JPWO2015008481A1 - Time-series data processing device, time-series data processing method, and program - Google Patents

Abstract

With the change in the processing contents of time series data, it is possible to specify the result data before and after switching between old and new without mixing the output of old and new result data while suppressing the increase in the amount of information added to the result data. Provide technology to output a lot of data. When there is an instruction to change from the old process to the new process, each processing unit 101 of the new process and the old process, the new and old identification information indicating the old process and the node identification information indicating the waiting node are included in the result data of the last old process. Is added to the result data of the first new process, and the control unit 121 adds new / old identification information representing the new process and node identification information, and selection information for selecting an output target from the own node. Of the result data arriving at the node and the result data in the delay unit 114, the last old process result data is selected before the first new process result data based on the additional information and the selection information, and is not selected. And a selection unit 113 for inserting the result data into the delay unit 114.

Description

  The present invention relates to a technique for processing time-series data, and more particularly to a time-series data processing apparatus, a time-series data processing method, and a computer-readable recording medium in which the content of processing on time-series data changes.

  First, “data processing for time-series data” will be described. In this specification, “data processing for time-series data” refers to processing for a series of data input one after another. The result data by the data processing on the time series data is also output one after another in the same manner. “To be input one after another” means that a plurality of data arrive at the processing device at different timings over a certain period. As an example of data processing for time series data, there is a process of taking the sum of the two most recent values of the time series data. Hereinafter, data processing for time series data is simply referred to as “time series data processing”.

  In the following description, the data array is represented as [1, 2, 3, 4], the data set is represented as {1, 2, 3, 4}, and the data set is represented as < 10 degrees, January 1>. In this way, the type of data is distinguished and represented by the type of parentheses. In addition, the numerical values of individual data such as 1 and 2 are expressed as “1” and “2”, and are distinguished from reference numerals.

  The period of the clock used for the operation of the time-series data processing device is called “cycle”, and this is used as a unit of period. “At the start of the cycle” means the first time of the period, and “at the end of the cycle” means the last time of the period. For example, when the frequency of the clock is 1 GHz, the first cycle for a certain reference time means a period between the reference time plus zero seconds and the reference time plus 1 nanosecond, and the Nth cycle is the reference time plus ( N-1) A period from nanoseconds to a reference time plus N nanoseconds. The start of the Mth cycle means a reference time plus (M-1) nanoseconds, and the end of the Mth cycle means a reference time plus M nanoseconds.

  Here, the input data is a set of one or more values. Examples include a set of temperature and date, a set of temperature only, a set of brand, stock price, and time. For example, a set of temperature and date is expressed as <10 degrees, January 1>. The input data and values handled by the device are also simply referred to as “data” when there is no risk of confusion.

  Time-series data processing is often performed on a plurality of input data or a window of time-series data. A “window” is a set of input data, and there are two types. One is a set of the most recent S input data, where S is a predetermined number (S is an integer equal to or greater than 1). The other is a set of input data in which a predetermined period is P, one element of the set of input data is regarded as time, and the time falls within the period P.

  For example, a set of temperature and date is input data, and <10 degrees, January 1>, <11 degrees, January 2>, <10 degrees, January 3>, <10 degrees, January 4 It is assumed that time series data of> is given. The windows that fit in the period from January 4 to the past three days are {<11 degrees, January 2>, <10 degrees, January 3>, <10 degrees, January 4>}.

  A set, a subset, and an element of input data in a window are referred to as “window data”. In time series data processing, this window data is often stored. In the following description, windows or window data are represented as arrays or sets.

  It is assumed that the time series data processing is executed using an IC (Integrated Circuit) chip having a plurality of partitions. A partition is an area where the contents of processing can be defined. An example of such an IC chip is an FPGA (Field Programmable Gate Array). In this case, circuit information or the like for configuring a circuit for executing processing is written in the partition. It is also possible to execute time-series data processing using a processor that reads a program from a memory and executes it. In this case, a program for specifying a procedure of time series data processing is written in the memory area.

  In time-series data processing, it may be desired to change the first user-defined process (old process) to the second user-defined process (new process). That is, during execution of the old process defined in the first partition on the IC chip, circuit information for executing the new process is written in the second partition, the new process is executed, and the old process is stopped. In this way, it may be desired to change the old process to the new process.

  As an example of changing the old process to the new process, consider an application that detects an emergency including a fire by averaging the values of temperature sensors distributed over a wide range over a certain period. In such an application, it is desirable to change the length of the period during which the temperature is averaged in order to increase the speed of following the temperature change when the temperature change becomes severe.

  As another example of changing the old process to the new process, consider an application that calculates an index for determining the timing of buying and selling financial products such as stock certificates. In such an application, there are a plurality of necessary indicators according to the brand of the financial product to be bought and sold and the buying and selling strategy, and the set of indicators to be output is different. Assume that such an application can calculate only a finite number of indicators simultaneously. Since the brands and strategies to be bought and sold change according to the time, it is desirable to change the index to be calculated according to the brands and trading strategies to be bought and sold at that time.

  An example of input data and output data when changing the old process to the new process will be described with reference to FIG. In this example, the old process that takes the sum of the most recent binary values is changed to a new process that takes the sum of the most recent three values. In addition, it is assumed that data is input every cycle, and the description will be made using time in units of one cycle.

  In FIG. 15, first, time series data “1” is input at the start of the first cycle. Therefore, the old process stores the value as window data. As a result, the window data stored in the old process changes from [] to [1].

  At the start of the second cycle, time series data “2” is input. Therefore, the old process stores the value as window data. As a result, the window data stored in the old process changes from [1] to [1,2]. The old process outputs the result data “3” using the window data [1, 2] because the latest two values are available.

  At the start of the third cycle, an instruction to change to the new process is given, and information for executing the new process is written into the apparatus. For example, it is assumed that circuit information for executing a new process is written in an area different from the old process in the FPGA. Then, the new process starts operation. It is assumed that the old process and the new process operate in parallel for a while.

  At the start of the third cycle, time series data “3” is input. Therefore, the old process stores the input data “3”. As a result, the window data stored in the old process changes from [1,2] to [2,3]. In the old process, the result data “5” is output using the window data [2, 3]. In the new process, the input data “3” is saved as window data. As a result, the window data stored in the new process changes from [] to [3].

  At the start of the fourth cycle, time series data “4” is input. Therefore, the old process stores the input data “4” and outputs the result data “7” using the window data [3, 4]. In the new process, the input data “4” is stored as window data. As a result, the window data stored by the new process changes from [3] to [3,4].

  At the start of the fifth cycle, time series data “5” is input. Therefore, the old process stores the input data “5” and outputs the result data “9” using the window data [4, 5]. In the new process, the input data “5” is stored. As a result, the window data stored by the new process changes from [3,4] to [3,4,5]. The new process outputs the result data “12” using the window data [3, 4, 5] because the latest three values are available.

  Thereafter, the old process is stopped or deleted, and the process change is completed.

  At the time of such a process change, there is a need to perform both processes while preventing the results of both the new process and the old process from being mixed and output to the outside of the apparatus. Specifically, the result sequence output to the outside of the apparatus is preferably the result sequence of the old process until the specific result of the old process, and then the result sequence of the new process starting from the specific result of the new process. In addition, after such a change in processing, there is a need to perform both processes while sending out the results of the old process that are still useful and usable after the change instruction to the new process. is there.

  As an example of the above needs at the time of processing change, consider an application for detecting an emergency situation including the above-mentioned fire. When the temperature change becomes severe, it is assumed that the width of the period in which the temperature is averaged is changed in order to increase the speed of following the temperature change. At this time, if the average of the temperature before changing the width of the period and the average of the temperature after changing the width of the period are mixed and output to the outside of the apparatus, the temperature change cannot be correctly followed. Therefore, when changing the duration of the temperature averaging period, it is necessary to prevent the results before and after the process change from being mixed and output outside the apparatus. In addition, after the process change is instructed and until the average temperature after the process change is output, the average temperature before the process change is still useful and usable. Therefore, when changing the width of the period during which the temperature is averaged, it is desirable to output more old process results from the apparatus until the new process results are output to the outside of the apparatus.

  As another example of the above-mentioned needs at the time of processing change, consider an application that issues a buy / sell command for financial products such as stock certificates. In order to change the brand to be traded, it is necessary to change the processing content of the trading instruction. Hereinafter, the brand before the change is described as “old brand”, and the brand after the change is described as “new brand”. At this time, if a trading instruction regarding the old brand and a trading instruction regarding the new brand are mixed and output to the outside of the apparatus, this may cause a mistake in trading. Therefore, when changing the processing contents of the trade instruction, it is necessary to prevent the results before and after the process change from being mixed and output to the outside of the apparatus. In addition, even after an instruction to change to processing related to a new brand, there is a case where it is desired to receive a trading instruction regarding an old brand. For example, it is assumed that a change to processing related to a new brand is instructed after a command for selling related to an old brand. In such a case, it is desirable that a trading instruction regarding the old brand is output until a buying instruction regarding the old brand is issued, and thereafter a trading instruction regarding the new brand is output. In this way, the sequence of results output from the processing related to the old brand that is not ready to be stopped is useful and available. Therefore, when changing from the process related to the old issue to the process related to the new issue, the result column from the old process that is not ready to be stopped is output. It is desirable to output the process related to the brand.

  Technologies related to the above processing are described in Patent Documents 1 to 6 and Non-Patent Documents 1 and 2.

  Patent Document 1 discloses a system using a reconfigurable device.

  Patent Document 2 discloses a time-series data compression apparatus that performs two types of processing.

  Patent Documents 3 and 4 disclose a processing apparatus that handles information to which a version number (updated version number) is added.

  Patent Documents 5 and 6 disclose processing apparatuses that handle information to which output scheduled times are added.

  Non-Patent Document 1 discloses a time-series data processing apparatus capable of two types of processing, old processing and new processing.

  Non-Patent Document 2 discloses a time-series data processing apparatus capable of two types of processing, old processing and new processing.

  Here, by combining the above related technologies as appropriate, it is possible to perform two types of processing, new processing and old processing, on time-series data, and identify and output the results of the old processing and the new processing. It is.

(Operation example of time-series data processing device (part 1) based on related technology)
For example, a method for identifying and outputting the result of the old process and the result of the new process by combining the related techniques described in Patent Documents 2 to 4 and Non-Patent Documents 1 and 2 will be described. FIG. 16 shows a configuration of a time-series data processing device 600 that can be realized by a combination of these related technologies. The time-series data processing device 600 includes a processing unit 601, a control unit 611, a selection unit 621, and a synchronization unit 631.

  The processing unit 601 performs user definition processing. Here, the time-series data processing device 600 includes a plurality of processing units 601. The processing unit 601 in which the user-defined process corresponding to the old process is written performs the old process, and the processing unit 601 in which the user-defined process corresponding to the new process is written performs the new process.

  The control unit 611 adds a processing number that uniquely identifies the type of user-defined processing and a version number that distinguishes old processing and new processing to the processing result of the processing unit 601. When a certain type of user-defined process is changed from the old process to the new process, the control unit 611 adds the same process number to the old process and the new process, and adds different version numbers. To do.

  When one or more result data arrives, the selection unit 621 selects one of the result data. Here, the time-series data processing device 600 includes one or more selection units 621, and each selection unit 621 receives result data output from the processing unit 601 or another selection unit 621. Then, based on the process number and version number added to each arrived result data, the selection unit 621 discards the result of the older version number for the same process number and selects the result data of the newer version number. . Each selection unit 621 transmits the selected data to the next selection unit 621 or synchronization unit 631.

  The synchronization unit 631 records the latest version number for each process. When the result data arrives, the synchronization unit 631 discards the result data having a version number older than the latest recorded version number, and outputs the result data of the new process to which the latest version number is added to the outside. . Further, when the result data of the version number newer than the latest recorded version number arrives, the synchronization unit 631 updates the recorded version number to the new version number.

  An example of the operation of the time-series data processing apparatus 600 configured as described above will be described with reference to FIGS. In this operation example, it is assumed that the processing unit 601-1 performs an old process and the processing unit 601-2 performs a new process. In addition, an output from the processing unit 601-1 is input to the selection unit 621-1, and an output from the processing unit 601-2 is input to the selection unit 621-2. Further, it is assumed that an output from the selection unit 621-1 is input to the selection unit 621-2, and an output from the selection unit 621-2 is input to the synchronization unit 631. In the following, data output from each processing unit 601 is expressed as <result data, processing number, version number>.

  First, in the second cycle, the processing unit 601-1 that has performed the old process outputs <o1, 10, 1> to the selection unit 621-1. In addition, the processing unit 601-2 that has performed the new process outputs <n2, 10, 2> to the selection unit 621-2. Here, it is assumed that the result data “n2” of the new process is to be output outside the apparatus after the result data “o1” of the old process. FIG. 17 shows output information inside the time-series data processing device 600 at the end of the second cycle.

  Next, in the third cycle, the selection unit 621-1 does not need to make a selection because the arrival result data is one of <o1, 10, 1>. Therefore, the selection unit 621-1 outputs <o1, 10, 1> to the selection unit 621-2. Similarly, the selection unit 621-2 outputs one of the received result data <n2, 10, 2> to the synchronization unit 631. FIG. 18 shows output information inside the time-series data processing device 600 at the end of the third cycle.

  Next, in the fourth cycle, the selection unit 621-2 outputs one result data <o1, 10, 1> to the synchronization unit 631 because there is only one result data. In addition, the synchronization unit 631 records “1” as the latest version number of the processing number “10”. Therefore, the synchronization unit 631 compares the arrival result data <n2, 10, 2> with the latest version number “1” corresponding to the processing number “10”, and determines the version number of the arrival result data. Is new. Then, the synchronization unit 631 updates the latest version number of the processing number “10” to “2” and outputs <n2, 10, 2> to the outside. FIG. 19 shows output information inside the time-series data processing device 600 at the end of the fourth cycle.

  Next, in the fifth cycle, the synchronization unit 631 compares the arrival result data <o1, 10, 1> with the latest version number “2” corresponding to the processing number “10”, and the arrival result Judge that the version number of the data is older. Therefore, the synchronization unit 631 discards the result data.

  As described above, the time-series data processing apparatus 600 based on the combination of related techniques can output the result of the new process to the outside after outputting the result of the old process to the outside. However, with this combination of related technologies, the purpose of outputting the result data “n2” of the new process after the result data “o1” of the old process cannot be achieved. As described above, in the combination of the related techniques, the output contents and order of the result data depend on the output timing of the old and new result data from each processing unit 601 and the arrival timing of the selection unit 621 or the synchronization unit 631. Therefore, although the time-series data processing apparatus 600 can output the old and new result data without mixing them, it cannot output the specified result data of the new process next to the specified result data of the old process. .

(Another operation example of time-series data processing device (part 1) based on related technology)
Another example of the operation of the time-series data processing apparatus 600 based on the combination of related technologies described above will be described with reference to FIGS. Here, similarly to the above-described operation example, it is assumed that the processing unit 601-1 performs an old process and the processing unit 601-2 performs a new process. Similarly to the operation example described above, it is assumed that the output from the selection unit 621-1 is input to the selection unit 621-2, and the output from the selection unit 621-2 is input to the synchronization unit 631. However, in this operation example, unlike the above-described operation example, the output from the processing unit 601-1 is input to the selection unit 621-2, and the output from the processing unit 601-2 is input to the selection unit 621-1. Shall.

  First, in the second cycle, the processing unit 601-1 that has performed the old processing outputs <o1, 10, 1> to the selection unit 621-2. Here, it is assumed that the result data “n2” of the new process is to be output outside the apparatus after the result data “o1” of the old process. FIG. 20 shows output information inside the time-series data processing device 600 at the end of the second cycle.

  Next, in the third cycle, the selection unit 621-2 outputs one result data <o1, 10, 1> that has arrived to the synchronization unit 631. In addition, the processing unit 601-1 that has performed the old process outputs <o2, 10, 1> to the selection unit 621-2. In addition, the processing unit 601-2 that has performed the new process outputs <n2, 10, 2> to the selection unit 621-1. FIG. 21 shows output information inside the time-series data processing device 600 in the third cycle.

  Next, in the fourth cycle, the selection unit 621-2 outputs one result data <o2, 10, 1> that has arrived to the synchronization unit 631. Further, the selection unit 621-1 outputs the arrived result data <n2, 10, 2> to the selection unit 621-2. Further, the synchronization unit 631 compares the arrived result data <o1, 10, 1> with the latest version number “1” corresponding to the processing number “10”, and the version number added to the result data is Judge that it is the latest. Therefore, the synchronization unit 631 outputs <o1, 10, 1> and leaves the version number in the table as it is. FIG. 22 shows output information inside the time-series data processing device 600 at the end of the fourth cycle.

  Next, in the fifth cycle, the synchronization unit 631 compares the arrived result data <o2, 10, 1> with the latest version number “1” corresponding to the processing number “10”, and outputs the result data. It is determined that the added version number is the latest. Therefore, the synchronization unit 631 outputs <o2, 10, 1> and leaves the version number in the table as it is.

  As described above, the time-series data processing apparatus 600 based on the combination of the related techniques can sequentially output the results of the old processing to the outside before outputting the results of the new processing to the outside. However, with this combination of related technologies, the purpose of outputting the result data “n2” of the new process after the result data “o1” of the old process is not achieved. As described above, in the time-series data processing apparatus 600, the output contents and order of the result data depend on the output timing of the old and new result data from each processing unit 601 and the arrival timing of the selection unit 621 or the synchronization unit 631. Therefore, although the time-series data processing apparatus 600 can output the old and new result data without mixing them, it cannot output the specified result data of the new process next to the specified result data of the old process. .

(Configuration of time-series data processing device (part 2) based on related technology)
For example, a method of identifying and outputting the result of the old process and the result of the new process by a combination of related techniques described in Patent Documents 5 to 6 will be described. FIG. 23 shows a configuration of a time-series data processing device 700 that can be realized by a combination of these related technologies. The time-series data processing device 700 includes a processing unit 701 and a network unit 731. The network unit 731 includes a transmission unit 732, a time comparison unit 733, and a delay unit 734.

  The processing unit 701 performs user definition processing. Here, the time-series data processing device 700 includes a plurality of processing units 701. The processing unit 701 in which the user-defined process corresponding to the old process is written performs the old process, and the processing unit 701 in which the user-defined process corresponding to the new process is written performs the new process. Further, the processing unit 701 adds the processing number, the network address, and the scheduled output time to the processing result and outputs the result. The process number is information that can uniquely identify the process. The network address is identification information of the network unit 731.

  The network unit 731 controls the output order of the arrived result data. Here, the time-series data processing device 700 includes one or more network units 731. Each network unit 731 receives data output from the processing unit 701 or another network unit 731. Each network unit 731 outputs data to another network unit 731 or an output port outside the apparatus.

  The delay unit 734 delays by holding the result data.

  When the network address added to the result data arriving at the transmission unit 732 indicates the own node, the time comparison unit 733 outputs the scheduled output time added to the arrival result data and the result data of the delay unit 734, respectively. Compare Then, the time comparison unit 733 selects the result data that has reached the scheduled output time, and causes the transmission unit 732 to output the result data. In addition, the time comparison unit 733 stores the result data not selected from the transmission unit 732 to the delay unit 734.

  The transmission unit 732 transmits the arrived result data to the next network unit 731 or an output port outside the apparatus according to the instruction of the time comparison unit 733 or stores it in the delay unit 734.

  An example of the operation of the time-series data processing apparatus 700 configured as described above will be described with reference to FIGS. In this operation example, it is assumed that the processing unit 701-1 performs an old process and the processing unit 701-2 performs a new process. In addition, it is assumed that the output from the processing unit 701-1 is input to the network unit 731-1 and the output from the processing unit 701-2 is input to the network unit 731-2. Further, it is assumed that the output from the network unit 731-1 is input to the network unit 731-2, and the output from the network unit 731-2 is output to the outside. The network address “1” indicates the network unit 731-1, and the network address “2” indicates the network unit 731-2. Hereinafter, data output from the processing unit 701 is expressed as <result data, process number, network address, scheduled output time>. The scheduled output time is expressed in units of one cycle.

  First, in the second cycle, the processing unit 701-1 that has performed the old process outputs the result data <o1, 10, 2, 4> to the network unit 731-1. In addition, the processing unit 701-2 that has performed the new process outputs <n2, 10, 2, 5> to the network unit 731-2. Here, it is assumed that the result data “n2” of the new process is to be output outside the apparatus after the result data “o1” of the old process. FIG. 24 shows output information inside the time-series data processing device 700 at the end of the second cycle.

  Next, in the third cycle, the time comparison unit 733-1 indicates that the network address “2” added to the result data <o1, 10, 2, 4> arriving at the transmission unit 732-1 indicates its own node. Therefore, the transmission unit 732-1 is instructed to output the output without comparing the scheduled output times. Therefore, the transmission unit 732-1 outputs <o1, 10, 2, 4> to the network unit 731-2. Also, the time comparison unit 733-2 indicates that the network address “2” added to the arrival result data <n2, 10, 2, 5> indicates the own node, and therefore the current time is the scheduled output time “5”. It is determined whether or not “is reached. Here, the current time (third cycle) has not reached the scheduled output time “5”. Therefore, the delay unit 734-2 stores the result data <n2, 10, 2, 5>. FIG. 25 shows output information inside the time-series data processing device 700 at the end of the third cycle.

  Next, in the fourth cycle, the time comparison unit 733-2 receives the result data <o1, 10, 2, 4> arriving at the transmission unit 732-2 and the result data <o1, 10-2, 44-2> For n2, 10, 2, 5>, since the attached network address “2” indicates the own node, it is determined whether or not the scheduled output time has been reached. Here, the current time (fourth cycle) has reached the scheduled output time “4”. Therefore, the time comparison unit 733-2 selects the result data <o1, 10, 2, 4> and instructs the transmission unit 732-2 to output it. Then, the transmission unit 732-2 outputs the result data <o1, 10, 2, 4> to the output port to the outside of the apparatus. In addition, the transmission unit 732-2 stores the result data <n2, 10, 2, 5> not selected in the delay unit 734-2. FIG. 26 shows output information inside the time-series data processing device 700 at the end of the fourth cycle.

  Next, in the fifth cycle, the time comparison unit 733-2 automatically adds the network address “2” to the result data <n2, 10, 2, 5> held in the delay unit 734-2. Since the node is indicated, it is determined whether or not the scheduled output time has been reached. Here, the current time (fifth cycle) has reached the scheduled output time “5”. Therefore, the time comparison unit 733-2 selects the result data <n2, 10, 2, 5> and instructs the transmission unit 732-2 to output it. Then, the transmission unit 732-2 outputs the result data <n2, 10, 2, 5> to the output port to the outside of the apparatus.

  As described above, the time-series data processing device 700 based on the combination of the related technologies outputs the specified result data of the old process (“o1” in this example) to the outside, and then specifies the result data of the new process. The purpose of outputting the output ("n2" in this example) to the outside is achieved.

Kyumars Sheykh Esmail et al., Changing Flights in Mid-air: A model for Safely Modifying Continuous Queries, In Proceedings of SIGMOD’11, June 2011 Masamichi Takagi, Takashi Takenaka and Hiroaki Inoue: Dynamic query switching for complex event processing on FPGAs, In Proceedings of FPL 2012

Japanese translation of PCT publication No. 2004-508617 JP 2002-049747 A JP 2001-075908 A Japanese Patent Laying-Open No. 2005-049102 Special table 2012-520648 gazette Special table 2011-525322 gazette

  However, Patent Documents 1 to 6 and Non-Patent Documents 1 and 2 described above have the following problems.

  Patent Document 1 does not mention a problem that occurs when a device configuration file (circuit information for configuring a predetermined circuit) is rewritten. Therefore, the related art described in Patent Document 1 cannot identify the results of the old and new processes when the new process and the old process are changed.

  The related technique described in Patent Document 2 applies one of two types of processing based on the result of a predetermined determination on the result of preprocessing. In this related technique, it is not distinguished by which process the final processing result is output. Thus, this related technique cannot identify the results of the two types of processing.

  Further, in the related techniques described in Patent Documents 3 and 4, by adding a version number, information old and new is identified and information of an older version number is discarded. For this reason, this related technique cannot output more useful information of older version number information.

  In addition, when the related technology described in Patent Document 2 and the related technology described in Patent Document 3 or Patent Document 4 are combined, two types of processing, new processing and old processing, are performed on time-series data, As a result, it is possible to determine whether the data is generated by new or old processing. However, even if these related technologies are combined, it is not possible to output more useful data that is the result data of the old process.

  Further, as described with reference to FIGS. 16 to 22, when the related technology described in Patent Documents 2 to 4 and the related technology described in Non-Patent Documents 1 to 2 are combined, the old process and the new process are performed. The result can be identified. As a result, as described above, it is possible to output the result data of the old process and the result data of the new process without mixing by combining these related techniques. However, as described above, even if these related techniques are combined, after the specified result of the old process is output to the outside, the specified result of the new process cannot be output to the outside.

  Further, as described with reference to FIGS. 23 to 26, when the related techniques described in Patent Documents 5 to 6 are combined, after specifying the result data of the old process, the result data of the new process is specified. Can be output. However, when these related technologies are combined, it is necessary to add and transmit and store the scheduled output time to the processing result data. Therefore, when the number of processes operating in one device increases, there arises a problem that resources necessary for transmission and storage increase. For example, when 64 bits are used for the scheduled output time, in the example of FIG. 24, a wire from the processing unit 701-1 to the network unit 731-1, a wire from the processing unit 701-2 to the network unit 731-2, a network A 64-bit resource is additionally required for each wire from the unit 731-1 to the network unit 731-2. Further, each of the delay unit 734-1 and the delay unit 734-2 requires an additional 64-bit resource.

  The present invention has been made to solve the above-described problems.With the change of processing contents in time-series data processing, it is possible to specify the result data before and after switching between the new and old output, and the result data by the old and new processing can be specified. An object is to provide a technique for outputting more useful items. In providing the above technique, the present invention suppresses an increase in the amount of information added to the processing result data, and does not mix the output of the result data by the old and new processing.

  The time-series data processing device according to the present invention executes a first process on a plurality of nodes including a selection unit and a delay unit, and input data, thereby outputting a first process result data. A processing unit, a second processing unit that outputs the result data of the second processing by executing the second processing on the input data, and a change from the first processing to the second processing When the instruction is received from the outside, the old process and the result data (the last old process result data) determined to be the last output to the outside of the apparatus among the result data of the first process Additional information including old process information and new process result data and node identification information indicating the node in the device that waits for the new process result data is added as new and old identification information indicating any of the new processes. As well as Information indicating new processing as the new / old identification information with respect to result data (result data of the first new processing) that is first determined to be output to the outside of the apparatus among the result data of the second processing; A control unit for adding additional information including the node identification information, and the selection unit stores selection information representing information for selecting result data to be output from the node, Select one of the result data output from the first processing unit or the second processing unit and arrived at its own node and the result data held in the delay unit and send it to another node or outside At the time of output, the result of the last old process is selected before selecting the result data of the first new process based on the additional information added to each result data and the stored selection information. As well as selecting the data, and inserts the results did not select the data to the delay unit, the delay unit, the included in each node, to delay by holding results data that is not selected by the selection unit.

  The time-series data processing method according to the present invention includes a second process for outputting the result data of the second process based on the input data from the first process for outputting the result data of the first process based on the input data. When the change instruction is received, the old process and the result data (the result data of the last old process) determined to be output to the outside of the apparatus at the end of the result data of the first process Additional information including old process information and new process result data and node identification information indicating the node in the device that waits for the new process result data is added as new and old identification information indicating any of the new processes. Then, among the result data obtained by the second process, the result data determined to be output first to the outside of the apparatus (result data of the first new process) is used as the new / old identification information. Additional information including processing information and node identification information is added, and each node stores selection information representing information for selecting result data to be output from the own node, and Each result when selecting one of the result data output from the first process or the second process and arriving at the own node and the held result data and outputting it to another node or outside Based on the additional information added to the data and the selection information stored, the result data of the last old process is selected before selecting the result data of the first new process, and the selection is made Delay the result data that you did not do.

  The computer-readable recording medium according to the present invention outputs the second process result data based on the input data from the first process that outputs the result data of the first process based on the input data. An old process is performed on the result data (the last old process result data) that is determined to be output to the outside of the apparatus at the end of the first process result data received from the process change instruction. And additional information including information indicating the old process as the old and new identification information indicating that the process is one of the new process and node identification information indicating the node in the apparatus that waits for the result data of the old process and the result data of the new process. In addition, with respect to the result data (result data of the first new process) that is determined to be output to the outside of the apparatus among the result data by the second process, the new and old identification information And additional information including the node identification information and information representing a new process, and in each node, selection information representing information for selecting result data to be output from the own node is stored, When selecting any one of the result data output from the first process or the second process and arriving at its own node and the result data held and outputting it to another node or outside, Based on the additional information added to each result data and the selection information stored, the result data of the last old process is selected before the result data of the first new process is selected. A program for causing the computer apparatus to execute delay of the result data not selected is recorded.

  The present invention provides a technique for enabling output of more useful results data of old and new processes by enabling specification of result data before and after switching between new and old output in accordance with changes in processing contents in time-series data processing. Can do. In providing the above technique, the present invention suppresses an increase in the amount of information added to the processing result data, and does not mix the output of the result data by the old and new processing.

It is a block diagram which shows the structure of the time series data processing apparatus as the 1st Embodiment of this invention. It is a flowchart which shows operation | movement of the control part of the time series data processing apparatus as the 1st Embodiment of this invention. It is a flowchart which shows operation | movement of the selection part of the time series data processing apparatus as the 1st Embodiment of this invention. It is a block diagram which shows the structure of the time series data processing apparatus as the 2nd Embodiment of this invention. It is a figure which shows an example of the result data structure output from a process part in the 2nd Embodiment of this invention. It is a flowchart which shows operation | movement of the control part of the time series data processing apparatus as the 2nd Embodiment of this invention. It is a flowchart which shows operation | movement of the selection part of the time series data processing apparatus as the 2nd Embodiment of this invention. In the specific example of operation | movement of the time series data processing apparatus as the 2nd Embodiment of this invention, it is a figure explaining the data output from each part at the end of a 2nd cycle. In the specific example of operation | movement of the time series data processing apparatus as the 2nd Embodiment of this invention, it is a figure explaining the data output from each part at the time of a 3rd cycle completion | finish. In the specific example of operation | movement of the time series data processing apparatus as the 2nd Embodiment of this invention, it is a figure explaining the data output from each part at the end of a 4th cycle. It is a figure explaining the data output from each part at the time of completion | finish of a 5th cycle in the specific example of operation | movement of the time series data processing apparatus as the 2nd Embodiment of this invention. In the specific example of operation | movement of the time series data processing apparatus as the 2nd Embodiment of this invention, it is a figure explaining the data output from each part at the time of a 6th cycle completion | finish. In the specific example of operation | movement of the time series data processing apparatus as the 2nd Embodiment of this invention, it is a figure explaining the data output from each part at the time of a 7th cycle completion | finish. It is a figure explaining an example of the additional resource required in order to implement | achieve the time series data processing apparatus as the 2nd Embodiment of this invention. It is a time transition diagram of the result data output by the old process and new process of the time series data processing apparatus of related technology. It is a block diagram which shows the structure of an example of the time series data processing apparatus realizable by the combination of a related technique. FIG. 17 is a diagram for describing data output from each unit at the end of the second cycle in an example of the operation of the time-series data processing device of the related technology illustrated in FIG. 16. FIG. 17 is a diagram for describing data output from each unit at the end of the third cycle in an example of the operation of the time-series data processing device of the related technology illustrated in FIG. 16. FIG. 17 is a diagram illustrating data output from each unit at the end of the fourth cycle in an example of the operation of the related-art time-series data processing device illustrated in FIG. 16. FIG. 17 is a diagram for describing data output from each unit at the end of the second cycle in another example of the operation of the time-series data processing device of the related technique illustrated in FIG. 16. FIG. 17 is a diagram illustrating data output from each unit at the end of the third cycle in another example of the operation of the time-series data processing device of the related technology illustrated in FIG. 16. FIG. 17 is a diagram illustrating data output from each unit at the end of the fourth cycle in another example of the operation of the time-series data processing device of the related technology illustrated in FIG. 16. It is a block diagram which shows the structure of the other time series data processing apparatus realizable by the combination of a related technique. FIG. 24 is a diagram illustrating data output from each unit at the end of the second cycle in an example of the operation of the time-series data processing device of the related technology illustrated in FIG. 23. FIG. 24 is a diagram for describing data output from each unit at the end of the third cycle in an example of the operation of the time-series data processing device of the related technology illustrated in FIG. 23. FIG. 24 is a diagram illustrating data output from each unit at the end of the fourth cycle in an example of the operation of the time-series data processing device of the related technology illustrated in FIG. 23.

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

(First embodiment)
The time-series data processing apparatus 100 as the first embodiment of the present invention is an apparatus capable of executing different user-defined processes in a plurality of areas. The time-series data processing device 100 is configured to be able to change a user-defined process to another user-defined process at the time of execution. Hereinafter, the user-defined process is also simply referred to as “process”. The process before the change is also referred to as “old process”, and the process after the change is also referred to as “new process”.

  For example, the time-series data processing device 100 can be configured by an IC chip having a plurality of areas. An example of such an IC chip is an FPGA (Field Programmable Gate Array). Such an IC chip can start the operation by writing circuit information for executing the new process in a region other than the region where the old process operates while the old process is operated in a certain region. Is possible. Such an IC chip can then stop the old process by deleting circuit information for executing the old process from the corresponding area. In this way, such an IC chip can realize a process change while shortening a period during which no result is output.

  FIG. 1 is a block diagram showing the configuration of the time-series data processing device 100. In FIG. 1, the time-series data processing apparatus 100 includes a plurality of processing units 101, one or more network units 111, and a control unit 121. The network unit 111 includes a transmission unit 112, a selection unit 113, and a delay unit 114. The network unit 111 constitutes an embodiment of a node in the present invention. Moreover, the transmission part 112 comprises one Embodiment of the selection part in this invention with the selection part 113. FIG.

  In FIG. 1, time series data is input to each processing unit 101 from the outside. Data output from each processing unit 101 is input to one of the network units 111. Each network unit 111 receives data output from the processing unit 101 or another network unit 111. Further, data output from each network unit 111 is input to another network unit 111 or output outside the apparatus. As described above, the time-series data processing device 100 is configured to bundle the result data output from each processing unit 101 into one by a network including one or more network units 111 and then output the result data to the outside of the device. It has become. In FIG. 1, two processing units 101, one network unit 111, and one control unit 121 are shown. However, the number of units included in the time-series data processing device of the present invention is limited. is not.

  The processing unit 101 generates and outputs result data by executing a defined process on the input time-series data. For example, circuit information for executing an old process may be written in a certain processing unit 101, and circuit information for executing a new process may be written in another processing unit 101. In addition, the process part 101 which performs an old process comprises one Embodiment of the 1st process part in this invention. The processing unit 101 that executes the new processing constitutes an embodiment of the second processing unit in the present invention.

  The control unit 121 receives information representing a process change instruction from the outside. The information indicating the change instruction may include information for specifying the processing unit 101 that is executing the old process and information for specifying the processing unit 101 in which the new process is written.

  In addition, the control unit 121 adds additional information including new / old identification information representing the old process and node identification information to the result data of the last old process. The result data of the last old process refers to the result data determined to be the last output data out of the apparatus among the result data output from the processing unit 101 that executes the old process. For example, the control unit 121 may determine that the result data output when the processing unit 101 that executes the old process can be stopped is the result data of the last old process. The stoppable situation may be a situation in which a series of processes of the processing unit 101 that executes the old process is completed.

  Further, the control unit 121 adds additional information including new / old identification information representing the new process and node identification information to the result data of the first new process. The result data of the first new process refers to the result data determined to be output to the outside of the apparatus first among the result data output from the processing unit 101 that executes the new process. For example, the control unit 121 may determine that the result data output after a predetermined cycle after the activation of the processing unit 101 that executes the new process is the result data of the first new process.

  The old and new identification information is information that can identify the old process and the new process. That is, the new / old identification information indicates that the process of outputting the result data is a process change target that progresses in the time-series data processing apparatus 100 and is either the old process or the new process.

  The node identification information is information for identifying the network unit 111 that waits for the result data of the new process and the result data of the old process. As described above, the time-series data processing device 100 bundles the result data output from each of the plurality of processing units 101 into one by a network composed of one or more network units 111, and then externally stores the result data. Output. Therefore, the control unit 121 sets the location of the common network unit 111 through which both result data passes as common node identification information added to the result data of the last old process and the result data of the first old process. Calculate it. Note that the node identification information may indicate that no waiting is required.

  In addition, after the first new process result data is output, the control unit 121 performs a new process on the new process result data output from the new process unit 101 immediately before the change of the new and old processes is completed. New and old identification information indicating processing and node identification information indicating that no waiting is required are added. When the change between the new and old processes is completed, the result data of the first new process (result data to which additional information including new and old identification information indicating the new process and node information indicating the meeting node is added) exists in the device. It means that it will not (output outside the device).

  Hereinafter, the result data to which the additional information is added may be simply referred to as “result data”. In the time-series data processing apparatus 100, the result data is input / output or held by each unit together with the additional information added.

  The selection unit 113 stores selection information representing information for selecting result data to be output from the own node. Here, the local node refers to the network unit 111 including the selection unit 113. For example, the selection unit 113 may store information indicating whether or not the result data of the last old process has passed through its own node as selection information. In this case, the selection unit 113 initializes the selection information to indicate that it has not been passed before starting the process change.

  The selection unit 113 selects one of the result data output from each processing unit 101 and arrives at its own node and the result data held in the delay unit 114 and outputs the selected result data to another node or the outside. . At this time, the selection unit 113 refers to the additional information added to each result data and the stored selection information, thereby selecting the last old process before selecting the result data of the first new process. Select the result data. That is, the selection unit 113 does not select the result data of the first new process before selecting the result data of the last old process.

  Specifically, when the node identification information added to the result data indicates the own node and the new and old identification information indicates a new process, the selection unit 113 does not indicate that the selection information has passed. It is decided not to select the result data.

  Further, the selection unit 113 indicates that the node identification information added to the result data indicates that no waiting is required, the new / old identification information represents a new process, and the delay unit 114 stores the result data of the first new process. If so, it is decided not to select this result data. Note that if the new and old identification information added to the result data stored in the delay unit 114 indicates a new process and the node identification information indicates the own node, the selection unit 113 results data of the first new process. Can be determined to be stored in the delay unit 114.

  Further, the selection unit 113 selects one of the other result data so that the output order of the results output from the same processing unit 101 is not changed and based on some priority. Here, “result data other than that” includes result data in which the node identification information indicates the own node and the old and new identification information indicates the new process when the selection information indicates passed. In the “other result data”, when the result data of the new process is not stored in the delay unit 114, the result data in which the node identification information indicates that no waiting is required and the old and new identification information indicates the new process Is included. In addition, “other result data” includes result data in which the added node identification information does not indicate the own node, or in which the old and new identification information indicates the old process. Further, the certain priority may be a priority set for each type of processing or round robin.

  In addition, when the node identification information added to the selected result data indicates the own node, and the new and old identification information indicates the old process, the selection unit 113 updates the selection information to indicate that it has passed.

  The selection unit 113 outputs the selected result data to the next network unit 111 or the outside of the apparatus via the transmission unit 112. In addition, the selection unit 113 causes the transmission unit 112 to insert the result data that has not been selected into the delay unit 114.

  The transmission unit 112 transmits the input result data or the result data stored in the delay unit 114 to the next network unit 111 or an output port to the outside of the apparatus under the control of the selection unit 113. In addition, the transmission part 112 outputs the additional information added with the result data.

  The delay unit 114 delays by holding the result data under the control of the selection unit 113. The delay unit 114 may be capable of holding a plurality of result data. In that case, the delay unit 114 stores the result data from the processing unit 101 of the old process so that the result data is extracted before the result data from the processing unit 101 of the new process. Further, the delay unit 114 stores result data from the same processing unit 101 so as to be extracted in the order of arrival. For example, the delay unit 114 may have a queue structure.

  The operation of the time-series data processing apparatus 100 configured as described above will be described with reference to FIGS.

  In the following description, a clock cycle that defines a unit time required to complete a predetermined operation is referred to as “one cycle”. The time-series data processing device 100 is assumed to be constituted by an IC chip having a plurality of areas (partitions). In each partition, setting information called configuration data and configuration data is set to create a circuit for executing user-defined processing.

  In the following, “instruct” means that an arbitrary part in the apparatus operates a signal of a wiring connected to another part, or an instruction content is given to a storage device existing in another part. This means writing a value to be represented. “Set” and “store” mean that an arbitrary part in the device writes a value in a storage device existing in another part. In addition, “send” and “send” means that any part in the device operates a signal of a wiring connected to another part, or a value is stored in a storage device existing in another part. Means to write. Further, “receive” and “receive” mean that any part in the apparatus is a target to which information is transmitted by another part.

  In addition, the time-series data processing device 100 starts the following operation when receiving an instruction to change the process from the old process to the new process from the outside. Note that the time-series data processing apparatus 100 may receive such a change instruction from a host computer or the like connected to the outside. In addition, it is assumed that user-defined information related to the new process to be changed is written to an arbitrary processing unit 101 configured by a partition different from the processing unit 101 that is executing the old process.

  First, the operation by the control unit 121 is shown in FIG.

  In FIG. 2, first, the control unit 121, for the result data of the last old process output by the processing unit 101 of the old process, the new / old identification information indicating the old process and the node identification information indicating the waiting node. Is added (S1). Note that at a later timing, the control unit 121 continues to transmit information that instructs the old processing unit 101 not to output the result data.

  Next, the control unit 121 adds new / old identification information indicating the new process and node identification information indicating the waiting node to the result data of the first new process output by the processing unit 101 of the new process. (S2). Note that at a timing earlier than this, the control unit 121 continues to transmit information instructing not to output result data to the processing unit 101 of the new process. Further, the control unit 121 outputs new data to the result data output by the processing unit 101 of the new process immediately after the result data of the first new process is output and immediately before the change of the new and old processes is completed. New and old identification information indicating processing and node identification information indicating that no waiting is required are added.

  And the control part 121 repeats operation | movement of S1 and S2 of FIG.

  Above, description of operation | movement of the control part 121 is complete | finished.

  Next, the operation of the network unit 111 is shown in FIG.

  In FIG. 3, first, the selection unit 113 initializes selection information to information indicating that it has not been passed (S11).

  Next, the selection unit 113 arbitrates the right to use the output port between the result data arriving at the transmission unit 112 of its own node and the oldest result data stored in the delay unit 114 ( S12). Note that “arbitration of the right to use an output port” specifically means a process of selecting one piece of data that can use an output port. The fact that data has been selected in this arbitration is also referred to as “selected by arbitration” or “data acquired a port”. When the delay unit 114 has a queue structure, the “oldest” is an entry that is first taken out from the queue.

  Specifically, in S12 of FIG. 3, when the selection information does not indicate that the selection has been completed, the selection unit 113 indicates that the added node identification information indicates the own node, and the old and new identification information indicates the new process. It is decided not to select the result data. In addition, the selection unit 113 indicates that the added node identification information indicates that no waiting is required, the new / old identification information indicates a new process, and the delay unit 114 stores the result data of the first new process. , Decide not to select this result data. Further, the selection unit 113 selects one of the other result data so that the output order of the results output from the same processing unit 101 is not changed and based on some priority. As described above, “result data other than that” includes result data in which the node identification information indicates the own node and the old and new identification information indicates the new process when the selection information indicates passed. It is. In the “other result data”, when the result data of the new process is not stored in the delay unit 114, the result data in which the node identification information indicates that no waiting is required and the old and new identification information indicates the new process Is included. In addition, “other result data” includes result data in which the added node identification information does not indicate the own node, or in which the old and new identification information indicates the old process. Further, the certain priority may be a priority set for each type of processing or round robin.

  Next, regarding the result data selected by the arbitration, the selection unit 113 indicates that the selection information has passed if the node identification information indicates the own node and the new and old identification information indicates the old process. Update (S13).

  Next, the selection unit 113 outputs the result data selected by the arbitration together with the added information to the next network unit 111 or an output port to the outside of the apparatus (S14).

  Next, the selection unit 113 inserts the result data not selected in the arbitration into the delay unit 114 (S15). At this time, the selection unit 113 inserts the result data from the old processing unit 101 into the delay unit 114 so that the result data is older than the result data from the new processing unit 101. In this case, the “older position” is a position taken out first by the selection unit 113. Further, the selection unit 113 inserts the other result data into the delay unit 114 so as to arrive at a position where the result data is extracted in the order of arrival.

  Then, the selection unit 113 repeats the operations from S12 to S15 in FIG.

  Above, description of operation | movement of the network part 111 is complete | finished.

  As described above, it should be noted that the time-series data processing apparatus 100 does not immediately change the output result data to the new process after receiving the process change instruction. For example, the time-series data processing device 100 continues outputting the result data of the old process until the old process can be stopped even after receiving the process change instruction. For example, when the old process is in progress, the old process is not in a state where it can be stopped. Further, when a series of data by the old process is being output sequentially, the old process is not in a state where it can be stopped. When the old process can be stopped, the time-series data processing apparatus 100 outputs the result data string starting from the result data of the first new process after outputting the result data of the last old process. Become.

  Next, effects of the first exemplary embodiment of the present invention will be described.

  The time-series data processing apparatus according to the first embodiment of the present invention can specify result data before and after switching between old and new output according to a change in processing contents in time-series data processing, and result data by each process of old and new Output more useful stuff. At that time, the time-series data processing apparatus does not mix the output of the result data by the old and new processes while suppressing an increase in the amount of information added to the process result data.

  The reason is that the control unit adds additional information including new / old identification information indicating the old process and node identification information indicating a meeting place of the new process result data to the result data of the last old process. Furthermore, the control unit adds additional information including new / old identification information indicating the new process and node identification information indicating a meeting place of the result data of the old process to the result data of the first new process. Then, the selection unit stores selection information for selecting result data output from the node. Further, when the selection unit selects either the arrived result data or the result data stored in the delay unit, the selection unit is based on the additional information added to each result data and the stored selection information. This is because the last old process result data is selected before the first new process result data is selected.

  Specifically, for example, the selection unit stores that the result data of the last old process has not been passed as selection information indicating whether or not the result data has been passed. Then, the selection unit selects selection information for the arrival result data or the result data stored in the delay unit, for which the added node identification information indicates the own node and the old and new identification information indicates the new process. This is because this is not selected unless the mark indicates that it has passed. In addition, the selection unit selects one of the other result data according to some priority, the node identification information added to the selected result data indicates the own node, and the old and new identification information is old. This is because when the process is indicated, the selection information is updated so as to indicate passing. This is because the selection unit outputs the selected result data from the transmission unit to the next network unit or the outside, and inserts the unselected result data into the delay unit.

  As a result, in the present embodiment, after the process change instruction, the result data of the last specified old process is output to the outside, and the result sequence of the new process starting with the result data of the first specified new process is output. Can be output outside the device. That is, in the present embodiment, before and after the process change, the result data of the old process and the result data of the new process are not mixed and output to the outside of the apparatus. Also, the present embodiment outputs more useful and usable old process result data until the old process can be stopped even after the new process starts to operate in response to the process change instruction. It will be.

  Moreover, the present embodiment can greatly reduce the amount of information that needs to be added to the result data by providing the effects as described above as compared with the related art. For example, in the combination of the related technologies described in Patent Documents 5 and 6, it is necessary to add at least the scheduled output time to any result data. This is because the amount of information of the old and new identification information and the node identification information is very small compared to the amount of information for representing such time information.

(Second Embodiment)
Hereinafter, a second embodiment of the present invention will be described in detail with reference to the drawings. Note that the first embodiment of the present invention assumes that one type of new / old change process proceeds in the apparatus, and is not related to the simultaneous progress of other new / old change processes and the process to be changed. It was effective for cases where processing was not considered. In the present embodiment, an example will be described in which one or more types of processing are simultaneously operated in the apparatus, and the old and new change processing is simultaneously performed for one or more types of processing.

  The configuration of a time-series data processing device 200 as the second embodiment of the present invention is shown in FIG. In FIG. 4, the time-series data processing device 200 is different from the time-series data processing device 100 according to the first embodiment of the present invention in place of the network unit 111 and the network unit 211 and the control unit 121. The point provided with the control part 221 differs. The network unit 211 is different from the network unit 111 according to the first embodiment of the present invention in that the network unit 211 includes a selection unit 213 instead of the selection unit 113.

  In FIG. 4, time series data is input to each processing unit 101 from the outside. Further, data output from each processing unit 101 is input to one of the network units 211. Each network unit 211 receives data output from the processing unit 101 or another network unit 211. Further, data output from each network unit 211 is input to another network unit 211 or output outside the apparatus. As described above, the time-series data processing device 200 is configured to bundle the result data output from each processing unit 101 into one by a network including one or more network units 211 and then output the result data to the outside of the device. It has become. In FIG. 4, two processing units 101, one network unit 211, and one control unit 221 are shown. However, the number of units included in the time-series data processing device of the present invention is limited. is not.

  When the last input data necessary for calculating the result data of the last old process is input, the control unit 221 transmits the next additional information to the processing unit 101 of the old process. Specifically, at this timing, the control unit 221 instructs the processing unit 101 of the old process to identify the type of processing to be changed between old and new from irrelevant processing, and new and old identification information representing the old processing, The additional information including the node identification information indicating the meeting node is transmitted. Thereby, the control unit 221 causes the processing unit 101 of the old process to output the result data of the last old process to which the process identification information, the old and new identification information, and the node identification information are added.

  In addition, when the last input data necessary for calculating the result data of the first new process is input, the control unit 221 transmits the next additional information to the processing unit 101 of the new process. Specifically, at this timing, the control unit 221 instructs the processing unit 101 of the new process to have the same process identification information as the corresponding old process, new / old identification information indicating the new process, and node identification indicating the waiting node. Additional information including information is transmitted. Thereby, the control unit 221 causes the processing unit 101 of the new process to output the result data of the first new process to which the process identification information, the old and new identification information, and the node identification information are added.

  In addition, the control unit 221 receives the last input data necessary for calculating the result data of the first new process and immediately before the change of the new and old processes is completed, Then, the following additional information is transmitted. Specifically, at this timing, the control unit 221 instructs the processing unit 101 of the new process to have the same process identification information as the corresponding old process, new / old identification information indicating the new process, and node identification indicating that no waiting is required. Additional information including information is transmitted. Note that the completion of the change between the new and old processes means that there is no result data of the first new process in the apparatus.

  In addition, at the timing after the change between the old and new processing is completed, the control unit 221 instructs the processing unit 101 of the new processing to process identification information indicating that it is not a new / old change target, new / old identification information indicating new processing, Additional information including node identification information indicating that no waiting is required is transmitted.

  Here, the process identification information is information that can identify the type of process to be changed between old and new from unrelated processes. That is, the process identification information is set to the same value for the old process and the new process to be changed. Further, the process identification information only needs to be able to identify the type of process to be changed between old and new and an irrelevant process not to be changed. That is, the process identification information is set to the same value for each irrelevant process. As described above, the process identification information is used to distinguish the type of process to be changed between old and new from irrelevant processes.

  The node identification information is information for identifying the network unit 211 that waits for the result data of the new process and the result data of the old process, similarly to the node identification information in the first embodiment of the present invention. In addition, the node identification information may indicate that no waiting is required.

  The data structure of information output from each processing unit 101 is shown in FIG. Hereinafter, result data to which additional information including process identification information, old and new identification information, and node identification information is added may be simply referred to as “result data”. In the time-series data processing apparatus 200, the result data is input / output or held by each functional block together with various additional information added.

  Further, the control unit 221 indicates the timing at which the last input data necessary for calculating the result data of the last old process is input, information indicating the processing status of the processing unit 101 of the old process or the processing unit 101 of the new process, Or you may determine based on the information showing the output condition of result data.

  Further, the control unit 221 indicates the timing at which the last input data necessary for calculation of the result data of the first new process is input, information indicating the processing status of the processing unit 101 of the old process or the processing unit 101 of the new process, Or you may determine based on the information showing the output condition of result data.

  Note that the processing status of the processing unit 101 refers to, for example, a status indicating whether a series of processing is in progress or a series of processing is completed in the processing unit 101 of the old process. Alternatively, the processing status of the processing unit 101 may be a status such as after a fixed cycle after the start of a new process. The output status of the result data refers to a status such as whether a series of result data is being output from the processing unit 101 of the old process or whether the output has been completed. Moreover, the control part 221 may determine such timing based on the combination of the information mentioned above and other information. Further, the control unit 221 may perform the determination process by notifying the processing status and output status of the processing unit 101 from the processing unit 101.

  In addition, the control unit 221 waits for each irrelevant processing unit 101 (not shown) that is not the old / new change target, the process identification information indicating that it is not the old / new change target, the new / old identification information indicating the new process, Additional information including node identification information indicating unnecessary is transmitted. Regarding the result data from the irrelevant processing unit 101, node identification information is mainly referred to by the selection unit 213 described later, and the old and new identification information and the processing identification information are not referred to. Therefore, the old and new identification information transmitted to the irrelevant processing unit 101 does not necessarily represent the new process, and may be information of any content. Further, the process identification information transmitted to the irrelevant processing unit 101 may have any content as long as it is information other than the information indicating the process to be changed.

  As the selection information for selecting the result data to be output from the own node, the selection unit 213 determines whether the result data of the last old process has passed through the own node for each type of process to be changed. Information indicating whether or not is stored. Specifically, the selection unit 213 holds a table having the process identification information as an index as selection information, and indicates whether or not the result data of the last old process has been passed in association with the process identification information. Information may be stored. Hereinafter, a table (selection information) using such processing identification information as an index is also referred to as a passage completion table.

  In addition, the selection unit 213 arbitrates the right to use the output port between the result data arriving at the transmission unit 112 of the own node and the oldest result data stored in the delay unit 114.

  Specifically, the selection unit 213 indicates that for the result data in which the node identification information indicates the own node and the old and new identification information indicates the new process, the entry in the passage completion table corresponding to the process identification information has not been passed. Is not selected, the result data is not selected. Further, the selection unit 213 indicates the result of the first new process in which the node identification information indicates that no waiting is required and the process identification information identical to the process identification information is added to the result data in which the old and new identification information indicates the new process. When data is stored in the delay unit 114, this result data is not selected. The selection unit 213 selects one of the other result data so that the output order of the results output from the same processing unit 101 is not changed and based on some priority. The “other result data” includes result data in which the node identification information indicates the own node and the old and new identification information indicates the new process, and an entry in the passage completion table corresponding to the process identification information is included. Contains result data indicating passed. The “other result data” is the result data in which the node identification information indicates that no waiting is required and the old and new identification information indicates a new process, and the same process identification information as the process identification information is added. The result data in which the result data of the first new process is not stored in the delay unit 114 is included. Further, “other result data” includes result data in which the node identification information does not indicate the own node or the old and new identification information indicates the old process.

  In addition, when the node identification information of the selected result data indicates the own node and the new and old identification information indicates the old process, the selection unit 213 displays an entry in the passage completion table corresponding to the process identification information of the result data, Update to indicate passed.

  The selection unit 213 also controls the transmission unit 112 to output the selected result data to the next network unit 211 or the outside.

  The selection unit 213 inserts the result data that has not been selected into the delay unit 114 and delays the result data. At this time, when the inserted result data includes a plurality of result data having the same process identification information, the selection unit 213 delays the old process result data so that the result data of the old process is older than the result data of the new process. What is necessary is just to insert in 114. As described above, the “older position” is a position taken out first by the selection unit 213. In addition, the selection unit 213 inserts the other result data into the delay unit 114 so as to arrive at a position where the result data is extracted in the order of arrival.

  The operation of the time-series data processing device 200 configured as described above will be described with reference to FIGS.

  In the following description, it is assumed that the time-series data processing device 200 is configured by an IC chip having a plurality of areas (partitions). In each partition, setting information called configuration data and configuration data is set to create a circuit for executing user-defined processing.

  In addition, the time-series data processing device 200 starts the following operation when receiving an instruction to change the process from the old process to the new process from the outside. Note that the time-series data processing device 200 may receive such a change instruction from a host computer or the like connected to the outside. In addition, it is assumed that user-defined information related to the new process to be changed is written to an arbitrary processing unit 101 configured by a partition different from the processing unit 101 that is executing the old process.

  First, the operation of the control unit 221 is shown in FIG.

  In FIG. 6, when the control unit 221 determines that it is the timing at which the last data necessary for calculation of the result data of the last old process is input, the control unit 221 converts the result data to the processing unit 101 of the old process. Additional information for adding is transmitted. Specifically, the control unit 221 sends processing identification information for identifying a new / old change target process, new / old identification information representing an old process, and node identification information indicating a waiting node to the processing unit 101 of the old process. Transmit (S21).

  As described above, the control unit 221 performs such timing determination based on information on the processing status of the processing unit 101 of the old process or the processing unit 101 of the new process or information on the output status. Also good.

  Further, at the timing after S21 in FIG. 6, the control unit 221 continues to instruct the processing unit 101 of the old process not to output the result data.

  Next, when the control unit 221 determines that it is the timing when the last data necessary for calculation of the result data of the first new process is input, the control unit 221 adds the result data to the processing unit 101 of the new process. Send additional information. Specifically, the control unit 221 sends the corresponding process identification information common to the old process, the new and old identification information indicating the new process, and the node identification information indicating the waiting node to the processing unit 101 of the new process. Transmit (S22).

  Note that at a timing after this timing and immediately before the change between the new and old processes is completed, the control unit 221 transmits the next additional information to be added to the result data to the processing unit 101 of the new process. . Specifically, the control unit 221 transmits corresponding processing identification information common to the old process, new / old identification information indicating the new process, and node identification information indicating that no waiting is required, to the processing unit 101 of the new process. To do.

  In addition, at the timing after the change between the old and new processing is completed, the control unit 221 instructs the processing unit 101 of the new processing to process identification information indicating that it is not a new / old change target, new / old identification information indicating new processing, Additional information including node identification information indicating that no waiting is required is transmitted.

  Further, as described above, the control unit 221 performs such timing determination based on information regarding the processing status of the processing unit 101 of the old process or the processing unit 101 of the new process or information regarding the output status. Also good.

  Further, at the timing before S22 in FIG. 6, the control unit 221 continues to instruct the processing unit 101 of the new process not to output the result data.

  Next, the control unit 221 transmits the next additional information to be added to the result data to each irrelevant processing unit 101 that is not a target for new and old changes. Specifically, the control unit 221 transmits, to each of these processing units 101, processing identification information indicating that it is not a new / old change target, new / old identification information indicating new processing, and node identification information indicating that no waiting is required. (S23).

  Note that the control unit 221 may use a value instructed by the host computer as the value of various types of identification information transmitted to each processing unit 101.

  Above, description of operation | movement of the control part 221 is complete | finished.

  Next, the operation of the selection unit 213 is shown in FIG.

  In FIG. 7, first, the selection unit 213 sets and initializes the entry in the passage completion table corresponding to the process identification information to be changed to information indicating that the passage has not been completed (for example, “N”) (S31). ). Note that the selection unit 213 determines, for each process identification information that identifies the type of process to be changed between old and new, at the timing when the result data of the first new process identified by the process identification information no longer exists in the apparatus. What is necessary is just to initialize the entry of the passage completion table corresponding to process identification information.

  Next, the selection unit 213 arbitrates the right to use the output port between the result data arriving at the transmission unit 112 of its own node and the oldest result data stored in the delay unit 114. Specifically, for the result data, the selection unit 213 indicates that the node identification information indicates the own node, the old and new identification information indicates the new process, and the passage completion table corresponding to the process identification information. If the entry indicates that the entry has not been passed, the selection unit 213 determines not to select the result data. The selection unit 213 selects any other result data so that the output order of the results output from the same processing unit 101 is not changed and according to some priority (S32).

Next, when the node identification information indicates the own node and the new and old identification information indicates the old process for the result data selected by the arbitration, the selection unit 213 passes corresponding to the process identification information. The entry in the completion table is updated to information (for example, “Y”) indicating that it has been passed (S33).
Next, the selection unit 213 outputs the result data selected by the arbitration from the transmission unit 112 to the next network unit 211 or an output port to the outside of the apparatus (S34).

  Next, the selection unit 213 inserts the result data not selected in the arbitration into the delay unit 114 (S35). At this time, when the result data to be inserted includes a plurality of result data having the same process identification information, the selection unit 213 places the result data of the old process at a position older than the result data of the new process. It inserts into the delay part 114 so that it may start. In addition, the selection unit 213 inserts the other result data into the delay unit 114 so as to arrive at a position where the result data is extracted in the order of arrival.

  Then, the selection unit 213 repeats the operations from S32 to S35 in FIG.

  Above, description of operation | movement of the selection part 213 is complete | finished.

(Specific example of operation)
Next, a specific example of the operation of the time-series data processing device 200 will be described with reference to FIGS.

  In the specific example described below, the time-series data processing device 200 includes three processing units 101 and two network units 211. Of the three processing units 101, the processing unit 101-1 defines an old process, the processing unit 101-2 defines a new process, and the processing unit 101-3 includes other processes not to be changed. Is defined.

  Specifically, the processing unit 101-1 defines an old process for calculating the product of the latest two input data for time series data. In the processing unit 101-2, a new process for calculating the sum of the latest two input data is newly defined. In the processing unit 101-3, a process for calculating the difference between the two most recent input data is defined as a process unrelated to the old process and the new process.

  Further, the old process by the processing unit 101-1 requires a calculation delay of 2 cycles, the new process by the processing unit 101-2 requires a calculation delay of 1 cycle, and the process by the processing unit 101-3 takes 1 It is assumed that a cycle calculation delay is required. Further, it is assumed that process identification information “1” is set for these old process and new process, and process identification information “3” is set for the process by the processing unit 101-3.

  Of the two network units 211, the data output from the processing units 101-1 and 101-3 is input to the network unit 211-1, and the processing units 101-2 and 101-2 are input to the network unit 211-2. Data output from the network unit 211-1 is input. The network unit 211-2 outputs data to an output port to the outside of the apparatus. The network unit 211-1 is identified by the node identification information “1”, and the network unit 211-2 is identified by the node identification information “2”.

  The network unit 211-1 includes a transmission unit 112-1, a selection unit 213-1, and a delay unit 114-1, and the network unit 211-2 includes a transmission unit 112-2, a selection unit 213-2, and a delay. It shall have the part 114-2.

  Further, in this specific example, it is assumed that the number of types of processes to be changed simultaneously is limited to one.

  In this specific example, when circuit information for executing a new process is written to the processing unit 101-2, the processing unit 101-2 starts the operation of the new process. Then, after the processing unit 101-2 starts the operation of the new processing, when the old processing can be stopped in the processing unit 101-1 of the old processing, the processing unit 101-1 stops the operation, and the processing unit 101-1 It is assumed that circuit information is deleted.

  Here, it is assumed that the timing at which the old process can be stopped is given from the processing unit 101-1 of the old process to the control unit 221. The timing is assumed to be several cycles after the writing of the circuit information for executing the new process to the processing unit 101-2 is completed and the new process starts outputting the result. It is also assumed that the timing at which the old process outputs the last result data is the same. In other words, the result of the old process can be used for several cycles even after the new process starts outputting the result data. This example makes it possible to output the result of this available old process.

  Next, details of a specific operation of the time-series data processing device 200 will be described.

  Here, it is assumed that v1 is given as input data to the time-series data processing device 200 in the first cycle, v2 in the second cycle, and v3 in the third cycle. Input data is given at the start of each cycle. These input data are assumed to be real numbers. Further, the result data of the old process using v1 and v2 is the result data of the last old process, and the result data of the new process using v2 and v3 is the result data of the first new process.

  First, the selection unit 213-1 and the selection unit 213-2 set the entry of the passage completion table to “N” indicating that the passage has not been completed in association with the process identification information “1” to be changed (S31). ). Thus, each selection unit 213 initializes selection information corresponding to the process identification information to be changed before the process change is started.

(First cycle)
In the first cycle, the processing unit 101-1 executes the old process.

  Further, data “v1” is input to the time-series data processing device 200, and the processing unit 101-1 holds this data.

(Second cycle)
In the second cycle, a new process is written from the outside into an area corresponding to the processing unit 101-2. Then, the processing unit 101-2 starts the operation of the new process.

  In addition, data “v2” is input to the time-series data processing device 200. Here, the data “v2” is the last input data necessary for calculating the result data of the last old process.

  Therefore, the control unit 221 uses the process identification information “1” common to the old process and the new process, the old and new identification information “o” representing the old process, and the node identification information “2” as the processing unit 101 of the old process. -1 (S21). The network unit 211-2 indicated by the node identification information “2” indicates a place where result data from the processing unit 101-1 of the old process and the processing unit 101-2 of the new process merge. Hereinafter, information transmitted from the control unit 221 to each processing unit 101 is represented as <processing identification information, old / new identification information, node identification information>. That is, the control unit 221 transmits <1, o, 2> to the processing unit 101-1.

  FIG. 8 shows output information inside the time-series data processing device 200 at the end of the second cycle.

  Here, as described above, in this specific example, the number of types of processes that change simultaneously is one. Therefore, the process identification information only needs to at least indicate whether the process is a process to be changed, and can be expressed by 1 bit. The node identification information for identifying the two network units 211-1 and 211-2 can be represented by 1 bit. In addition, since the new and old identification information can be represented by 1 bit, the information added to the result data by the control unit 221 can be represented by 3 bits in total.

(3rd cycle)
In the third cycle, data “v3” is input to the time-series data processing device 200. Here, the data “v3” is the last input data necessary for calculating the result data of the first new process.

  Therefore, the control unit 221 includes information <1, n, which includes process identification information “1” common to the old process and the new process, old and new identification information “n” representing the new process, and node identification information “2”. 2> is transmitted to the processing unit 101-2 of the new process.

  Further, the processing unit 101-2 completes calculation (new processing) of the sum of “v2” and “v3” because the calculation delay is one cycle. Then, the processing unit 101-2 adds <v2 + v3, 1, n, 2> to which the information <1, n, 2> received from the control unit 221 is added to the result data “v2 + v3”. The data is output to the network unit 211-2 (S22). In this way, information output from the processing unit 101 is represented as <result data, process identification information, old and new identification information, node identification information>.

  Note that the control unit 221 continues to instruct the processing unit 101-2 not to output in the previous cycle.

  Further, since the calculation delay is two cycles, the processing unit 101-1 completes the calculation of the product of “v1” and “v2” (old process). Then, the processing unit 101-1 adds <v1 * v2,1, o, 2> to which the information <1, o, 2> received from the control unit 221 is added to the result data “v1 * v2”. The data is output to the network unit 211-1 (S21).

  Note that the control unit 221 continues to instruct the processing unit 101-1 not to perform output in a cycle after this.

  Further, the control unit 221 includes, for the processing unit 101-3, <3, n, −> including processing identification information “3”, “n” indicating new processing, and “-” indicating that no waiting is required. Send.

  Then, the processing unit 101-3 completes the calculation of the difference between “v2” and “v3” because the calculation delay is one cycle. Then, the processing unit 101-3 adds <v2-v3,3, n,-> to which the information <3, n,-> received from the control unit 221 is added to the result data “v2-v3”. The data is output to the network unit 211-1 (S23).

  FIG. 9 shows output information inside the time-series data processing device 200 at the end of the third cycle.

(4th cycle)
In the fourth cycle, the selection unit 213-1 performs arbitration of the right to use the output port for the input <v1 * v2,1, o, 2> and <v2-v3,3, n,->. Here, since the node identification information added to the former result data <v1 * v2,1, o, 2> does not represent the own node, the selection unit 213-1 refers to the passage completion table for the result data. Without being a candidate for output. Further, since the node identification information added to the latter result data <v2-v3,3, n,-> indicates that no waiting is required, the selection unit 213-1 also refers to the passage completion table for this result data. Without being a candidate for output. Here, it is assumed that the priority of the latter result data is high in <v1 * v2,1, o, 2> and <v2-v3,3, n,->. Therefore, the selection unit 213-1 selects <v2-v3,3, n,-> as an output target (S32).

  Further, the selection unit 213-1 does not update the passage completion table because the node identification information of the selected data <v2-v3,3, n,-> does not represent the own node.

  The selection unit 213-1 outputs the selected <v2-v3,3, n,-> to the network unit 211-2 (S34).

  In addition, the selection unit 213-1 inserts <v1 * v2,1, o, 2> that has not been selected into the delay unit 114-1 (S35).

  In addition, the selection unit 213-2 mediates the right to use the output port for the input <v2 + v3,1, n, 2>. The node identification information added to the result data represents the own node, and the old and new identification information represents the new process. Therefore, the selection unit 213-2 refers to the entry of the passage completion table corresponding to the process identification information “1”, and does not select the result data because the value is “N” (S32).

  Then, the selection unit 213-2 does not update the passage completion table because there is no selected result data.

  Then, since there is no selected result data, the selection unit 213-2 does not output to the outside of the apparatus, and does not select the result data <v2 + v3,1, n, 2> that is not selected. 2 is inserted (S35).

  FIG. 10 shows output information inside the time-series data processing device 200 at the end of the fourth cycle.

(5th cycle)
In the fifth cycle, the selection unit 213-1 arbitrates the right to use the output port for <v1 * v2,1, o, 2> stored in the delay unit 114-1. Since the node identification information added to the result data does not represent the own node, the selection unit 213-1 selects the result data without referring to the passage completion table (S32).

  Further, the selection unit 213-1 does not update the passage completion table because the node identification information added to the result data does not represent the own node.

  Then, the selection unit 213-1 outputs the result data <v1 * v2,1, o, 2> to the network unit 211-2 (S34).

  The selection unit 213-2 outputs the input <v2-v3,3, n,-> and <v2 + v3,1, n, 2> stored in the delay unit 114-1. Mediates port usage rights. Here, in the latter <v2 + v3,1, n, 2>, the node identification information represents the own node, and the old and new identification information represents the new process. Therefore, the selection unit 213-2 refers to the entry of the passage completion table corresponding to the process identification information “1”, and does not select the result data because the value is “N” (S32).

  In addition, since the former <v2-v3,3, n,-> node identification information indicates that no waiting is required, the selection unit 213-2 selects this result data without referring to the passage completion table (S32). ).

  Further, the selection unit 213-2 does not update the passage completion table because the node identification information of the selected data <v2-v3,3, n,-> does not represent the own node.

  Then, the selection unit 213-2 outputs the result data <v2-v3,3, n,-> to the outside of the apparatus (S34).

  In addition, the selection unit 213-2 inserts the unselected data <v2 + v3,1, n, 2> into the delay unit 114-2 (S35).

  Output information inside the time-series data processing device 200 at the end of the fifth cycle is shown in FIG.

(6th cycle)
In the sixth cycle, the selection unit 213-2 receives the input <v1 * v2, 1, o, 2> and <v2 + v3,1, n, 2> stored in the delay unit 114-2. For the output port.

  Here, the node identification information of the latter result data <v2 + v3,1, n, 2> represents the own node, and the old and new identification information represents a new process. Therefore, the selection unit 213-2 refers to the entry of the passage completion table corresponding to the process identification information “1”, and does not select the result data because the value is “N” (S32).

  In addition, the selection unit 213-2 obtains this result because the node identification information of the former result data <v1 * v2,1, o, 2> represents the own node, and the old and new identification information represents the old process. Data is selected (S32).

  Further, the selection unit 213-2 selects the process identification information because the selected node identification information of <v1 * v2,1, o, 2> represents the own node and the old and new identification information represents the old process. The entry in the passage completion table corresponding to “1” is updated to “Y” (S33).

  Then, the selecting unit 213-2 outputs the result data <v1 * v2,1, o, 2> to the outside of the apparatus (S34).

  Further, the selection unit 213-2 inserts <v2 + v3,1, n, 2> that has not been selected into the delay unit 114-2 (S35).

  FIG. 12 shows output information inside the time-series data processing device 200 at the end of the sixth cycle.

(7th cycle)
In the seventh cycle, the selection unit 213-2 mediates the right to use the output port for <v2 + v3,1, n, 2> stored in the delay unit 114-2.

  Here, the node identification information of <v2 + v3,1, n, 2> represents the own node, and the old and new identification information represents a new process. Therefore, the selection unit 213-2 refers to the entry of the passage completion table corresponding to the process identification information “1”, and selects the result data because the value is “Y” (S32).

  Then, the selection unit 213-2 outputs the result data <v2 + v3,1, n, 2> to the outside of the apparatus (S34).

  Output information inside the time-series data processing device 200 at the end of the seventh cycle is shown in FIG.

  Above, description of an example of the specific operation | movement of the time series data processing apparatus 200 is complete | finished.

  In this specific example, the time-series data processing device 200 is arranged in the order of increasing time, <v2-v3,3, n,->, <v1 * v2, 1, o, 2>, <v2 + v3, 1, n. , 2> result data was output. As described above, the time-series data processing device 200 includes the specific result data <v2 + v3, 1, n, new process after the specific result data <v1 * v2, 1, o, 2> of the old process. 2> can be output.

  The time-series data processing device 200 also provides useful and usable old process result data <v1 * v2 after the second cycle in which a new process is defined for the processing unit 101-2 in accordance with a process change instruction. , 1, o, 2> could be output to the outside of the apparatus in the sixth cycle.

  In contrast, for example, in the related art described above, the old process is stopped at the time of writing the new process. Alternatively, Patent Documents 1 to 4 described above do not describe outputting the result data of the old process that is useful and usable even after the start of the operation of the new process. Therefore, this embodiment can output more useful and usable old process result data as compared with the related art.

  Next, the effect of the second exemplary embodiment of the present invention will be described.

  The time-series data processing apparatus according to the second embodiment of the present invention can specify the result data before and after the output switching between the new and old processes in accordance with the change of the processing contents in the time-series data processing. Output more useful stuff. At that time, the time-series data processing apparatus does not mix the output of the result data by the old and new processes while suppressing an increase in the amount of information added to the process result data.

  The reason is that the control unit adds process identification information for identifying the type of process, new / old identification information indicating the old process, and waiting node identification information to the result data of the last old process. At the same time, the control unit adds the same process identification information as the old process, the new and old identification information indicating the new process, and the node identification information for waiting for the result data of the old process to the result data of the first new process. It is. And the selection part in the node designated as a meeting place memorize | stores the information showing that it has not passed by matching with the process identification information of change object. If the node identification information of the result data arriving at the own node or the result data stored in the delay unit indicates the own node and the new and old identification information indicates a new process, the selection unit indicates the process identification information. This is because the selection information corresponding to is not selected unless it has been passed. Further, the selection unit selects and outputs one of the other result data according to the priority. In addition, when the node identification information of the selected result data indicates the own node and the new and old identification information indicates the old process, the selection unit updates the selection information corresponding to the process identification information to “passed”. It is.

  As a result, after the processing change instruction, this embodiment outputs the result data string of the old process until the specified result of the old process is externally output until the old process can be stopped. The new process result string starting from the designated first new process result data is output to the outside of the apparatus. That is, before and after the process change, the result data of the old process and the new process are not mixed and output outside the apparatus. Also, the present embodiment outputs more useful and usable result data of the old process even after the new process starts operating in accordance with the process change instruction.

  In addition, this embodiment can greatly reduce the amount of information that needs to be added to the result data by producing the above-described effect when changing from the old process to the new process. The reason will be described.

Here, N is the number of processes that operate simultaneously in the apparatus, and n is the number of processes in which new and old change processes proceed simultaneously. Further, for simplicity of explanation, N and n + 1 are assumed to be powers of 2, respectively. Then, among the information added to the result data, the process identification information only needs to be able to identify n types of new and old change processes that are simultaneously progressing and processes that are not the target of the old and new change processes. Therefore, the process identification information can be represented by log ₂ (n + 1) bits. Log ₂ (X) represents the logarithm of X with 2 as the base. Further, since the new / old identification information only needs to be able to distinguish between the new process and the old process, it can be represented by 1 bit. In addition, when the network configured by the network unit is a binary tree, the node identification information may represent the depth of the network. The depth of the network is represented by log ₂ (N). Therefore, the node identification information can be represented by ceil (log ₂ (log ₂ (N) +1)) bits. Here, ceil (X) represents the smallest integer not lower than X. The number of edges of the network part is (2N−1). Therefore, the number of wires added to the time-series data processing apparatus of the present embodiment in order to cope with the change from the old process to the new process is expressed by the following equation (1).

(2N-1) * (log ₂ (n + 1) + 1 + ceil (log ₂ (log ₂ (N) +1)))
... (1)
Furthermore, the selection information stored in each selection unit is n bits because the number of processes for which the old and new change processes are simultaneously required is required. In addition, since the number of processes simultaneously operating in the apparatus is N, if the network is a binary tree, the number of network units is N-1. Therefore, the number of selection units is N-1. Therefore, the number of storage elements added to the selection unit in the present embodiment in order to cope with the change from the old process to the new process is expressed by the following equation (2).

(N-1) * n (2)
The delay unit stores result data to which various information is added. At this time, there is a possibility that a plurality of old and new processing pairs join at the same time in the delay unit of the same node. For this reason, at least a storage element corresponding to information added to n result data is added to the delay unit in order to cope with a change from the old process to the new process. As described above, since the number of network units is N-1, the number of delay units is N-1. Therefore, the minimum number of storage elements added to the delay unit in the present embodiment in order to cope with the change from the old process to the new process is expressed by the following equation (3).

(N−1) * n * (log ₂ (n + 1) + 1 + ceil (log ₂ (log ₂ (N) +1))) (3)
On the other hand, a time-series data processing device that combines related techniques described in Patent Documents 5 to 6 includes information for identifying processing and a network node that performs waiting (synchronization) in the output data of each processing unit. It is necessary to add information to be identified and scheduled output time. Here, if the number of processes is N and the number of simultaneous progresses of new and old process changes is n, the information for identifying the process can be represented by log ₂ (n + 1) bits. Further, the identification information of the nodes of the network can be represented by ceil (log ₂ (log ₂ (N) +1)) bits if the network is a binary tree. Further, the scheduled output time can be represented by 64 bits, for example. The number of edges of the network part is (2N−1) as described above. Therefore, the number of wires added in the related art to cope with the change from the old process to the new process is expressed by the following equation (4).

(2N−1) * (log ₂ (n + 1) + 64 + ceil (log ₂ (log ₂ (N) +1))) (4)
Further, the result data with various information added is stored in the delay section of the related art. Considering the possibility of multiple old and new processing pairs joining at the same node at the same time, at least the storage elements for the information added to the n result data are added to the delay unit to cope with the change from the old process to the new process. Added. Here, since the number of processes is N, if the network is a binary tree, the number of delay units is N-1. For this reason, the minimum number of storage elements added to the delay unit in the related art in order to cope with the change from the old process to the new process is expressed by the following equation (5).

(N-1) * n * (log ₂ (n + 1) + 64 + ceil (log ₂ (log ₂ (N) +1))) (5)
Here, for example, the number N of processes simultaneously operating in the apparatus is 128. Further, the number n of old and new change processes that proceed simultaneously in the apparatus is limited to one. Then, in the present embodiment, the number of wires that need to be added to the entire apparatus is calculated as follows using Equation (1).
(2N-1) * (log ₂ (n + 1) + 1 + ceil (log ₂ (log ₂ (N) +1)))
= 1275 (N = 128, n = 1)
On the other hand, in the combination of related technologies, the number of wires that need to be added to the entire apparatus is calculated as follows using Equation (4).
(2N-1) * (log ₂ (n + 1) + 64 + ceil (log ₂ (log ₂ (N) +1)))
= 17340 (N = 128, n = 1)
Further, in the present embodiment, the number of storage elements added as a whole apparatus is calculated as follows using Expression (2) and Expression (3).
(N−1) * n + (N−1) * n * (log ₂ (n + 1) + 1 + ceil (log ₂ (log ₂ (N) +1))) = 762 (N = 128, n = 1)
On the other hand, in the combination of related technologies, the number of storage elements added as a whole apparatus is calculated as follows using Equation (5).
(N−1) * n * (log ₂ (n + 1) + 64 + ceil (log ₂ (log ₂ (N) +1))) = 8636 (N = 128, n = 1)
This comparison is shown in FIG. As described above, the time-series data processing apparatus according to the present embodiment can reduce the amount of information added to the result data as compared with the related technology, and as a result, the additional resources necessary for the entire apparatus can be greatly reduced. doing.

  In the above-described second embodiment of the present invention, a specific example has been described in which the number of types of new and old change processes that proceed simultaneously in the apparatus is limited to one. However, a plurality of types of processes in the apparatus are old and new. The change process may proceed simultaneously.

  Further, in each of the above-described embodiments of the present invention, the example in which the time-series data processing device of the present invention is configured by hardware having reconfigurable circuit elements such as an FPGA has been mainly described. In addition, the time-series data processing device stores the operation of each unit as a computer program of the present invention in a storage device (storage medium) of the computer device so that the CPU reads and executes the computer program. It may be. In such a case, the present invention is constituted by the code of the computer program or a storage medium.

  Moreover, each embodiment mentioned above can be implemented in combination as appropriate.

While the present invention has been described with reference to the embodiments, the present invention is not limited to the above embodiments. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.
This application claims the priority on the basis of Japanese application Japanese Patent Application No. 2013-150426 for which it applied on July 19, 2013, and takes in those the indications of all here.

100, 200 Time-series data processing device 101 Processing unit 111, 211 Network unit 112 Transmission unit 113, 213 Selection unit 114 Delay unit 121, 221 Control unit 600, 700 Time-series data processing device 601, 701 Processing unit 611 Control unit 621 Selection Unit 631 synchronization unit 701 processing unit 731 network unit 732 transmission unit 733 time comparison unit 734 delay unit

The present invention relates to a technique for processing time-series data, and more particularly to a time-series data processing apparatus, a time-series data processing method, and a program in which the content of processing on time-series data changes.

In addition, the program of the present invention provides a change instruction from the first process that outputs the result data of the first process based on the input data to the second process that outputs the result data of the second process based on the input data. Of the first process result data and the result data determined to be output to the outside of the apparatus last (result data of the last old process), either the old process or the new process Additional information including information indicating the old process as the old and new identification information indicating that it is, and node identification information indicating the node in the apparatus that waits for the result data of the old process and the result data of the new process, and Information indicating the new process as the new / old identification information for the result data (result data of the first new process) that is first determined to be output to the outside of the apparatus among the result data of process 2 And additional information including the node identification information, and at each of the nodes, selection information representing information for selecting result data to be output from the own node is stored, and the first process or When selecting one of the result data output from the second process and arriving at its own node and the result data held and outputting it to another node or outside, it is added to each result data. Based on the additional information and the stored selection information, the result data of the last old process is selected before selecting the result data of the first new process, and the result of not selecting The computer apparatus is caused to delay the data.

Claims (10)

A plurality of nodes comprising selection means and delay means;
First processing means for outputting result data of the first processing by executing the first processing on the input data;
Second processing means for outputting result data of the second processing by executing the second processing on the input data;
When the change instruction from the first process to the second process is received from the outside, the result data determined to be output to the outside of the apparatus at the end of the result data of the first process (last In the device that waits for old processing result data and new processing result data as old / new identification information indicating that it is either old processing or new processing. And additional information including node identification information indicating the current node, and among the result data by the second process, the result data determined to be output to the outside of the apparatus first (result of the first new process) Control means for adding additional information including information indicating new processing as the new and old identification information and the node identification information to the data),
The selection means stores selection information representing information for selecting result data to be output from the own node, and is output from the first processing means or the second processing means and arrives at the own node. When the result data selected from the result data and the result data held in the delay means are selected and output to another node or outside, the additional information added to each result data is stored. Based on the selection information and selecting the result data of the last old process before selecting the result data of the first new process, and inserting the result data not selected into the delay means,
The delay means is delayed by holding result data included in each node and not selected by the selection means,
Time-series data processing device.   The selection means stores, as the selection information, information indicating that the result data of the last old process has not passed through its own node, and the node identification information added to the selected result data is automatically stored. When the new and old identification information indicates the old process, the selection information is updated to indicate that it has passed, and the result data that has arrived and the result data held in the delay means are added. The result data is not selected when the node identification information indicates a self-node and the new and old identification information indicates a new process, and the selection information does not indicate the passed. The time-series data processing device according to 1. In addition to the new / old identification information and the node identification information, the control means sets the type of the new / old change target process to the old / old change target for the result data of the last old process and the result data of the first new process. Adding the additional information including process identification information for identifying from a process that is not,
The selection means stores the selection information for each processing identification information, and when the node identification information added to the selected result data indicates the own node, and the old and new identification information represents an old process, The selection information corresponding to the process identification information added to the result data is updated so as to indicate the passing, and the arrived result data and the result data held in the delay means are added to the node When the identification information indicates the own node and the old and new identification information indicates a new process, and the selection information corresponding to the process identification information added to the result data does not indicate the passed, 3. The time-series data processing apparatus according to claim 2, wherein no result data is selected.   When the last input data necessary for calculation of the result data of the last old process is input, the control means transmits the additional information to the first processing means, so that the additional information is When the added result data of the last old process is output to the first processing means and the last input data necessary for calculation of the result data of the first new process is input, the second data 2. The data processing apparatus according to claim 1, wherein the second processing unit outputs the result data of the first new process to which the additional information is added by transmitting the additional information to a processing unit. 4. The time-series data processing device according to any one of items 3.   The control means is information about the processing status of the first processing means or the second processing means, the timing at which the last input data necessary for calculating the result data of the last old processing is input, Alternatively, it is determined based on information related to the output status of the result data, and the timing at which the last input data necessary for calculating the result data of the first new process is input is determined by the first processing means or 5. The time-series data according to claim 1, wherein the time-series data is determined based on information on a processing status of the second processing unit or information on an output status of the result data. Processing equipment.   The control means transmits information instructing the first processing means not to output the result data after the output of the result data of the last old process, and to the second processing means. 6. The time-series data processing device according to claim 1, wherein information instructing not to output the result data is transmitted before the output of the result data of the first new process. . When receiving a change instruction from the first process that outputs the result data of the first process based on the input data to the second process that outputs the result data of the second process based on the input data,
Of the result data of the first process, the result data determined to be the last output to the outside of the apparatus (the result data of the last old process) is either the old process or the new process Additional information including information indicating the old process as the old and new identification information indicating, and node identification information indicating the node in the apparatus that waits for the result data of the old process and the result data of the new process,
Information indicating new processing as the new / old identification information with respect to result data (result data of the first new processing) that is first determined to be output to the outside of the apparatus among the result data of the second processing; , Adding additional information including the node identification information,
In each of the nodes, selection information representing information for selecting result data to be output from the own node is stored, and the result that is output by the first process or the second process and arrives at the own node When selecting one of the data and the result data to be held and outputting it to another node or the outside, the additional information added to each result data and the stored selection information A time-series data processing method that selects the result data of the last old process before selecting the result data of the first new process, and delays the result data not selected based on the result data.   Each node stores, as the selection information, information indicating that the result data of the last old process has not passed through its own node, and the node identification information added to the selected result data is automatically stored. When the new and old identification information indicates an old process, the selection information is updated so as to indicate passing, and the node identification added to the arrived result data and the retained result data is indicated. The time-series data processing according to claim 7, wherein when the information indicates a local node and the new / old identification information indicates a new process, the result data is not selected if the selection information does not indicate the passed. Method. When receiving a change instruction from the first process that outputs the result data of the first process based on the input data to the second process that outputs the result data of the second process based on the input data,
Of the result data of the first process, the result data determined to be the last output to the outside of the apparatus (the result data of the last old process) is either the old process or the new process Additional information including information indicating the old process as the old and new identification information indicating, and node identification information indicating the node in the apparatus that waits for the result data of the old process and the result data of the new process,
Information indicating new processing as the new / old identification information with respect to result data (result data of the first new processing) that is first determined to be output to the outside of the apparatus among the result data of the second processing; , Adding additional information including the node identification information,
In each of the nodes, selection information representing information for selecting result data to be output from the own node is stored,
When selecting any one of the result data output from the first process or the second process and arriving at its own node and the result data held and outputting it to another node or outside, Based on the additional information added to each result data and the selection information stored, the result data of the last old process is selected before the result data of the first new process is selected. , Delaying the result data not selected,
A computer-readable recording medium on which a program for causing a computer device to execute is recorded. The storage of the selection information stores, as the selection information, information indicating that the result data of the last old process has not passed through its own node,
When the result data is selected and output, when the node identification information added to the selected result data indicates the own node and the new and old identification information indicates an old process, the selection information is passed. For the result data that has been updated to indicate completion and arrived and retained result data, when the added node identification information indicates the own node and the new and old identification information indicates a new process, the selection information The computer-readable recording medium according to claim 9, wherein when the data does not indicate the passing, the result data is not selected.

Images