer, and is a program for processing a data access request from a client.
- the access log acquisition means 512 is a program that generates and records an access log that is a history of access to the distributed storage node (more specifically, the data recording unit 514 provided in the data node computer).
- the access log recording unit 513 is a storage medium that stores the access log generated by the access log acquisition unit 512.
- the access log acquisition unit 512 and the access log recording unit 513 correspond to an example of the characteristic estimation source data acquisition / holding unit 11 described above.
- the data recording unit 514 is a storage medium and software that actually stores data held and stored in the distributed storage.
- the data structure control unit 515 is software that controls the storage format of the data stored in the data recording unit 514 in accordance with instructions from the control command issuing unit 551 of the control command node computer 55.
- the data structure control means 515 corresponds to an embodiment of the data structure strategy control execution means 44 described above.
- the data moving means 516 is software that performs a process of moving the data stored in the data node computer 51 to another node.
- the data moving means 516 includes communication means for exchanging data with other data node computers 51.
- the data moving unit 516 corresponds to an example of a part of the above-described data arrangement strategy control executing unit 43 (in this example, the movement source side). Further, in this embodiment, it is assumed that the information held and managed by the access destination determination unit 571 of the meta-database node 57 needs to be changed at the time of data movement.
- the data node computer power supply control means 517 is a program for shifting a node to a stopped state (or starting another node) using an OS command of the data node computer 51 or the like.
- the data node computer power supply control means 517 corresponds to an embodiment of the system component control execution means 41 described above.
- the characteristic record DB node computer 54 is a computer that performs the functions of the characteristic recording unit 2 and the characteristic estimation processing means 12.
- the characteristic record DB node computer 54 includes an estimation processing unit 541, a DB access processing unit 542, and a characteristic record management DB 543.
- the characteristic estimation processing means 541 is a program that operates on the characteristic record DB node computer 54, and estimates a characteristic model for each data and job on the distributed storage system to be controlled.
- the characteristic estimation processing unit 541 corresponds to an embodiment of the characteristic estimation processing unit 12 described above.
- the characteristic record management DB 543 is a storage medium and database infrastructure software that stores data and jobs on the distributed storage system in association with their identifiers and characteristic model information.
- the DB access processing means 542 is an access program for the database. In the present embodiment, the characteristic recording management DB 543 and the DB access processing means 542 realize the characteristic recording unit 2.
- the control command node computer 55 is a computer that realizes the control strategy determination / command issue means 31. More specifically, the control command issuing unit 551 performs the process. Although not shown, it is assumed that the control command issuing unit 551 has the configuration shown in FIG. 6, for example.
- the job scheduler computer 56 is a computer that performs a process of assigning a job to be processed on the distributed storage system to be controlled to each computer. Actually, the task assignment means 561 performs the processing.
- the job scheduler computer 56 includes task allocation strategy control means 562 in addition to task assignment means 561.
- the task allocation strategy control unit 562 is instructed by the control command issuing unit 551 of the control command node computer 55 and performs control related to job allocation (task allocation) to the task allocation unit 561.
- the task placement strategy control means 562 corresponds to an embodiment of the task placement strategy control execution means 33 described above.
- the meta database node 57 is a computer that determines the access destination of the distributed storage, and the access destination determination means 571 is software that performs its function. As a specific example, the access destination determination unit 571 calculates a data node that is actually stored from a global file path, or calculates a node to be accessed from a global block address.
- the information acquisition means 313 of the control instruction issuing means 551 of the control instruction node computer 55 sends information indicating what data is stored for each node to the other information management unit 21. Acquired from the access destination determination means 571 of the corresponding meta database node 57. When the data list for each node is acquired, future access amount information for each data is calculated from the characteristic model for each data. At this point, it is assumed that the characteristic model has already been estimated based on the access log for each data and each job.
- the control command issuing unit 551 (more specifically, the information acquiring unit 313) of the control command node computer 55 searches the characteristic record management DB 543 via the DB access processing unit 542 of the characteristic record DB node computer 54 for each data. And obtain a characteristic model.
- the control command issuing unit 551 (more specifically, the condition determining unit 314) calculates the future access amount (access amount of the latest ⁇ time) of each data using the acquired characteristic model.
- the future access amount may be calculated in advance.
- the future access amount information for each data is not calculated by acquiring the characteristic model, but is searched from the table calculated and held when the characteristic model is estimated, for example, as shown in FIG. Get it.
- FIG. 11 is an explanatory diagram showing another example of characteristic information stored in the characteristic record management DB 543.
- FIG. 11 shows an example in which access amount information at a future time is stored for each data.
- the access amount of the latest ⁇ time is calculated for each node, and it is determined whether or not the result is below a specified constant value. That's fine. As a result, if there is a node that is below a certain value, it is assumed that the condition is met. What is necessary is just to determine as a control strategy which performs the control processing content with the highest priority among the conditions finally matched.
- the rule “when there is a node whose access amount is less than a certain value in recent ⁇ time” is “predicting access before stopping the node, stopping the node whose access amount is less than a certain value”
- the control command issuing unit 551 executes the control processing content to be executed. Determine as a strategy and issue necessary control instructions.
- the control command issuing means 551 again calculates the access tendency (accessed presence / absence of access) from the characteristic model of the data to the future time for the data stored in the data computer node 51-1, If there is expected data, a control instruction to that effect is issued to the data moving means 516 in order to move the data to another node (other data computer nodes 51-2, 51-3, etc.). At that time, the control command issuing unit 551 issues a control command for changing the information to be stored and managed to the access destination determining unit 571 of the meta database node 57.
- control command issuing unit 551 issues a control command to the task allocation strategy control unit 562 in the job scheduler computer 56 so as not to assign a task to the data computer node 51-1.
- control command issuing means 551 issues a power stop command to the data node computer power control 517 of the data computer node 51-1.
- control rules such as “when there is data that will not be accessed for a certain time in the future” and “compress the relevant data”, and if it is the control processing content with the highest priority,
- the control command issuing unit 551 issues a control command for compressing the corresponding data as a data structure change command to the data structure control unit 515.
- control execution means classified into four types according to the control target are used.
- present invention is not limited to this, and there may be at least one type of desired control execution means.
- FIG. 12 is a block diagram showing an outline of a control system according to the present invention.
- the control system according to the present invention includes characteristic estimation means 701, one or more control execution means 702, control strategy determination means 703, and control command issue means 704.
- the characteristic estimation unit 701 (for example, the characteristic estimation process execution unit 12) is configured to execute a task or a task based on characteristic estimation source data related to a task operating on a computer system to be controlled or data held by the computer system. Estimate the characteristics of the data.
- the control execution means 702 (for example, the system component control execution means 41, the task arrangement strategy control execution means 42, the data arrangement strategy control execution means 43 or the data structure strategy control execution means 44) is executed by the computer according to the issued control command. Stop / operate control for system components, change task allocation, change data allocation, or change data structure.
- the control strategy determining unit 703 (for example, the control strategy determining / command issuing unit 31 and the condition determining unit 314) operates the computer system at a future time derived from the task or data characteristics estimated by the characteristic estimating unit 701. As a control strategy to be executed based on the situation, one or more of stop / operation control, task allocation change, data allocation change, and data structure change for the system components of the computer system to be controlled The content of the control process combining the above is determined.
- the control command issuing unit 704 (for example, the control strategy determining / command issuing unit 31 and the command issuing unit 311) issues a control command to the control execution unit 701 according to the control processing content determined by the control strategy determining unit 706.
- FIG. 13 is a block diagram showing another configuration example of the control system according to the present invention.
- the control system according to the present invention has four types of control: system component control execution means 721, task placement control execution means 722, data placement control execution means 723, and data structure control execution means 724.
- Execution means 702 may be provided.
- the control instruction issuing unit 704 according to the control processing content determined by the control strategy determining unit 706, task allocation control execution unit 702, task allocation control execution unit 703, data allocation control execution unit 704, A control command may be issued to the data structure control execution unit 705.
- the system component control execution unit 722 (for example, the system component control execution unit 41) performs stop / operation control on the system component of the computer system to be controlled.
- Task allocation strategy control execution unit 723 (for example, task allocation strategy control execution unit 42) changes the task allocation of the computer system to be controlled.
- the data arrangement strategy control execution means 724 (for example, the data arrangement strategy control execution means 43) changes the data arrangement of the computer system to be controlled.
- Data structure strategy control execution means 725 changes the data structure of the computer system to be controlled.
- the control system further includes external environment information acquisition means (for example, information acquisition means 313) that acquires external environment information that is information indicating the real-world situation from a predetermined information source, and the control strategy determination means includes characteristic estimation. Based on the condition of the computer system at a future time derived from the task or data characteristics estimated by the means and the real-world situation indicated by the external environment information acquired by the external environment information acquisition means The control strategy to be executed may be determined.
- external environment information acquisition means for example, information acquisition means 313
- control system includes a condition pair control process content storage unit (for example, process rule pair control process content storage unit 312) that associates the control process contents to be executed as the control strategy with the conditions for executing the control process contents.
- the control strategy determining means is indicated by the operating status of the computer system at a future time, derived from the task or data characteristics estimated by the characteristic estimating means, or external environment information acquired by the external environment information acquiring means
- the control strategy to be executed may be determined by determining whether the real-world situation or a combination thereof matches the condition stored in the condition-control processing content storage unit.
- the control system also includes classifying means for classifying and classifying data or tasks held by the computer system to be controlled according to the degree of similarity with other data or tasks, and estimating characteristics.
- the means may estimate characteristics for each class classified by the classifying means.
- the characteristic estimation means calculates a model likelihood based on the characteristic estimation source data for one or more arbitrary characteristic models, which are models representing the use of data or tasks or behavior, and the model likelihood is predetermined.
- the characteristics of the data or task may be estimated by selecting one characteristic model or a combination thereof that is greater than or equal to the threshold.
- one or more computer systems to be controlled perform stop / operation control, change task arrangement, change data arrangement, or change data structure for system components of the computer system according to the issued control command.
- the control execution means may be provided.
- the control command issuing means may issue a control command to the control execution means included in the computer system to be controlled according to the control processing content determined by the control strategy determining means.
- the present invention can be suitably applied to uses such as a distributed computing system, a distributed storage system, and a cloud computing back-end system.
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Debugging And Monitoring (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Description
・天気予報
・システムが動作しているデータセンタの電力情報
・システムを構成する計算機ノードの負荷情報
・システムを構成する計算機ノード上で動作する任意のサービスのログ情報
・WWW(ワールドワイドウェブ)上に存在する情報(例えば、Webページ、マイクロブログと呼ばれるようなサービス)
・TVの放送内容を示したサービス、TV放送の字幕情報
11 特性推定元データ取得・保持手段
12 特性推定処理実行手段
121 適合モデル発見手段
122 モデルセット記録部
123 モデル生成手段
124 クラス適合状態モデル発見手段
125 データ適合モデル発見手段
2 特性記録管理部
21 その他の情報管理部
3 制御戦略決定・命令発行手段
31 制御戦略決定・命令発行手段
311 命令発行手段
312 処理ルール対制御処理内容格納部
313 情報取得手段
314 条件判定手段
315 処理ルール対制御処理内容入出力手段
4 制御戦略実行手段群
41 システム構成要素制御実行手段
42 タスク配置戦略制御実行手段
43 データ配置戦略制御実行手段
44 データ構造戦略制御実行手段
701 特性推定手段
702 制御実行手段
703 制御戦略決定手段
704 制御命令発行手
721 システム構成要素制御実行手段
722 タスク配置戦略制御実行手段
723 データ配置戦略制御実行手段
724 データ構造戦略制御実行手段
Claims (10)
- 制御対象とされるコンピュータシステム上で動作するタスクまたは前記コンピュータシステムが保持しているデータに関係する特性推定元データを基に、前記タスクまたはデータの特性を推定する特性推定手段と、
発行された制御命令に従い、前記コンピュータシステムのシステム構成要素に対する停止・稼働制御、タスク配置の変更、データ配置の変更またはデータ構造の変更を行う1つ以上の制御実行手段と、
前記特性推定手段によって推定されたタスクまたはデータの特性から導出される、将来の時間における前記コンピュータシステムの稼働状況を基に、実行する制御戦略として、前記コンピュータシステムのシステム構成要素に対する停止・稼働制御、タスク配置の変更、データ配置の変更、データ構造の変更のうちのいずれか1つ以上を組み合わせた制御処理内容を決定する制御戦略決定手段と、
前記制御戦略決定手段が決定した制御処理内容に従い、前記制御実行手段に対して制御命令を発行する制御命令発行手段とを備える
ことを特徴とする制御システム。 - 制御対象とされるコンピュータシステムのシステム構成要素に対する停止・稼働制御を行うシステム構成要素制御実行手段、前記コンピュータシステムのタスク配置の変更を行うタスク配置制御実行手段、前記コンピュータシステムのデータ配置の変更を行うデータ配置制御実行手段、前記コンピュータシステムのデータ構造の変更を行うデータ構造制御実行手段、の4種の制御実行手段を備える
請求項1に記載の制御システム。 - 所定の情報源から実世界の状況を示す情報である外部環境情報を取得する外部環境情報取得手段を備え、
制御戦略決定手段は、特性推定手段によって推定されたタスクまたはデータの特性から導出される、将来の時間における前記コンピュータシステムの稼働状況と、前記外部環境情報取得手段によって取得された外部環境情報によって示される実世界の状況とを条件に用いて、実行する制御戦略を決定する
請求項1または請求項2に記載の制御システム。 - 制御戦略として実行する制御処理内容と、前記制御処理内容を実行する条件とを対応づけた条件対制御処理内容記憶手段を備え、
制御戦略決定手段は、特性推定手段によって推定されたタスクまたはデータの特性から導出される、将来の時間における前記コンピュータシステムの稼働状況、または前記外部環境情報取得手段によって取得された外部環境情報によって示される実世界の状況、もしくはその組み合わせが、前記条件対制御処理内容記憶手段に記憶されている条件に合致するか否かを判定することによって、実行する制御戦略を決定する
請求項1から請求項3のうちのいずれか1項に記載の制御システム。 - 制御対象とされるコンピュータシステムが保持しているデータまたはタスクを、他のデータまたはタスクとの間の類似度に応じて分類しクラス化するクラス化手段を備え、
特性推定手段は、前記クラス化手段によって分類された各クラスに対して、それぞれ特性を推定する
請求項1から請求項4のうちのいずれか1項に記載の制御システム。 - 特性推定手段は、データまたはタスクの利用または動作の傾向を表現したモデルである1つ以上の任意の特性モデルに対して特性推定元データを基にモデル尤度を算出し、モデル尤度が所定の閾値以上の1の特性モデルまたはその組み合わせを選択することによって、データまたはタスクの特性を推定する
請求項1から請求項5のうちのいずれか1項に記載の制御システム。 - 制御対象とされるコンピュータシステムが、
発行された制御命令に従い、当該コンピュータシステムのシステム構成要素に対する停止・稼働制御、タスク配置の変更、データ配置の変更またはデータ構造の変更を行う1つ以上の制御実行手段を備え、
制御命令発行手段は、制御対象とされるコンピュータシステムが備える前記制御実行手段に対して、制御戦略決定手段が決定した制御処理内容に従い制御命令を発行する
請求項1から請求項6のうちのいずれか1項に記載の制御システム。 - 当該コンピュータシステム上で動作するタスクまたは前記コンピュータシステムが保持しているデータに関係のある情報である特性推定元データを基に、前記タスクまたはデータの特性を推定する特性推定手段と、
発行された制御命令に従い、当該コンピュータシステムのシステム構成要素に対する停止・稼働制御、タスク配置の変更、データ配置の変更またはデータ構造の変更を行う1つ以上の制御実行手段と、
前記特性推定手段によって推定されたタスクまたはデータの特性から導出される、将来の時間における当該コンピュータシステムの稼働状況を基に、実行する制御戦略として、システム構成要素に対する停止・稼働制御、タスク配置の変更、データ配置の変更、データ構造の変更のうちのいずれか1つ以上を組み合わせた制御処理内容を決定する制御戦略決定手段と、
前記制御戦略決定手段が決定した制御処理内容に従い、前記制御実行手段に対して制御命令を発行する制御命令発行手段とを備える
ことを特徴とするコンピュータシステム。 - 制御対象とされるコンピュータシステム上で動作するタスクまたは前記コンピュータシステムが保持しているデータに関係のある情報である特性推定元データを基に、前記タスクまたはデータの特性を推定し、
推定されたタスクまたはデータの特性から導出される、将来の所定のまたは指定された時間における前記コンピュータシステムの稼働状況を基に、実行する制御戦略として、前記コンピュータシステムのシステム構成要素に対する停止・稼働制御、タスク配置の変更、データ配置の変更、データ構造の変更のうちのいずれか1つ以上を組み合わせた制御処理内容を決定し、
決定した制御処理内容に従い、前記コンピュータシステムのシステム構成要素に対する停止・稼働制御、タスク配置の変更、データ配置の変更またはデータ構造の変更を行う1つ以上の制御実行手段に対して、制御命令を発行する
ことを特徴とする制御方法。 - コンピュータに、
制御対象とされるコンピュータシステム上で動作するタスクまたは前記コンピュータシステムが保持しているデータに関係のある情報である特性推定元データを基に、前記タスクまたはデータの特性を推定する処理、
推定されたタスクまたはデータの特性から導出される、将来の時間における前記コンピュータシステムの稼働状況を基に、実行する制御戦略として、前記コンピュータシステムのシステム構成要素に対する停止・稼働制御、タスク配置の変更、データ配置の変更、データ構造の変更のうちのいずれか1つ以上を組み合わせた制御処理内容を決定する処理、および
決定した制御処理内容に従い、前記コンピュータシステムのシステム構成要素に対する停止・稼働制御、タスク配置の変更、データ配置の変更またはデータ構造の変更を行う1つ以上の制御実行手段に対して、制御命令を発行する処理
を実行させるための制御プログラム。
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US14/003,846 US20130346983A1 (en) | 2011-03-11 | 2012-03-07 | Computer system, control system, control method and control program |
JP2013504552A JP5692355B2 (ja) | 2011-03-11 | 2012-03-07 | コンピュータシステム、制御システム、制御方法および制御プログラム |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2011-054291 | 2011-03-11 | ||
JP2011054291 | 2011-03-11 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2012124295A1 true WO2012124295A1 (ja) | 2012-09-20 |
Family
ID=46830387
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2012/001563 WO2012124295A1 (ja) | 2011-03-11 | 2012-03-07 | コンピュータシステム、制御システム、制御方法および制御プログラム |
Country Status (3)
Country | Link |
---|---|
US (1) | US20130346983A1 (ja) |
JP (1) | JP5692355B2 (ja) |
WO (1) | WO2012124295A1 (ja) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP6318751B2 (ja) * | 2014-03-20 | 2018-05-09 | 富士通株式会社 | 情報処理装置、アクション切替方法、及びアクション切替プログラム |
JP6107801B2 (ja) * | 2014-12-12 | 2017-04-05 | 日本電気株式会社 | 情報処理装置、情報処理システム、タスク処理方法、及び、プログラム |
CN109725996B (zh) * | 2018-06-15 | 2023-07-21 | 中国平安人寿保险股份有限公司 | 任务模式选择与任务执行方法、装置、设备及存储介质 |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2006268684A (ja) * | 2005-03-25 | 2006-10-05 | Fujitsu Ltd | 需要予測装置、需要予測方法および需要予測プログラム |
JP2009169780A (ja) * | 2008-01-18 | 2009-07-30 | Hitachi Ltd | リソース状態予測装置、リソース状態予測方法、およびプログラム |
WO2010137455A1 (ja) * | 2009-05-29 | 2010-12-02 | インターナショナル・ビジネス・マシーンズ・コーポレーション | 計算機システム、方法、およびプログラム |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7346927B2 (en) * | 2002-12-12 | 2008-03-18 | Access Business Group International Llc | System and method for storing and accessing secure data |
US8185899B2 (en) * | 2007-03-07 | 2012-05-22 | International Business Machines Corporation | Prediction based priority scheduling |
US8326669B2 (en) * | 2007-04-19 | 2012-12-04 | International Business Machines Corporation | System and method for selecting and scheduling corrective actions for automated storage management |
US8352621B2 (en) * | 2009-12-17 | 2013-01-08 | International Business Machines Corporation | Method and system to automatically optimize execution of jobs when dispatching them over a network of computers |
US8850243B2 (en) * | 2011-03-02 | 2014-09-30 | Tso Logic Inc. | Non-intrusive power management |
-
2012
- 2012-03-07 US US14/003,846 patent/US20130346983A1/en not_active Abandoned
- 2012-03-07 JP JP2013504552A patent/JP5692355B2/ja active Active
- 2012-03-07 WO PCT/JP2012/001563 patent/WO2012124295A1/ja active Application Filing
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2006268684A (ja) * | 2005-03-25 | 2006-10-05 | Fujitsu Ltd | 需要予測装置、需要予測方法および需要予測プログラム |
JP2009169780A (ja) * | 2008-01-18 | 2009-07-30 | Hitachi Ltd | リソース状態予測装置、リソース状態予測方法、およびプログラム |
WO2010137455A1 (ja) * | 2009-05-29 | 2010-12-02 | インターナショナル・ビジネス・マシーンズ・コーポレーション | 計算機システム、方法、およびプログラム |
Also Published As
Publication number | Publication date |
---|---|
US20130346983A1 (en) | 2013-12-26 |
JPWO2012124295A1 (ja) | 2014-07-17 |
JP5692355B2 (ja) | 2015-04-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US9652161B2 (en) | System, method, and medium of optimizing load reallocation in an in-memory data management grid | |
US20200034292A1 (en) | Adaptable data caching mechanism for in-memory cluster computing | |
Rao et al. | Performance issues of heterogeneous hadoop clusters in cloud computing | |
CN103067425A (zh) | 虚拟机创建方法、虚拟机管理系统及相关设备 | |
US20160011980A1 (en) | Distributed processing method and system | |
JP2012118987A (ja) | メモリ使用量照会ガバナのためのコンピュータ実装方法、コンピュータ・プログラム、およびシステム(メモリ使用量照会ガバナ) | |
Costa et al. | Towards automating the configuration of a distributed storage system | |
US10298709B1 (en) | Performance of Hadoop distributed file system operations in a non-native operating system | |
Zhang et al. | Aurora: Adaptive block replication in distributed file systems | |
US10579419B2 (en) | Data analysis in storage system | |
JP5692355B2 (ja) | コンピュータシステム、制御システム、制御方法および制御プログラム | |
WO2016122595A1 (en) | Chunk monitoring | |
JP5810918B2 (ja) | スケジューリング装置、スケジューリング方法及びプログラム | |
Salehian et al. | Comparison of spark resource managers and distributed file systems | |
US8732394B2 (en) | Advanced disk drive power management based on maximum system throughput | |
Chen et al. | Data prefetching and eviction mechanisms of in-memory storage systems based on scheduling for big data processing | |
CN104765572A (zh) | 一种节能的虚拟存储服务器系统及其调度方法 | |
EP3555767B1 (en) | Partial storage of large files in distinct storage systems | |
US11169720B1 (en) | System and method for creating on-demand virtual filesystem having virtual burst buffers created on the fly | |
Wang et al. | An adaptive elasticity policy for staging based in-situ processing | |
Wang et al. | NBUFlow: a dataflow based universal task orchestration and offloading platform for low-cost development of IoT systems with cloud-edge-device collaborative computing | |
Zhao et al. | A holistic cross-layer optimization approach for mitigating stragglers in in-memory data processing | |
US10168763B2 (en) | Modification of when workloads access data units and/or on which storage devices data units are stored to conserve power | |
Lee et al. | Adaptive Data Replication Scheme Based on Access Count Prediction in Hadoop | |
Yuan et al. | Research of scheduling strategy based on fault tolerance in Hadoop platform |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 12757291 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2013504552 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 14003846 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 12757291 Country of ref document: EP Kind code of ref document: A1 |