WO2014068980A1 - 分散データ処理システム、及び、分散データ処理方法 - Google Patents
分散データ処理システム、及び、分散データ処理方法 Download PDFInfo
- Publication number
- WO2014068980A1 WO2014068980A1 PCT/JP2013/006437 JP2013006437W WO2014068980A1 WO 2014068980 A1 WO2014068980 A1 WO 2014068980A1 JP 2013006437 W JP2013006437 W JP 2013006437W WO 2014068980 A1 WO2014068980 A1 WO 2014068980A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- processing
- data
- types
- pair
- predetermined arithmetic
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/10—File systems; File servers
- G06F16/18—File system types
- G06F16/182—Distributed file systems
-
- 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]
Definitions
- the present invention relates to a distributed data processing system and a distributed data processing method, and more particularly to a distributed data processing system and a distributed data processing method for processing each pair of a plurality of types of data.
- Patent Document 1 describes an example of an operation management system that models a system using time series information of system performance and detects a failure of the system using a generated model.
- the operation management system described in Patent Document 1 generates a system correlation model by determining a correlation function for each pair of a plurality of metrics based on measurement values of a plurality of metrics of the system. Then, the operation management system detects the destruction of the correlation (correlation destruction) using the generated correlation model, and determines the cause of the failure based on the correlation destruction. In this way, the technique of analyzing a failure factor based on correlation destruction is called invariant relation analysis.
- correlation functions are calculated for all pairs of multiple metrics.
- the number of pairs for which the correlation function is calculated is proportional to the square of the number of metrics. For this reason, when the scale of the system (the number of metrics) is large, the number of pairs for calculating the correlation function becomes enormous, and it becomes difficult to generate a correlation model within a predetermined time.
- Non-Patent Document 1 Hadoop Distributed File System
- MapReduce MapReduce
- Patent Document 2 discloses a method of determining a node to execute processing based on a communication delay between nodes in a distributed processing system such as Hadoop.
- each pair of a plurality of metrics is assigned to any node, and each node calculates a correlation function for the assigned pair.
- each node needs to acquire metric data related to the assigned pair from the node where the data is arranged, and data transfer between the nodes occurs frequently.
- Hadoop there is a problem that distributed processing for each pair of multiple types of data cannot be executed efficiently.
- An object of the present invention is to provide a distributed data processing system and a distributed data processing method capable of solving the above-described problems and efficiently executing distributed processing for each pair of a plurality of types of data.
- a distributed data processing system is a distributed data processing system that performs predetermined arithmetic processing on each pair of N types (N is a natural number of 2 or more) of data, and the N types Management means for allocating each of the first to N ⁇ 1th of the plurality of processing means to any of the plurality of processing means, and each of the i th (i is 1 or more and N ⁇ 1 or less) assigned by the management means
- a plurality of processing means for executing the predetermined arithmetic processing on a pair of i + 1 to Nth data.
- a management apparatus is a management apparatus in a distributed data processing system that performs a predetermined arithmetic process on each pair of N types of data (N is a natural number of 2 or more).
- N is a natural number of 2 or more.
- Each of the first to N ⁇ 1 of the seeds is assigned to the i-th data (i is a natural number greater than or equal to 1 and less than or equal to N ⁇ 1) assigned by the management device and each of the i + 1 to Nth data.
- Management means for assigning to any one of a plurality of processing devices that execute the predetermined arithmetic processing for the pair.
- a processing apparatus is a processing apparatus in a distributed data processing system that performs predetermined arithmetic processing on each pair of N types (N is a natural number of 2 or more) of data, I-th data (i is a natural number greater than or equal to 1 and less than or equal to N-1) assigned by a management device that assigns each of the first to N ⁇ 1 of the seeds to any of the plurality of processing devices, and i + 1 To N-th data pair, the processing means for executing the predetermined arithmetic processing.
- N is a natural number of 2 or more
- I-th data i is a natural number greater than or equal to 1 and less than or equal to N-1 assigned by a management device that assigns each of the first to N ⁇ 1 of the seeds to any of the plurality of processing devices, and i + 1 To N-th data pair, the processing means for executing the predetermined arithmetic processing.
- a distributed data processing method is a distributed data processing method for performing predetermined arithmetic processing on each pair of N types (N is a natural number of 2 or more) of data, and , Each of the first to N ⁇ 1 of the N types is assigned to any of a plurality of processing means, and in each of the plurality of processing means, the i-th (i is The predetermined calculation process is performed on a pair of data of a natural number of 1 or more and N ⁇ 1 or less) and each of i + 1 to Nth data.
- the first computer-readable recording medium is a distributed data processing system that performs predetermined arithmetic processing on each pair of N types (N is a natural number of 2 or more) of data.
- a program for executing processing assigned to any of the plurality of processing devices that execute the predetermined arithmetic processing is stored for a pair of data of a natural number of 1 or less and each of the i + 1 to Nth data.
- the second computer-readable recording medium is a distributed data processing system that performs predetermined arithmetic processing on each pair of N types of data (N is a natural number of 2 or more).
- i stores a program for executing the predetermined calculation process for a pair of data of a natural number of 1 or more and N ⁇ 1 or less) and each of the i + 1 to Nth data.
- the effect of the present invention is that the distributed processing for each pair of a plurality of data can be executed efficiently.
- FIG. 2 is a block diagram showing the configuration of the operation management system 500 in the first embodiment of the present invention.
- the operation management system 500 generates a correlation model of the analysis target system 600 based on the performance information collected from the analysis target system 600, and detects a failure or an abnormality of the analysis target system 600 using the generated correlation model. Do.
- the analysis target system 600 includes one or more monitored devices that execute service processing such as a WEB server, an application server (AP server), and a database server (DB server).
- the monitored device measures actual measurement data (measurement values) of a plurality of types of performance values at regular intervals (predetermined performance information collection cycle), and transmits them to the operation management system 500.
- the performance value item for example, the usage rate and usage of computer resources such as CPU (Central Processing Unit) usage rate, memory usage rate, disk access frequency, network usage rate, and the like are used.
- CPU Central Processing Unit
- a set of monitored devices and performance value items is a metric (performance type or simply a type), and a set of metric values of N types (N is a natural number of 2 ⁇ N) measured at the same time. Is performance information.
- the metric corresponds to an element in Patent Document 1.
- the operation management system 500 includes a distributed data processing system (correlation model generation system) 100, an information collection device 200, a correlation destruction detection device 300, and a failure analysis device 400.
- the information collection device 200 collects performance information from the monitored device of the analysis target system 600 at a predetermined performance information collection cycle, and transmits it to the management device 110 of the distributed data processing system 100.
- the distributed data processing system 100 generates a correlation model of the analysis target system 600 based on the performance information.
- the correlation model includes a correlation function for each pair of N metrics.
- the correlation function is a function that predicts the time series of the value of the other metric from the time series of the value of one metric in the metric pair, and indicates the correlation of the metric pair.
- the distributed data processing system 100 includes a management device 110, a plurality of processing devices 120 (120_1, 120_2, etc, And a processing result collection device 130.
- the management device 110 includes a management unit 111 (or master) and a data storage unit 112.
- the data storage unit 112 stores the time series of the performance information received from the information collection device 200 as the processing data 113.
- FIG. 5 is a diagram illustrating an example of the processing data 113 according to the first embodiment of this invention.
- the performance information includes measured values (data) of N types of metrics (m 1 , m 2 ,..., M N ).
- the management unit 111 assigns each of the first to N ⁇ 1th of the N types of metrics to the processing device 120.
- the processing device 120 includes a processing unit 121 (or a worker) and a temporary storage unit 122.
- the processing unit 121 calculates a correlation function for a pair of a metric assigned by the management apparatus 110 and another metric.
- the processing unit 121 determines that the i-th metric (m i ) and the (i + 1) to N-th metric (m j , i + 1 ⁇ j ⁇ N).
- the correlation function is calculated for each pair.
- the processing unit 121 acquires the measurement value of each metric pair from the management apparatus 110, and determines the coefficient of the correlation function by performing system identification processing in the same manner as the operation management apparatus of Patent Document 1.
- the processing unit 121 stores the metric measurement value acquired from the management apparatus 110 in the temporary storage unit 122.
- the temporary storage unit 122 temporarily stores (caches) the metric data acquired from the data storage unit 112.
- FIG. 6 is a diagram illustrating an example of data stored in the temporary storage unit 122 according to the first embodiment of this invention.
- the temporary storage unit 122 includes a temporary storage unit 122a (first temporary storage unit) that stores data of the i-th metric (m i ), and i + 1 to N-th metrics (m j , It may be divided into a temporary storage unit 122b (second temporary storage unit) that stores data of (i + 1 ⁇ j ⁇ N).
- the temporary storage unit 122b may store data of a predetermined number of metrics from the i + 1th to Nth metrics (m j , i + 1 ⁇ j ⁇ N).
- the temporary storage unit 122b may store the metric data in, for example, a FIFO (First-In First-Out) format. Further, in this case, the temporary storage unit 122b may store the metric data in a format other than the FIFO, such as a format in which as many metrics with large i as possible remain.
- a FIFO First-In First-Out
- the temporary storage unit 122b may store the metric data in a format other than the FIFO, such as a format in which as many metrics with large i as possible remain.
- the processing result collection device 130 includes a processing result collection unit 131 and a processing result storage unit 132.
- the processing result collection unit 131 acquires the correlation function calculated in each processing device 120 and stores it in the processing result storage unit 132 as the processing result 133.
- the processing result storage unit 132 stores the processing result 133.
- the processing result 133 indicates a correlation model of the analysis target system 600.
- FIG. 7 is a diagram illustrating an example of the processing result 133 according to the first embodiment of this invention.
- f (i, j) indicates a correlation function for pairs of input metric m i and the output metric m j.
- the coefficients ⁇ and ⁇ are determined for the pair of mi and m j .
- another function expression may be used as the correlation function.
- Correlation destruction detection apparatus 300 detects the correlation destruction of the correlation included in the correlation model using newly input performance information, as in Patent Document 1.
- the failure analysis apparatus 400 estimates the failure factor based on the detection result of the correlation destruction, as in Patent Document 1.
- the management device 110, the processing device 120, the processing result collection device 130, the information collection device 200, the correlation destruction detection device 300, and the failure analysis device 400 each include a CPU and a storage medium that stores the program. It may be a computer that operates based on the control.
- the management device 110, the processing device 120, the processing result collection device 130, the information collection device 200, the correlation destruction detection device 300, and the failure analysis device 400 are each a virtual computer (virtual machine) constructed on a computer. ) Also, some of the management device 110, the processing device 120, the processing result collection device 130, the information collection device 200, the correlation destruction detection device 300, and the failure analysis device 400 may constitute one device. .
- FIG. 3 is a flowchart showing the processing of the management unit 111 in the first embodiment of the present invention.
- the processing data 113 as shown in FIG. 5 relating to the N types of metrics is stored in the data storage unit 112 of the management device 110. Assume.
- the management unit 111 of the management apparatus 110 initializes the metric number to 1 (step S101).
- the management unit 111 waits for a request from the processing unit 121 of each processing device 120 (step S102).
- the management unit 111 transmits the metric number to the processing unit 121 (step S104).
- the management unit 111 adds 1 to the metric number (step S105).
- the management unit 111 acquires the metric data of the requested metric number from the data storage unit 112, and transmits the metric data to the processing unit 121. (Step S106).
- the management unit 111 repeatedly executes the processes of steps S102 to S106.
- FIG. 4 is a flowchart showing the processing of the processing unit 121 in the first embodiment of the present invention.
- the processing unit 121 of each processing device 120 requests the metric number i from the management unit 111 (step S201), and acquires the metric number i.
- Processing unit 121 data of the metric m i is determined whether the temporary storage section 122b (step S202).
- the processing unit 121 acquires the data of the metric m i from the temporary storage section 122b, it is stored in the temporary storage unit 122a (step S203).
- step S202 If not in the temporary storage unit 122b (step S202 / N), the processing unit 121, the management unit 111, requesting the data of the metric m i (step S204), and acquires the data of the metric m i.
- Processing unit 121 stores the data of the metric m i in the temporary storage unit 122a (step S205).
- the processing unit 121 initializes the metric number j to i + 1 (step S206).
- the processing unit 121 determines whether or not the data of the metric m j is in the temporary storage unit 122b (Step S207).
- the processing unit 121 When not in the temporary storage unit 122b (step S207 / N), the processing unit 121 requests the data of the metric m j from the management unit 111 (step S208), and acquires the data of the metric m j .
- the processing unit 121 stores the data of the metric m j in the temporary storage unit 122b (Step S209).
- the metric data is stored in the temporary storage unit 122b, for example, in the FIFO format.
- Processing unit 121 to the pair of metrics m i and m j, and calculates the correlation function f (i, j) (step S210).
- the processing unit 121 determines that the correlation function f (j, j, i) is also calculated.
- the processing unit 121 transmits the calculated correlation function to the processing result collection unit 131 of the processing result collection device 130 (step S211).
- the processing result collection unit 131 sets the correlation function acquired from the processing device 120 in the processing result 133 of the processing result storage unit 132.
- the processing unit 121 adds 1 to the metric number j (step S212). If j is N or less (step S213 / Y), the processing of steps S207 to 212 is repeated (step S213).
- processing unit 121 repeats the processing of steps S201 to S213.
- the management unit 111 sequentially assigns metric numbers from 1 to N-1 to the processing unit 121.
- the processing unit 121_1 acquires the data of the metric m 1 from the management unit 111 and stores it in the temporary storage unit 122a.
- the processing unit 121_1 acquires the data of the metric m 2 from the management unit 111 and stores it in the temporary storage unit 122b.
- the processing unit 121_1 calculates a correlation function f (1,2) and a correlation function f (2,1) for the metric pair (m 1 , m 2 ).
- the processing unit 121_1, from the management unit 111 acquires data of the metric m 3, is stored in the temporary storage unit 122b.
- the processing unit 121_1 calculates the correlation function f (1, 3) and the correlation function f (3, 1) for the metric pair (m 1 , m 3 ). In this way, the processing unit 121_1 performs metric pairs (m 1 , m 2 ), (m 1 , m 3 ), (m 1 , m 4 ), ..., (m 1 , m N ) Correlation functions f (1,2), f (1,3), f (1,4), ..., f (1, N), and correlation functions f (2,1), f (3,1), f (4,1),..., f (N, 1) (correlation function group 134_1 and correlation function group 134_2 in the processing result 133 of FIG. 7) are calculated.
- Processing unit 121_2, from the management unit 111 acquires data of the metric m 2, is stored in the temporary storage unit 122a.
- the processing unit 121_2, from the management unit 111 acquires data of the metric m 3, is stored in the temporary storage unit 122b.
- the processing unit 121_2 calculates the correlation function f (2, 3) and the correlation function f (3, 2) for the metric pair (m 2 , m 3 ).
- the processing unit 121_2 performs the correlation function f (2, 3) on the metric pairs (m 2 , m 3 ), (m 2 , m 4 ),..., ( M 2 , m N ). , F (2,4),..., F (2, N), and correlation functions f (3,2), f (4,2),..., F (N, 2) (processing result 133 in FIG. 7) Correlation function group 135_1 and correlation function group 135_2) are calculated.
- the processing unit 121_1 acquires a metric number k (k is i ⁇ k ⁇ N ⁇ 1 acquired last time) from the management unit 111.
- the processing unit 121_1 performs correlation functions f (k, k + 1), f (k, k) on the metric pairs (m k , m k + 1 ), (m k , m k + 2 ),..., ( M k , m N ).
- k + 2),..., f (k, N) and correlation functions f (k + 1, k), f (k + 2, k),..., f (N, k) are calculated.
- the processing unit 121_1 acquires the data of the metric m k from the temporary storage unit 122b, saves it in the temporary storage unit 122a, and uses it. Further, when there is data of metrics m k + 1 to m N in the temporary storage unit 122b, the processing unit 121_1 uses those data.
- the correlation function is calculated for all pairs of N types of metrics by the plurality of processing units 121.
- the processing result 133 (correlation model) as shown in FIG. Stored in the unit 132.
- FIG. 1 is a block diagram showing a characteristic configuration of the first embodiment of the present invention.
- the distributed data processing system 100 performs predetermined arithmetic processing on each pair of N types of data (N is a natural number of 2 or more).
- the distributed data processing system 100 includes a management unit 111 and a plurality of processing units 121.
- the management unit 111 assigns each of the first to N ⁇ 1 of the N types to any of the plurality of processing units 121.
- Each of the plurality of processing units 121 is predetermined for a pair of i-th (i is a natural number greater than or equal to 1 and less than or equal to N ⁇ 1) data and i + 1 to Nth data assigned by the management unit 111.
- the calculation process is executed.
- the management unit 111 assigns each of the first to N ⁇ 1 of the N types to any one of the plurality of processing units 121, and each of the plurality of processing units 121 assigns them by the management unit 111. This is because a predetermined calculation process is performed on a pair of the i-th data and each of the i + 1 to Nth data.
- the management unit 111 sequentially assigns each of the first to N ⁇ 1 of the N types, and the processing unit 121 stores the i + 1 to Nth data in the temporary storage unit 122b, whereby the processing unit In 121, the k-th data (k is i ⁇ k ⁇ N ⁇ 1) acquired last time and the N-th data from k + 1 are stored in the temporary storage unit 122b, so that the number of times of data transfer is further increased. Reduced.
- the load related to I / O (Input / Output) of the management device 110, each processing device 120, and the processing result collection device 130 is reduced by reducing the number of times of data transfer. .
- the processing unit 121 is dynamically added so that the calculation of correlation functions for all pairs of N metrics can be completed within the processing completion time.
- FIG. 8 is a block diagram showing the configuration of the operation management system 500 according to the second embodiment of the present invention.
- the distributed data processing system 100 includes an operating processing device 120 (120_1, 120_2, etc And a stopped processing device 120 (120_4, 120_5,).
- the processing apparatus 120 in operation calculates the correlation function by performing the processes of steps S201 to S213 (FIG. 4).
- the management device 110 further includes a processing device control unit 114 (or control unit) and an operation state storage unit 115.
- the operation state storage unit 115 stores operation state information 116 indicating the operation state of the processing device 120.
- FIG. 10 is a diagram illustrating an example of the operation state information 116 in the second exemplary embodiment of the present invention.
- the operation state information 116 includes an identifier of the processing device 120 and an operation state (operating or stopped) of the processing device 120.
- the processing device control unit 114 calculates a prediction processing time necessary for calculating correlation functions for all pairs of N types of metrics by the active processing device 120, and based on the prediction processing time, the processing device control unit 114 is stopped. The processing device 120 is activated (the processing device 120 is added).
- FIG. 9 is a flowchart showing the processing of the processing device control unit 114 in the second embodiment of the present invention.
- the processing devices 120_1 to 3 are operating, the processing devices 120_4 to 6 are stopped, and the operating state information 116 as shown in FIG. Further, it is assumed that the correlation function is calculated by the processing apparatuses 120_1 to 3 in operation.
- the processing device control unit 114 of the management device 110 performs processing for all pairs of N types of metrics after a predetermined elapsed time from the request for the first metric number from the processing unit 121 (start of calculation of the correlation function by the processing unit 121).
- a prediction processing time is calculated (step S301).
- the processing device control unit 114 calculates the predicted processing time for all pairs based on the elapsed time and the number of pairs for which the correlation function has been calculated in the processing result 133.
- the processing device control unit 114 When the calculated predicted processing time exceeds the processing completion time (step S302 / Y), the processing device control unit 114 refers to the operating state information 116 and is necessary for calculating the correlation function for all pairs within the processing completion time.
- the number of processing devices 120 is calculated (step S303).
- the processing completion time is set in advance by an administrator or the like based on the time when the correlation model changes.
- the processing device control unit 114 Based on the above-described elapsed time, the number of pairs for which correlation functions have been calculated, and the number of active processing devices 120 acquired from the operating state information 116, the processing device control unit 114 performs all processing within the processing completion time.
- the number of processing devices 120 necessary to calculate the correlation function for the pair is calculated.
- the processing device control unit 114 operates the stopped processing device 120 so that the number of operating processing devices 120 is the calculated number (step S304).
- the processing device control unit 114 updates the operation state related to the activated processing device 120 in the operation state information 116.
- the newly operating processing device 120 performs the processing of steps S201 to S213, thereby calculating a correlation function.
- the processing device control unit 114 sets the predicted processing time to 3 / Calculate as 2 ⁇ T. Since the processing device control unit 114 needs to calculate 2/3 correlation functions of all the pairs for the remaining 1 ⁇ 2 ⁇ T of the predetermined processing completion time, the processing device control unit 114 determines the number of processing devices 120 required. It is calculated as 6 which is 2 times. The processing device control unit 114 operates the stopped processing devices 120_4 to 6-6. Then, in addition to the processing devices 120_1 to 3, the processing devices 120_4 to 6 calculate a correlation function.
- the processing device control unit 114 may present the calculated predicted processing time to an administrator or the like, and operate the stopped processing device 120 in accordance with an instruction from the administrator or the like.
- processing device control unit 114 may calculate the number of necessary processing devices 120 based on the load status of each processing device 120 and operate the processing devices 120 that are stopped.
- the processing device control unit 114 may add the processing device 120 by deploying a new virtual machine on the computer.
- the calculation of correlation functions for all pairs of N types of metrics can be completed within the processing completion time.
- the reason is that when the processing device control unit 114 calculates the number of processing devices 120 necessary to calculate the correlation function for all pairs within the processing completion time when the predicted processing time is larger than the processing completion time, This is because the addition of the device 120 accelerates the correlation function calculation process.
- the correlation function calculation process can be easily accelerated.
- the reason is that the processing of the management device 110 and each processing device 120 does not depend on the number of processing devices 120, and the processing devices 120 can be easily added.
- the case where the correlation function is calculated for each pair of the plurality of types of data by the plurality of processing devices 120 as the distributed processing has been described as an example. As long as the process is performed on the pair, a calculation process other than the calculation of the correlation function may be performed.
- one management device 110 and one processing result collection device 130 are provided, but a plurality of management devices 110 and a plurality of processing result collection devices 130 may be used.
- allocation of metrics to the processing device 120 and transmission of metric data are executed in a distributed manner by the plurality of management devices 110.
- the collection of correlation functions from the processing device 120 is also executed in a distributed manner by the plurality of processing result collection devices 130.
- DESCRIPTION OF SYMBOLS 100 Distributed data processing system 110 Management apparatus 111 Management part 112 Data storage part 113 Processing data 114 Processing apparatus control part 115 Operation state storage part 116 Operation state information 120 Processing apparatus 121 Processing part 122 Temporary storage part 130 Processing result collection apparatus 131 Processing result Collection unit 132 Processing result storage unit 133 Processing result 200 Information collection device 300 Correlation destruction detection device 400 Failure analysis device 500 Operation management system 600 Analysis target system
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Debugging And Monitoring (AREA)
- Computer And Data Communications (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Description
次に、本発明の第1の実施の形態について説明する。
次に、本発明の第2の実施の形態について説明する。
110 管理装置
111 管理部
112 データ記憶部
113 処理データ
114 処理装置制御部
115 稼動状態記憶部
116 稼動状態情報
120 処理装置
121 処理部
122 一時記憶部
130 処理結果収集装置
131 処理結果収集部
132 処理結果記憶部
133 処理結果
200 情報収集装置
300 相関破壊検出装置
400 障害分析装置
500 運用管理システム
600 分析対象システム
Claims (20)
- N種(Nは、2以上の自然数)のデータの内の各ペアに対して所定の演算処理を行う分散データ処理システムであって、
前記N種の内の1番目からN-1番目の各々を、複数の処理手段のいずれかに割り当てる管理手段と、
各々が、前記管理手段により割り当てられたi番目(iは、1以上かつN-1以下の自然数)のデータとi+1からN番目の各々のデータとのペアに対して前記所定の演算処理を実行する前記複数の処理手段と、
を備える分散データ処理システム。 - さらに、前記N種のデータを記憶するデータ記憶手段を備え、
前記複数の処理手段の各々は、前記データ記憶手段から前記i番目のデータと前記i+1からN番目のデータとを取得し、前記所定の演算処理を実行する、
請求項1に記載の分散データ処理システム。 - さらに、前記複数の処理手段の各々に、第1の一時記憶手段を備え、
前記複数の処理手段の各々は、前記データ記憶手段から取得した前記i番目のデータを前記第1の一時記憶手段に記憶させ、前記第1の一時記憶手段に記憶されたデータを用いて、前記i番目のデータとi+1からN番目の各々のデータとのペアに対して前記所定の演算処理を実行する、
請求項1または2に記載の分散データ処理システム。 - さらに、前記複数の処理手段の各々に、第2の一時記憶手段を備え、
前記複数の処理手段の各々は、前記データ記憶手段から取得した前記i+1からN番目のデータの少なくとも一部を前記第2の一時記憶手段に記憶させ、前記管理手段により新たにk番目(kは、1以上かつN-1以下の自然数)が割り当てられた場合に、前記第2の一時記憶手段に記憶されたデータを用いて、前記k番目のデータとk+1からN番目の各々のデータとのペアに対して前記所定の演算処理を実行する、
請求項1乃至3のいずれかに記載の分散データ処理システム。 - 前記管理手段は、前記N種の内の1番目からN-1番目の各々を順番に、前記複数の処理手段のいずれかに割り当てる、
請求項1乃至4のいずれかに記載の分散データ処理システム。 - さらに、前記複数の処理手段による前記所定の演算処理が実行されているときに、新たな前記処理手段を追加する制御手段を備える、
請求項1乃至5のいずれかに記載の分散データ処理システム。 - 前記制御手段は、前記複数の処理手段による前記N種のデータの内の全ペアに対する前記所定の演算処理の予測処理時間が所定の処理完了時間を越える場合、前記全ペアに対する前記所定の演算処理が前記所定の処理完了時間以内で完了するために必要な前記処理手段の数を算出し、前記処理手段の数が当該算出した数になるように、前記処理手段を追加する、
請求項6に記載の分散データ処理システム。 - 前記N種のデータは、システムにおけるN種のメトリックの計測値であり、
前記複数の処理手段の各々は、前記管理手段により割り当てられたi番目のメトリックの計測値とi+1からN番目の各々のメトリックの計測値とのペアに対して相関関数を算出する、
請求項1乃至7のいずれかに記載の分散データ処理システム。 - N種(Nは、2以上の自然数)のデータの内の各ペアに対して所定の演算処理を行う分散データ処理システムにおける管理装置であって、
前記N種の内の1番目からN-1番目の各々を、管理装置により割り当てられたi番目(iは、1以上かつN-1以下の自然数)のデータとi+1からN番目の各々のデータとのペアに対して前記所定の演算処理を実行する複数の処理装置のいずれかに割り当てる管理手段
を備えた管理装置。 - N種(Nは、2以上の自然数)のデータの内の各ペアに対して所定の演算処理を行う分散データ処理システムにおける処理装置であって、
前記N種の内の1番目からN-1番目の各々を複数の処理装置のいずれかに割り当てる管理装置により割り当てられたi番目(iは、1以上かつN-1以下の自然数)のデータと、i+1からN番目の各々のデータとのペアに対して前記所定の演算処理を実行する処理手段
を備えた処理装置。 - N種(Nは、2以上の自然数)のデータの内の各ペアに対して所定の演算処理を行う分散データ処理方法であって、
管理手段において、前記N種の内の1番目からN-1番目の各々を、複数の処理手段のいずれかに割り当て、
前記複数の処理手段の各々において、前記管理手段により割り当てられたi番目(iは、1以上かつN-1以下の自然数)のデータとi+1からN番目の各々のデータとのペアに対して前記所定の演算処理を実行する、
分散データ処理方法。 - 前記複数の処理手段の各々において、前記N種のデータを記憶するデータ記憶手段から、前記i番目のデータと前記i+1からN番目のデータとを取得し、前記所定の演算処理を実行する、
請求項11に記載の分散データ処理方法。 - 前記複数の処理手段の各々において、前記データ記憶手段から取得した前記i番目のデータを、前記複数の処理手段の各々に対応する第1の一時記憶手段に記憶させ、前記第1の一時記憶手段に記憶されたデータを用いて、前記i番目のデータとi+1からN番目の各々のデータとのペアに対して前記所定の演算処理を実行する、
請求項11または12に記載の分散データ処理方法。 - 前記複数の処理手段の各々において、前記データ記憶手段から取得した前記i+1からN番目のデータの少なくとも一部を、前記複数の処理手段の各々に対応する第2の一時記憶手段に記憶させ、前記管理手段により新たにk番目(kは、1以上かつN-1以下の自然数)が割り当てられた場合に、前記第2の一時記憶手段に記憶されたデータを用いて、前記k番目のデータとk+1からN番目の各々のデータとのペアに対して前記所定の演算処理を実行する、
請求項11乃至13のいずれかに記載の分散データ処理方法。 - 前記管理手段において、前記N種の内の1番目からN-1番目の各々を順番に、前記複数の処理手段のいずれかに割り当てる、
請求項11乃至14のいずれかに記載の分散データ処理方法。 - さらに、制御手段において、前記複数の処理手段による前記所定の演算処理が実行されているときに、新たな前記処理手段を追加する、
請求項11乃至15のいずれかに記載の分散データ処理方法。 - 前記制御手段において、前記複数の処理手段による前記N種のデータの内の全ペアに対する前記所定の演算処理の予測処理時間が所定の処理完了時間を越える場合、前記全ペアに対する前記所定の演算処理が前記所定の処理完了時間以内で完了するために必要な前記処理手段の数を算出し、前記処理手段の数が当該算出した数になるように、前記処理手段を追加する、
請求項16に記載の分散データ処理方法。 - 前記N種のデータは、システムにおけるN種のメトリックの計測値であり、
前記複数の処理手段の各々において、前記管理手段により割り当てられたi番目のメトリックの計測値とi+1からN番目の各々のメトリックの計測値とのペアに対して相関関数を算出する、
請求項11乃至17のいずれかに記載の分散データ処理方法。 - N種(Nは、2以上の自然数)のデータの内の各ペアに対して所定の演算処理を行う分散データ処理システムにおける管理装置のプログラムを格納する記録媒体であって、
コンピュータに、
前記N種の内の1番目からN-1番目の各々を、管理装置により割り当てられたi番目(iは、1以上かつN-1以下の自然数)のデータとi+1からN番目の各々のデータとのペアに対して前記所定の演算処理を実行する複数の処理装置のいずれかに割り当てる、
処理を実行させるプログラムを格納する、コンピュータが読み取り可能な記録媒体。 - N種(Nは、2以上の自然数)のデータの内の各ペアに対して所定の演算処理を行う分散データ処理システムにおける処理装置のプログラムを格納する記録媒体であって、
コンピュータに、
前記N種の内の1番目からN-1番目の各々を複数の処理装置のいずれかに割り当てる管理装置により割り当てられたi番目(iは、1以上かつN-1以下の自然数)のデータと、i+1からN番目の各々のデータとのペアに対して前記所定の演算処理を実行する、
処理を実行させるプログラムを格納する、コンピュータが読み取り可能な記録媒体。
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP13851403.9A EP2916222A4 (en) | 2012-11-01 | 2013-10-30 | DISTRIBUTED DATA PROCESSING SYSTEM AND DISTRIBUTED DATA PROCESSING METHOD |
US14/439,082 US10296493B2 (en) | 2012-11-01 | 2013-10-30 | Distributed data processing system and distributed data processing method |
CN201380057593.9A CN104769551B (zh) | 2012-11-01 | 2013-10-30 | 分布式数据处理系统和分布式数据处理方法 |
JP2014544322A JP6070717B2 (ja) | 2012-11-01 | 2013-10-30 | 分散データ処理システム、及び、分散データ処理方法 |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2012241993 | 2012-11-01 | ||
JP2012-241993 | 2012-11-01 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2014068980A1 true WO2014068980A1 (ja) | 2014-05-08 |
Family
ID=50626928
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2013/006437 WO2014068980A1 (ja) | 2012-11-01 | 2013-10-30 | 分散データ処理システム、及び、分散データ処理方法 |
Country Status (5)
Country | Link |
---|---|
US (1) | US10296493B2 (ja) |
EP (1) | EP2916222A4 (ja) |
JP (1) | JP6070717B2 (ja) |
CN (1) | CN104769551B (ja) |
WO (1) | WO2014068980A1 (ja) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20190253357A1 (en) * | 2018-10-15 | 2019-08-15 | Intel Corporation | Load balancing based on packet processing loads |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH07325806A (ja) * | 1994-06-01 | 1995-12-12 | Nec Corp | マルチプロセッサシステムのプロセス割当方法 |
JP2006146668A (ja) * | 2004-11-22 | 2006-06-08 | Ntt Data Corp | 運用管理支援装置及び運用管理支援プログラム |
JP2009199533A (ja) * | 2008-02-25 | 2009-09-03 | Nec Corp | 運用管理装置、運用管理システム、情報処理方法、及び運用管理プログラム |
JP2010097489A (ja) | 2008-10-17 | 2010-04-30 | Nec Corp | 分散データ処理システム、分散データ処理方法および分散データ処理用プログラム |
WO2011070910A1 (ja) * | 2009-12-07 | 2011-06-16 | 日本電気株式会社 | データ配置・計算システム、データ配置・計算方法、マスタ装置、及びデータ配置方法 |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7490323B2 (en) * | 2004-02-13 | 2009-02-10 | International Business Machines Corporation | Method and system for monitoring distributed applications on-demand |
JP5408442B2 (ja) * | 2010-01-21 | 2014-02-05 | 株式会社日立製作所 | 並列分散処理方法、及び、計算機システム |
CN102713862B (zh) * | 2010-02-15 | 2015-12-02 | 日本电气株式会社 | 故障原因提取装置、故障原因提取方法和程序记录介质 |
JP5532150B2 (ja) * | 2011-01-24 | 2014-06-25 | 日本電気株式会社 | 運用管理装置、運用管理方法、及びプログラム |
WO2012128388A1 (ja) * | 2011-03-23 | 2012-09-27 | 日本電気株式会社 | 運用管理システム、運用管理方法、及びプログラム |
JP5843459B2 (ja) * | 2011-03-30 | 2016-01-13 | 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International Business Machines Corporation | 情報処理システム、情報処理装置、スケーリング方法、プログラムおよび記録媒体 |
-
2013
- 2013-10-30 EP EP13851403.9A patent/EP2916222A4/en not_active Withdrawn
- 2013-10-30 WO PCT/JP2013/006437 patent/WO2014068980A1/ja active Application Filing
- 2013-10-30 JP JP2014544322A patent/JP6070717B2/ja active Active
- 2013-10-30 US US14/439,082 patent/US10296493B2/en active Active
- 2013-10-30 CN CN201380057593.9A patent/CN104769551B/zh not_active Expired - Fee Related
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH07325806A (ja) * | 1994-06-01 | 1995-12-12 | Nec Corp | マルチプロセッサシステムのプロセス割当方法 |
JP2006146668A (ja) * | 2004-11-22 | 2006-06-08 | Ntt Data Corp | 運用管理支援装置及び運用管理支援プログラム |
JP2009199533A (ja) * | 2008-02-25 | 2009-09-03 | Nec Corp | 運用管理装置、運用管理システム、情報処理方法、及び運用管理プログラム |
JP4872944B2 (ja) | 2008-02-25 | 2012-02-08 | 日本電気株式会社 | 運用管理装置、運用管理システム、情報処理方法、及び運用管理プログラム |
JP2010097489A (ja) | 2008-10-17 | 2010-04-30 | Nec Corp | 分散データ処理システム、分散データ処理方法および分散データ処理用プログラム |
WO2011070910A1 (ja) * | 2009-12-07 | 2011-06-16 | 日本電気株式会社 | データ配置・計算システム、データ配置・計算方法、マスタ装置、及びデータ配置方法 |
Non-Patent Citations (2)
Title |
---|
HADOOP, 27 September 2012 (2012-09-27), Retrieved from the Internet <URL:http://hadoop.apache.org/index.html> |
See also references of EP2916222A4 |
Also Published As
Publication number | Publication date |
---|---|
CN104769551A (zh) | 2015-07-08 |
JPWO2014068980A1 (ja) | 2016-09-08 |
JP6070717B2 (ja) | 2017-02-01 |
EP2916222A4 (en) | 2016-04-13 |
EP2916222A1 (en) | 2015-09-09 |
US20150293936A1 (en) | 2015-10-15 |
US10296493B2 (en) | 2019-05-21 |
CN104769551B (zh) | 2018-07-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108009016B (zh) | 一种资源负载均衡控制方法及集群调度器 | |
Jain et al. | Working vacations queueing model with multiple types of server breakdowns | |
JP5532150B2 (ja) | 運用管理装置、運用管理方法、及びプログラム | |
US7467291B1 (en) | System and method for calibrating headroom margin | |
CN112003763B (zh) | 网络链路的监测方法、监测装置、监测设备及存储介质 | |
JP5768796B2 (ja) | 運用管理装置、運用管理方法、及び、プログラム | |
Bashar | Autonomic scaling of cloud computing resources using BN-based prediction models | |
JP5954430B2 (ja) | 運用管理装置、及び、運用管理方法 | |
EP3465966B1 (en) | A node of a network and a method of operating the same for resource distribution | |
WO2016178316A1 (ja) | 計算機調達予測装置、計算機調達予測方法、及び、プログラム | |
Wang et al. | Design and implementation of an analytical framework for interference aware job scheduling on apache spark platform | |
Gong et al. | Finding constant from change: Revisiting network performance aware optimizations on iaas clouds | |
CN111431996B (zh) | 用于资源配置的方法、装置、设备和介质 | |
JP6176390B2 (ja) | 情報処理装置、解析方法、及び、プログラム記録媒体 | |
Shen et al. | Performance modeling of big data applications in the cloud centers | |
CN113158435B (zh) | 基于集成学习的复杂系统仿真运行时间预测方法与设备 | |
US20090217282A1 (en) | Predicting cpu availability for short to medium time frames on time shared systems | |
Ever | Performability analysis of cloud computing centers with large numbers of servers | |
JP6070717B2 (ja) | 分散データ処理システム、及び、分散データ処理方法 | |
JP2006092053A (ja) | システム使用率管理装置及びそれに用いるシステム使用率管理方法並びにそのプログラム | |
Rajput et al. | Study of performance evolution of three tier architecture based cloud computing system | |
Tutov | Models and methods of resources allocation of infocommunication system in cloud data centers | |
JP2018032245A (ja) | 計算機システム及びリソース制御方法 | |
JP5747389B2 (ja) | 計算機資源割当装置及び方法及びプログラム | |
CN115480908A (zh) | 一种内存池化方法以及相关装置 |
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: 13851403 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2013851403 Country of ref document: EP |
|
ENP | Entry into the national phase |
Ref document number: 2014544322 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 14439082 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |