JPH11259559A - Financial information processing system - Google Patents

Abstract

(57) [Summary] [PROBLEMS] To provide a financial information processing system capable of performing a mark-to-market process for a plurality of portfolios in a short time. A calculation process reads a predetermined one portfolio from a database storing a plurality of portfolios and executes a mark-to-market process on the read one portfolio. Only one control process 31 always resides on the calculation server 30. The control process 31 determines the number of calculation processes 32 that can be executed at the same time when the price-request processing request is sent from the dealer terminal 20, and starts the calculation processes 32 by the determined number. Then, the started calculation processes 32 are caused to execute the mark-down processing in parallel for a plurality of portfolios corresponding to the mark-up processing request. The result of the price washing process is displayed on the display device 50 of the dealer terminal 20.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a financial information processing system for supporting dealings in financial derivative products (derivatives) such as swaps and options and risk management operations.

[0002]

2. Description of the Related Art In a derivative transaction, a dealer uses a computer in order to quickly and accurately grasp a risk state associated with a business, such as a current asset value or a change in the asset value. In addition, a mark-to-market process is performed for a plurality of portfolios. Here, the portfolio refers to a group of contract data classified according to a certain standard. In addition, the contract data is data that records the details and conditions of transactions of derivative products that have been completed in the past. Usually, in the mark-to-market process, the present value of an asset is evaluated by a certain formula using market data such as interest rates and exchange rates. In particular, by performing mark-to-market processing on a portfolio basis, dealers can:
The current risk status can be evaluated in more detail.

[0003]

By the way, market condition data such as interest rates and exchange rates fluctuate from moment to moment, and accordingly, the current asset value and the risk status change steadily. For this reason, it is desired that the dealer performs a mark-up process on a plurality of portfolios as many times as possible in a day in association with the actual market movement, and always evaluates the latest risk status. However, portfolios that are subject to mark-to-market processing typically contain dozens to thousands of contract data, and the number of contract data tends to increase in the future. For this reason, the mark-to-market process for a plurality of portfolios can take tens of minutes, and in some cases, hours. Reduction of the processing time has been an issue.

The present invention has been made based on the above circumstances, and has as its object to provide a financial information processing system capable of performing a mark-to-market process for a plurality of portfolios in a short time.

[0005]

A financial information processing system according to the present invention for achieving the above object has a storage means for storing a plurality of portfolios including a plurality of contract data, and a predetermined one from the storage means. One or more portfolios are read, a plurality of calculation process means for executing a mark-to-market process on the read one or more portfolios, and an instruction to perform a mark-to-market process on one or more portfolios is received. Sometimes, the number of the calculation process means that can be executed simultaneously is determined, the calculation process means is started up by the determined number, and the mark-up processing for the one or more designated portfolios is started up. Control process means to be executed in parallel by each of the calculation process means; and Output means for outputting the result of the mark-to-market process for one or more portfolios, wherein each of the calculation process means executes a plurality of threads when executing the mark-to-market process for one or more portfolios. And causing the threads to independently execute predetermined calculations on contract data included in the portfolio.

When the control process means receives an instruction to perform a mark-to-market process for one or a plurality of portfolios, the control process means determines the number of calculation process means that can be executed at the same time, and executes the calculation process means by the determined number. Launch. Then, each of the started calculation process means is caused to execute the mark-to-market processing in a multi-process. By performing the price washing process in a multi-process as described above, according to the present invention, a quick price washing process can be performed, and the time until a result is obtained can be greatly reduced.

According to another aspect of the present invention, there is provided a computer-readable recording medium for reading one or more predetermined portfolios from storage means for storing a plurality of portfolios including a plurality of contract data. A plurality of calculation programs for executing the mark-to-market processing on the read one or a plurality of portfolios, and, when receiving an instruction to perform the mark-to-market processing on one or a plurality of portfolios, Determine the number of calculation programs, launch the calculation programs by the determined number, and execute the mark-to-market processing for the designated one or more portfolios in parallel by the launched calculation programs. And a control program, wherein each of the calculation programs has one or more In performing Nearai process for O, it generates a plurality of threads, is characterized in that for the execution independently a predetermined calculation for commitments data included in the portfolio each thread.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing an overall configuration of a financial information processing system according to a first embodiment of the present invention. As shown in FIG. 1, the financial information processing system according to the first embodiment includes a network 10 in which a plurality of dealer terminals 20, a plurality of calculation servers 30 located in a central portion of the network 10, and a database 40 are connected. Have been. Each dealer terminal 20 is installed in a dealing room in which the dealer performs a transaction, and each of the calculation servers 30
The database 40 is installed in, for example, another room.
Inter-process communication is performed between the two. Each calculation server 30 is a multiprocessor-compatible parallel computer, and executes a mark-to-market process with respect to a plurality of portfolios using multi-processes and multi-threads, as described later.

[0009] The financial information processing system of the first embodiment mainly calculates contract data related to derivative transactions. Derivative transactions involve trading products (derivative products) whose value is determined according to the price of the underlying asset, and the actual settlement is first. The contract data is data that records the details and conditions of a derivative product transaction that has been completed in the past. Contract data varies from financial institution to financial institution, but can range from tens of thousands to hundreds of thousands, but these are grouped according to certain criteria. A collection of execution data divided for each group is a portfolio. The grouping is performed, for example, for each product such as a swap or swaption, or for each department. In addition, one portfolio usually contains tens to thousands of contract data.

[0010] The dealer needs to quickly evaluate the present value of the portfolio currently held, risks that may occur due to market fluctuations, and the like. In addition, when there is a new business inquiry, the dealer needs to know the market value evaluation of the asset value currently held in a short time in order to quickly determine whether or not the inquiry should be received. In addition, an assessment needs to be made of the extent to which the asset may have changed at some point in the future.

[0011] However, the value of the derivative described in the contract data changes in accordance with the floating interest rate and the underlying asset value, not the actual accumulation of money. Moreover, in general, the evaluation requires advanced calculation processing based on economic theory that makes full use of advanced mathematics, and when the number of contracts becomes very large, the valuation of assets using contract data requires a high-performance computer. There is a need to do. The calculation server 30 performs a calculation for evaluating the current asset using the contract data. The calculation process for evaluating the current asset is called a mark-to-market process. In particular, in the first embodiment, a mark-to-market process is performed for a plurality of portfolios on a portfolio basis. By performing the mark-to-market processing on a portfolio basis in this way, the dealer can evaluate the current asset value in detail. In the first embodiment, it is assumed that a plurality of portfolios to be subjected to mark-to-market processing are composed of portfolios of the same type. That is, it is assumed that the number of contract data, the types of products such as swaps and swaptions, and the like are almost the same for each portfolio.

FIG. 2 shows the dealer terminal 20 and the calculation server 30.
Are shown one by one. A display device 50, a keyboard 60, and the like are connected to the dealer terminal 20. On the display device 50, the execution status of the price washing process, the processing result, and the like are displayed. The keyboard 60 is used by the dealer to identify the portfolio for which the price-marking process is to be performed.
It is used when D is specified. The dealer terminal 20 is also connected to a printer (not shown) that outputs the result of the price-wash process to paper. Note that a printer server via a network may be used.

[0013] Inside the dealer terminal 20, a financial derivative asset for transmitting a mark-to-market processing request to the calculation server 30 via the network 10, receiving the execution status of the mark-to-market processing, receiving the result of the mark-to-market processing, and the like. Value evaluation application software (hereinafter simply referred to as “client application”) 21 is running. When the dealer wants to execute the mark-to-market process on a plurality of portfolios during the dealing business, he or she inputs an instruction to that effect on the dealer terminal 20 together with the portfolio ID from the keyboard 60 or the like. Then, the client application 21 receives this and transmits it to the calculation server 30 as a mark-to-price processing request. This price-request processing request includes an ID for specifying the portfolio. When the calculation server 12 receives the price-request processing request, the price-reserve processing is actually started. The specific processing contents will be described later. In addition, on the dealer terminal 20, in addition to the client application 21, various application programs for managing risk of derivative transactions are running.

A plurality of portfolios are stored in the database 40 shown in FIG. The contract data included in each portfolio is provided with the ID of the portfolio. Market conditions data such as current or past interest rates and exchange rates are also stored in the database 40.
When a request for mark-to-market processing is sent from the dealer terminal 20 to the calculation server 30, the calculation server 30 determines the I
The execution data corresponding to D is read from the database 40 onto the main memory of the calculation server 30. At this time,
Predetermined market condition data is also read out on the main memory.

On the calculation server 30, as shown in FIG. 2, a control process 31 and a calculation process 32 are operating. The calculation process 32 reads one predetermined portfolio from the database 40 under the control of the control process 31 and executes a mark-to-market process on the read one portfolio. The control process 31 is always resident on the calculation server 30, and waits for a request for a mark-to-price process from the dealer terminal 20. The control process 31 determines the number of calculation processes 32 that can be executed at the same time when the price-request processing request is sent from the dealer terminal 20, and starts the calculation processes 32 by the determined number. Then, the started calculation processes 32 are caused to execute the mark-down processing in parallel for a plurality of portfolios corresponding to the mark-up processing request. That is, a mark-to-market process is performed on a plurality of portfolios by a multi-process. In the first embodiment, for example, each of the calculation servers 30 is equipped with eight CPUs for executing a price washing process by a plurality of calculation processes 32.
Note that a main CPU that operates the calculation process 32 itself is used.
May be provided separately.

Each calculation process 32 always notifies the control process 31 of the processing status. Control process 31
Sends the processing status to the client application 21
And causes the display device 50 of the terminal to display it. FIG. 3 shows a control process 31 on the calculation server 30.
FIG. 4 is a diagram showing a relationship between the calculation process 32 and the calculation process 32; FIG. 3 shows a situation in which a client application 21 running on one dealer terminal 20 issues a request for a markdown process to a control process 31 on a calculation server 30. In this case, in the first embodiment, the control process 31 first starts only one calculation process 32 as shown in FIG.
In step 2, a mark-to-market process is executed on a representative one of the plurality of portfolios corresponding to the mark-to-market processing request on a trial basis. Then, the CPU operating rate during the execution is monitored, and the maximum value of the CPU operating rate is obtained. Here, the CPU operating rate is defined as 100 when all eight CPUs that execute the price-washing process are fully operated.
% Means what percentage the CPU is operating during the execution of the process. The fact that the CPU operating rate is low means that the processing capacity of the calculation server 30 has room.

Next, the control process 31 can be executed simultaneously based on the maximum value of the CPU operating rate obtained as described above, that is, if some calculation processes 32
Calculation process 3 in which the maximum CPU operation rate becomes somewhat smaller than 100% when the
Determine the number of 2. Here, the maximum CPU operation rate is 10
The reason why the number of the calculation processes 32 is not determined so as to be 0% is that when the maximum CPU operation rate is 100% or more,
This is because too much load is applied to the CPU, and the processing efficiency is rather deteriorated. For example, assuming that the maximum value of the CPU operation rate obtained by trial execution is about 30%, the control process 31 determines the number of simultaneously executable calculation processes 32 to be three.

The control process 31 actually starts up the calculation processes 32 by the number of the determined calculation processes 32. In FIG. 3B, as an example, the control process 31
Has started three calculation processes 32, 32, 32. Then, the control process 31 causes the started calculation processes 32 to execute a mark-to-market process in parallel with respect to portfolios other than the portfolio executed on a trial basis. In the first embodiment, since a plurality of portfolios to be subjected to mark-to-market processing consist of the same type of portfolio, each calculation process 32
May process the portfolio in any order.

Next, the contents of the mark-to-market processing for one portfolio executed by each calculation process 32 will be described. The mark-to-market process for one portfolio is roughly divided into three processes: a pre-process, a contract calculation process, and a post-process. The pre-processing is a preparatory process for reading out the portfolio to be processed by the calculation process 32 from the database 40 onto the main memory of the calculation server 30, and for each execution data of the read out portfolio, to enter into the execution of the execution of the execution of the execution of the execution of the execution of the execution of the execution of the execution of the execution. Is what you do. In this preparation process, specifically, the contract data is processed using interest rate data or the like so that it can be easily calculated by a thread described later, and the processed contract data is written to the main memory.

It is to be noted that each calculation process 32 is executed by the database 4
Each of the portfolios read from 0 is temporarily stored in a different area on the main memory. In the first embodiment, each calculation process 32 can simultaneously read data stored in the main memory and the like by using a multi-engine for supporting multi-access.

The contract calculation processing is to execute a predetermined calculation for each contract data in a multi-thread. When each calculation process 32 enters the contract calculation process, first, the number of threads is determined. The thread is a unit of processing divided according to the number of CPUs provided in the calculation server 30, and is set by software by the calculation process 32. That is, the number of threads is the number of CPUs exclusively used by the calculation process 32. When the number of threads is determined, the calculation process 32 generates threads by that number, and each thread independently executes a predetermined calculation on each execution data included in the portfolio in parallel.

The post-process is a process after the execution of the contract calculation process. In fact, at the end of the execution, the final result for the portfolio has not been calculated. In this post-processing, the results processed by each thread are collected to calculate the final result for the portfolio, and the obtained result is written to the database 40.

FIG. 4 is a diagram schematically showing the operation of one calculation process 32. 4, the input-side shared memory 38 is a main memory including a RAM and the like provided in the calculation server 30. The portfolio held by the calculation process 32 is stored in the input side shared memory 38, and is set as one table. In the example of FIG. 4, n pieces of execution data D ₁ , D ₂ ,.
, D _n are stored. The portfolio is originally stored in the database 40, which is extracted by the calculation process 32 and stored in the shared memory 38 on the input side.

When actually executing the mark-to-market process, the calculation process 32 first fetches a predetermined portfolio from the database 40 and writes it on the input side shared memory 38 as preprocessing. After that, the contract calculation process is started, and a thread is set by software. In FIG. 4, four threads 33, 33, 33, 33 are set as an example. Since one thread 33 is a unit that performs processing independently using one CPU, four threads 3
3, 33, 33, 33 read contract data from the input side shared memory 38 one by one, and a total of four C
Parallel processing with a degree of parallelism of 4 is performed using the PU. That is, the n pieces of execution data D ₁ , D ₂ ,..., D _n are not processed one by one, but the n pieces of execution data D ₁ , D ₂ ,. .., D _n are processed. By performing the parallel processing with the multi-thread as described above, the time required for the contract calculation processing is reduced.

At the time of reading contract data, while a thread is reading contract data from the input side shared memory 38, another thread can read contract data under the exclusive control of the lock mechanism 34. I cannot do it. Further, in order to prevent the same contract data from being processed by a plurality of threads, the contract data once read by one thread is not read by another thread. The lock mechanism 34
Is implemented by an OS standard exclusive lock mechanism. This will be described later. The lock mechanism 34 is
This may be realized by creating a lock file and determining whether each thread can access by accessing the lock file.

The data calculated by the CPU for each thread is written into the shared memory 39 on the output side, for example, as matrix data. This output side shared memory 39 is also the main memory of the calculation server 30. At this time, if each thread writes data randomly, there is a problem that the entire data cannot be consistent. In this case as well, the lock mechanism 35 operates and a thread writes data to the output side shared memory 39. When it is, exclusive control is performed so that other threads cannot write data.

FIG. 5 shows the shared memory 3 on the input side shown in FIG.
FIG. 8 is a view showing a table of contract data developed on 8. Contract data is placed in each row of the table, and a 1-bit processed flag is provided at the right end of each row. Each thread sets the processed flag "1" to the contract data for which it is executing the process, and indicates that the contract data has already been processed. When looking at this table, other threads skip the execution data for which the processed flag “1” is set, and search for and read execution data for which the processed flag is “0”. At this time, in the first embodiment, among the contract data whose processed flag is “0”, the contract data at the top is indicated by the pointer. The thread looks for this pointer to find the data to read. This allows
The thread can easily find the contract data to read. Instead of using the pointer, the thread may scan the processed flag and read out the flag as soon as it finds "0".

Next, a description will be given of a processing procedure when a mark-to-market process is performed on a plurality of portfolios in the financial information processing system of the first embodiment. FIG. 6 is a flowchart illustrating a processing procedure when a mark-to-market process is performed on a plurality of portfolios. When a dealer sends an instruction from a predetermined dealer terminal 20 to perform a mark-to-market process on a plurality of portfolios to the calculation server 30, the control process 31 of the calculation server 30 starts up one calculation process 32 and performs the calculation. The process 32 causes the mark-to-market processing for one specific portfolio to be executed on a trial basis (step 11). The control process 31 monitors the CPU operating rate during the execution of the trial price washing process, and obtains the maximum value of the CPU operating rate. Then, the optimal number of calculation processes 32 is determined based on the obtained maximum value of the CPU operation rate (step 12). Thereafter, the control process 31 starts up the calculation processes 32 by the number determined in step 12, and causes each calculation process 32 to execute a mark-to-market process in parallel in a predetermined order for a plurality of portfolios (step 13).

FIG. 7 is a flowchart showing the procedure of processing performed in one calculation process. When the control process 31 starts a plurality of calculation processes 32 in the process of step 13 in FIG. 6, the process shown in FIG. 7 is started in each calculation process 32. When the processing procedure of FIG. 7 is started, first, the calculation process 32 reads the portfolio to be processed specified by the control process 31 from the database 40 onto the main memory of the calculation server 30 and determines the number of threads. Then, threads are generated by the determined number (step 21). Next, the calculation process 32 causes each thread to execute an operation for executing a contract calculation in parallel (step 22). That is, arithmetic operations are simultaneously and independently executed by the CPUs corresponding to the number of threads. When the parallel processing ends, the thread is deleted (step 23). Thus, the processing of one calculation process 32 ends.

FIG. 8 shows step 22 of the processing procedure shown in FIG.
3 is a flowchart showing a detailed processing procedure of a part of FIG. In step 30, the thread first determines whether or not the lock mechanism 34 shown in FIG. 4 is in a locked state, that is, whether or not another thread has locked the lock mechanism 34. If it is not locked and it is in a state where it can read contract data from the input side shared memory 38, the lock mechanism 34 is locked so that other threads cannot read it (step 31). Next, as shown in FIG. 5, the thread searches for the contracted data whose processing completion flag indicated by the pointer is “0” from the contracted data, and performs a process of fetching the contracted data. "1" is set to the data processing completion flag (step 32). As described above, when the thread reads the execution data and sets the processed flag of the execution data to “1”, it is not necessary to eliminate the access to the input-side shared memory 38, so that the lock is released (step 33). Then, move on to the actual calculation process (step
34). When the calculation is completed, the thread determines whether or not the lock mechanism 35 shown in FIG. 4 is in a locked state, that is, whether or not another thread has locked the lock mechanism 35 (step 35). If it is not locked and it is in a state where it can write the calculation result data to the output side shared memory 39, it locks the lock mechanism 35 so that other threads cannot write (step
36). Next, the thread writes the calculation result data in the output side shared memory 39 (step 37), and then releases the lock (step 38). In step 39, it is determined whether there is unprocessed contract data. If any are unprocessed, repeat the same process as described above,
If there is no unprocessed contract data, the process ends.

Next, a case in which the financial information processing system of the first embodiment is used to execute a mark-to-market process for a plurality of portfolios by the above-described multi-process and multi-thread processing method (processing method of the present invention) A description will be made of a comparison between a case where processing is performed by a processing method using only multi-threads without using multi-processes (multi-thread processing method) and a case where processing is performed by a conventional processing method not using multi-processing or multi-threading. FIG. 9A is a diagram showing the relationship between the CPU operation rate and the processing time when a mark-to-market process is performed on a plurality of portfolios by a conventional processing method, and FIG. CP when the mark-to-market process was performed on the portfolio
FIG. 9C is a diagram showing the relationship between the U operating rate and the processing time, and FIG. 9C is a diagram showing the relationship between the CPU operating rate and the processing time when the mark-up processing is performed on a plurality of portfolios by the processing method of the present invention. It is. Here, FIGS. 9 (a), (b) and (c)
, The vertical axis represents the CPU operation rate, and the horizontal axis represents the processing time.

In the conventional processing method, one CPU sequentially processes one portfolio at a time, and also processes one contract data included in the portfolio. In the first embodiment, the CPU operating rate is defined as 100% when all eight CPUs are fully operated. In the conventional processing method, only one CPU is operating at all times,
Naturally, as shown in FIG. 9A, the CPU operation rate is considerably small over the entire period in which pre-processing, contract calculation processing, and post-processing are performed for each portfolio. Therefore, the processing time until the mark-down processing is completed for all portfolios is very long.

On the other hand, in the multi-thread processing method, the mark-to-market processing is sequentially performed for each portfolio one by one. In the price-wash processing, the execution calculation processing is performed by multi-thread. For this reason, as shown in FIG. 9B, the CPU operation rate is high during the period in which the contract calculation process is performed for each portfolio. On the other hand, during the period of performing the pre-processing or the post-processing, the CPU operation rate is the same in the conventional processing method and the multi-thread processing method. In FIG. 9A or 9B, the processing amount when executing the contract calculation processing for one portfolio is represented by a curve representing the CPU utilization rate and a horizontal axis corresponding to the period during which the contract calculation processing is being performed. It is represented by the area enclosed by. Such throughput, of course,
It does not change whether the conventional processing method is used or the multi-thread processing method is used. For this reason, in the case where price-pickup processing is performed by the multi-thread processing method, the CPU operation rate during the period in which the contract calculation processing is performed is high. Processing time is reduced. Therefore, the processing time until the mark-down processing is completed for all portfolios is reduced.

By the way, one portfolio currently contains tens to thousands of contract data. When a portfolio including contract data of such a scale is subjected to a mark-to-market processing by a multi-thread processing method using the financial information processing system of the first embodiment, the CPU operation rate during the contract calculation processing is usually, for example, about If it is 20 to 40%, there is still room for the processing capacity of the CPU. The processing method of the present invention pays attention to this point and
Of the plurality of calculation processes 32
To execute a mark-to-market process on a plurality of portfolios in parallel. Accordingly, several mark-to-market processings for the portfolio are executed at the same time.
The U operation rate becomes extremely high as shown in FIG. For this reason, the processing time until the mark-down processing for a plurality of portfolios is completed is significantly reduced as compared with the case where the mark-down processing is performed by the multi-thread processing method.

In the financial information processing system according to the first embodiment, quick processing is possible by causing a plurality of calculation processes to execute mark-to-market processing on a plurality of portfolios in parallel, and the time required for obtaining a result is obtained. Can be greatly reduced. For this reason, the dealer can perform a mark-up process on a plurality of portfolios several times a day in association with the actual market movement, and can always quickly evaluate the latest risk status.

Next, a financial information processing system according to a second embodiment of the present invention will be described. In the financial information processing system of the second embodiment, a method for determining the number of calculation processes that can be executed simultaneously is different from the method of the first embodiment. In the first embodiment, a case is considered in which a plurality of portfolios to be subjected to mark-to-market processing are a group of portfolios of the same type. However, in the second embodiment, different portfolios may be included in a plurality of portfolios to be subjected to mark-to-market processing. In the second embodiment, components having the same functions as those of the first embodiment are denoted by the same reference numerals, and a detailed description thereof will be omitted.

In the second embodiment, the control process 31 of the calculation server 30 has data relating to the maximum value of the CPU operation rate when the mark-to-market process is executed for one representative portfolio in advance. The control process 31
Upon receiving an instruction from the dealer terminal 20 to execute a mark-to-market process for a plurality of portfolios, an initial value of the number of simultaneously executable calculation processes 32 is determined based on data relating to the maximum value of the CPU operation rate. . And
The calculation process 32 is started up by the determined initial value, and the price washing process is executed in a multi-process.

Further, the control process 31, in advance with the data of the lower limit value T _a and the upper limit value T _b for CPU utilization. The control process 31 includes a plurality of calculation processes 32
The CPU operating rate is constantly monitored during the execution of the price washing process. Then, the CPU operation rate is in the range [T _a ,
_Tb ], the control process 31 causes the plurality of calculation processes 32 to execute the processing as it is. On the other hand, the control process 31, when the CPU operation rate is determined to have become smaller than the lower limit value T _a, decides to increase one the number of computation process to be executed simultaneously. Also,
If it is determined that the CPU operation rate has become larger than the upper limit value _Tb , it is determined that the number of calculation processes 32 to be executed simultaneously is reduced by one. Then, when such a determination is made, the control process 31 causes the determined number of calculation processes 32 to execute the mark-down processing in parallel. As described above, in the second embodiment, the control process 31 automatically determines the number of the calculation processes 32 to be simultaneously executed so that the CPU operation rate falls within a certain range, so that the number of the calculation processes 32 can be reduced. Even when a plurality of portfolios include different types of portfolios, the mark-to-market process can be executed efficiently.

FIG. 10 is a flowchart showing a processing procedure when a mark-to-market process is performed on a plurality of portfolios in the financial information processing system of the second embodiment. When a dealer sends an instruction to the calculation server 30 to perform a mark-to-market process on a plurality of portfolios from the dealer terminal 20, the control process 31 executes the control process 31 at the same time based on data on the maximum CPU operating rate provided in advance. The number of possible calculation processes 32 is determined (step 41). Next, the control process 31 starts up the calculated processes 32 by the determined number, and causes each of the calculated processes 32 to execute the mark-down processing in parallel (step 42). Then, during the execution of the price washing process (step 43), the control process 31 constantly monitors the CPU operating rate and determines that the CPU operating rate is within _a predetermined range [T _a , T
_b ] is determined (step 44). C
If the PU operation rate falls within the predetermined range [T _a , T _b ], the calculation process 32 is caused to execute the mark-down processing as it is (step 42). On the other hand, when the CPU operating rate is within a predetermined range [T
_a, if it is determined that not within T _b], the control process 31 determines to change the number of calculation processes 32 to be executed simultaneously (step 45), the number of computation process 32 that the change Nearai Execute the process (step4
2). When the control process 31 determines that the mark-to-market process has been completed for all portfolios to be processed (step 43), the process exits from the flow of FIG.

In the financial information processing system according to the second embodiment, as in the first embodiment, a plurality of calculation processes execute a mark-to-market process on a plurality of portfolios in parallel, thereby enabling quick processing. Is possible,
The time until a result is obtained can be greatly reduced.
In the financial information processing system according to the second embodiment, the control process determines whether or not the CPU operating rate is within a predetermined range while executing the mark-to-market processing in the multi-process. Accordingly, by changing the number of calculation processes, it is possible to improve the efficiency of the mark-to-market processing even when the portfolios of the mark-to-market processing include different types of portfolios. Rapid processing is possible.

Next, a financial information processing system according to a third embodiment of the present invention will be described. In the financial information processing system of the third embodiment, the method of determining the number of calculation processes that can be executed simultaneously is different from the methods of the first and second embodiments. In the first embodiment, a case is considered in which a plurality of portfolios to be subjected to mark-to-market processing are a group of portfolios of the same type. However, in the third embodiment, similarly to the second embodiment, a plurality of portfolios to be subjected to a mark-to-market process may include different types of portfolios. In the third embodiment, components having the same functions as those in the first embodiment are denoted by the same reference numerals, and a detailed description thereof will be omitted.

In the financial information processing system according to the third embodiment, the database 40 stores a plurality of portfolios and stores processing characteristic data describing characteristics of mark-to-market processing for the portfolios in a table format. FIG.
FIG. 4 is a diagram for explaining a processing characteristic table which is a collection of processing characteristic data of a portfolio. As shown in FIG. 11, the processing characteristic table includes the portfolio ID, the portfolio type, the maximum CPU operating rate during execution of the mark-to-market processing for the portfolio, and the processing time required for the mark-to-market processing for the portfolio. And the number of contract data included in the portfolio. Here, the portfolio type column describes, for example, whether the contract data included in this portfolio is only a swap or is composed of a swap and a swaption.

The processing characteristic data of the portfolio is
It is not necessary to create all portfolios in advance and store them in the database 40. In the third embodiment, for a portfolio for which processing characteristic data has not been created, as described below, a mark-to-market request is issued,
At the time of actually executing the mark-to-price processing, processing characteristic data is created.

When the control process 31 of the calculation server 30 receives an instruction from the dealer terminal 20 to execute a mark-to-market process for a plurality of portfolios, the database 40 of the processing characteristic data of the portfolio corresponding to the instruction. Is read out onto the main memory of the calculation server 30. Then, the control process 31
For portfolios for which processing characteristic data is not stored in the database 40, one calculation process 32 individually executes a mark-to-market process, and processing characteristic data is created and stored in the database 40. On the other hand, for a portfolio whose processing characteristic data is stored in the database 40, the control process 31 determines the number of simultaneously executable calculation processes 32, the processing order of the portfolio, and the like based on the processing characteristic data read from the database 40. To determine. Then, the calculation processes 32 are started up by the determined number, and the calculation processes 32
Then, a price wash process is executed in parallel according to the determined processing order.

Next, how to determine the number of the calculation processes 32 and the processing order of the portfolio will be specifically described. For example, it is assumed that, of a plurality of portfolios to be processed, two portfolios include many swaptions, and the other portfolios include only swaps. Here, a portfolio including a large number of swaptions has a maximum CPU operating rate of, for example, 40 to 50.
% And the portfolio containing only swaps is CP
The maximum value of the U operation rate is, for example, 20 to 30%. In this case, the control process 31 first starts up two calculation processes 32, and causes the two calculation processes 32 to simultaneously execute a mark-to-market process on a portfolio including many swaptions. Next, after the mark-down processing for the two portfolios including a large number of such swaptions is completed, another calculation process 32 is started, and the mark-up processing for the portfolio including only the swap is added to the three calculation processes 32 in total. At the same time. As described above, the control process 31 determines the number of the calculation processes 32 and the processing order of the portfolios so that the mark-to-market processing for the plurality of portfolios is performed efficiently.

FIG. 12 is a flowchart showing a processing procedure when a mark-to-market process is performed on a plurality of portfolios in the financial information processing system of the third embodiment. When a dealer sends an instruction from the dealer terminal 20 to execute a mark-to-market process on a plurality of portfolios to the calculation server 30, the control process 31 sends processing characteristic data for the portfolio corresponding to the instruction from the database 40 to the main memory. Then, it is determined whether or not processing characteristic data is stored in the database 40 for all of the portfolios to be processed (step 51). If it is determined that there is processing characteristic data for all portfolios to be processed, the process proceeds to step 54. On the other hand, if it is determined that there is at least one portfolio having no processing characteristic data, the control process 31
Is started, and the calculation program 32 individually executes a mark-to-market process on the portfolio without the processing characteristic data (step 52). Then, the control process 31
At the time of such processing, the CPU operating rate, the processing time, and the like are monitored, processing characteristic data for the portfolio is newly created, and this is stored in the database 40 (st
ep53). Thereafter, the process proceeds to step 54. If there is no processing characteristic data for all of the portfolios to be processed, since all the results have been obtained by the mark-to-market processing in step 52, the processing ends in step 53 without proceeding to the steps after step 54.

Next, in step 54, the control process 31
Based on the processing characteristic data read in the processing of step 51, the number of simultaneously executable calculation processes 32 is determined, and the processing order of the portfolio is determined, for example, in descending order of the operation rate (step 54). And control process 3
1 starts the calculation processes 32 as many as the determined number, and makes each calculation process 32 execute step processing according to the determined processing order.
A mark-to-price process is executed in parallel on the portfolio corresponding to the processing characteristic data read out in step 51 (step 55).

As described above, a mark-to-market process is individually performed to create processing characteristic data for a new portfolio, while a portfolio for which processing characteristic data has already been created is multi-processed. A washing process is performed. Therefore, if all of the portfolios to be processed this time are portfolios that have been at least once price-marked, the current price-marking processing does not go through the steps 52 and 53 at all. Since the processing is performed from step 54, the processing time is greatly reduced.

It should be noted that the portfolio is not always the same, and the number of contract data included in the portfolio may increase in the future or the type may change in accordance with the increase. In order to deal with such a case, the processing characteristic data may be updated at regular intervals. In the financial information processing system of the third embodiment, similar to the first embodiment, quick processing is possible by causing a plurality of calculation processes to execute mark-to-market processing on a plurality of portfolios in parallel. Yes, the time it takes to get results can be significantly reduced. Further, in the financial information processing system according to the third embodiment, the number of calculation processes that can be executed simultaneously and the processing order of the portfolio are determined based on the processing characteristic data of the portfolio, so that a plurality of Even when portfolios of different types are included in the portfolio, the efficiency of the mark-to-price processing can be improved.

The present invention is not limited to the above embodiments, and various modifications can be made within the scope of the invention. In each of the embodiments described above, the case where the execution calculation process is performed by multi-threads has been described, but the execution calculation process does not necessarily need to be performed by multi-threads. However, in this case, it goes without saying that the processing time is slower than in the case where the price washing process is performed by multi-process and multi-thread.

Further, in each of the above embodiments, a case has been described in which each computation server is equipped with eight CPUs for executing the price-wash processing. However, in general, any number of CPs other than eight can be used.
U may be mounted on each calculation server. In extreme cases, each calculation server may use one CPU.

[0052]

As described above, according to the financial information processing system according to the present invention, quick processing is possible by causing a plurality of calculation processes to execute a mark-to-market process on a plurality of portfolios in parallel. Yes, the time it takes to get results can be significantly reduced. For this reason, the dealer can perform a mark-up process on a plurality of portfolios several times a day in association with the actual market movement, and can always quickly evaluate the latest risk status.

[Brief description of the drawings]

FIG. 1 is a diagram showing an overall configuration of a financial information processing system according to a first embodiment of the present invention.

FIG. 2 is a diagram showing one dealer terminal and one calculation server in the financial information processing system.

FIG. 3 is a diagram showing a relationship between a control process and a calculation process on a calculation server.

FIG. 4 is a diagram schematically showing an operation of a calculation process.

5 is a diagram showing a table of contract data developed on the input-side shared memory shown in FIG. 4;

FIG. 6 is a flowchart showing a processing procedure when performing a mark-to-market process on a plurality of portfolios in the financial information processing system of the first embodiment.

FIG. 7 is a flowchart illustrating a procedure of processing performed in one calculation process.

FIG. 8 is a flowchart detailing a part of the processing procedure shown in FIG. 7;

FIG. 9 is a diagram showing a relationship between a CPU operation rate and a processing time when a mark-to-market process for a plurality of portfolios is performed in each of the conventional processing method, the multi-thread processing method, and the processing method of the present invention.

FIG. 10 is a flowchart showing a processing procedure when a mark-to-market process is performed on a plurality of portfolios in the financial information processing system of the second embodiment.

FIG. 11 is a diagram illustrating a processing characteristic table which is a collection of processing characteristic data of a portfolio.

FIG. 12 is a flowchart showing a processing procedure when performing a mark-to-market process on a plurality of portfolios in the financial information processing system of the third embodiment.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 Network 20 Dealer terminal 21 Client application 30 Calculation server 31 Control process 32 Calculation process 33 Thread 34,35 Lock mechanism 38 Input shared memory 39 Output shared memory 40 Database 50 Display device 60 Keyboard

Claims (7)

[Claims] 1. A storage unit for storing a plurality of portfolios including a plurality of contract data, a predetermined one or a plurality of portfolios are read from the storage unit, and a mark-to-market process is performed on the read one or a plurality of portfolios. A plurality of calculation process means to be executed, and when receiving an instruction to perform a mark-to-market process for one or a plurality of portfolios, determine the number of the calculation process means that can be executed simultaneously, and determine only the determined number. Control process means for starting up the calculation process means, and causing the price calculation processing for the designated one or more portfolios to be executed in parallel by each of the started calculation process means; and Output means for outputting the result of the mark-to-market processing for the obtained one or more portfolios, Wherein each of the calculation process means generates a plurality of threads when executing a mark-to-market process for one or a plurality of portfolios, and performs a predetermined calculation on the execution data included in the portfolio in each of the threads. A financial information processing system characterized by being executed independently. 2. The control process means, when receiving the instruction, activates one of the calculation process means,
The started calculation processing means is caused to execute the mark-down processing, and the maximum CPU operation rate during execution of the price-measurement processing by the one calculation processing means is determined. Based on the maximum value of
2. The financial information processing system according to claim 1, wherein the number of said calculation process units that can be executed simultaneously is determined. 3. When the control process means causes each of the calculation process means to execute a mark-to-market process in parallel for a plurality of portfolios, the CPU operation rate during the execution of the parallel process becomes a predetermined upper limit value. If it is determined that the CPU processing rate has become larger, the number of the calculation process means that can be executed at the same time is reduced. 2. The financial information processing system according to claim 1, wherein the number of possible calculation processing means is increased. 4. The control process means, upon receiving the instruction, based on data relating to the maximum value of the CPU operating rate when executing a mark-to-market process on one representative portfolio provided in advance. 4. The financial information processing system according to claim 3, wherein the number of said calculation process units that can be executed simultaneously is determined. 5. The storage means stores processing characteristic data for a portfolio including at least information relating to a CPU operation rate. When the control process means receives the instruction, the control process means performs processing on a portfolio corresponding to the instruction. Of the processing characteristic data stored in the storage means, and for the portfolio in which the processing characteristic data is stored in the storage means, based on the processing characteristic data read from the storage means. Determining the number of the calculation process units that can be executed simultaneously and the processing order of the portfolio, and causing each of the calculation process units to execute the mark-to-order processing in parallel according to the determined number and the processing order. The financial information processing system according to claim 1, wherein 6. The control process means, for a portfolio in which the processing characteristic data is not stored in the storage means, causes one of the calculation processing means to individually execute a mark-to-market process and creates the processing characteristic data. And storing the data in the storage means.
The financial information processing system as described. 7. A plurality of calculations for reading a predetermined one or a plurality of portfolios from storage means for storing a plurality of portfolios including a plurality of contract data, and executing a mark-to-market process on the read one or a plurality of portfolios When receiving an instruction to perform a mark-to-market process for one or more portfolios, determine the number of the calculation programs that can be executed simultaneously, and launch the calculation programs by the determined number, A control program for executing the mark-to-market processing for the instructed one or more portfolios in parallel by each of the launched calculation programs, and wherein each of the calculation programs comprises one or more portfolios. When executing a mark-to-price process for a plurality of threads, a plurality of threads are generated, and A computer-readable recording medium, characterized in that to execute independently a predetermined calculation for commitments data contained in over portfolio.