JP5731332B2 - Apparatus and method for calculating transaction processing time - Google Patents

Description

  The present invention relates to an apparatus and method for calculating transaction processing time. In particular, the present invention relates to an apparatus and method for calculating a processing time excluding a transaction lock waiting time for achieving a target transaction performance.

As information processing technology using a queue model, for example, there are technologies of Patent Documents 1 to 5.
In the technique of Patent Document 1, conditions relating to the average processing time and the number of data are set, and an access control table that defines which access method of the contention avoidance method and the pointer method is selected for each condition is defined as M Determined by verifying with the / M / 1 model.

  In the technique of Patent Document 2, the size Bj of the area for accumulating data in the data stream j in the primary buffer without saving to the secondary buffer is approximated by performing multi-element M / M / 1 type approximation. The expressed waiting time w1j in the data saving process from the primary buffer to the secondary buffer and the waiting time w2j in the data restoration process from the secondary buffer to the primary buffer are set.

  The technique of Patent Document 3 has a broadcasting station site that centrally manages the consistency of replicated data for arbitrary objects existing in the system, and a plurality of distributed data distributed over super multi-points, each having replicated data for arbitrary objects. Model in an M / M / 1 type queue in a distributed database system including a DB site and a satellite for broadcasting information about whether or not to permit updating from the broadcasting station site to each DB site. It is disclosed that the value of the total processing time Eall of the access operation sequence between each DB site, broadcasting station site, and satellite is evaluated.

  In the technique of Patent Document 4, the load generation unit generates a plurality of transactions, the transaction processing unit processes the plurality of transactions, and the computer calculates the number of processing requests for each transaction at a rate generated per unit time. Process requests are executed in small increments starting from a small amount. At each stage, the average processing wait time of transactions is calculated from the transaction processing time and the number of transaction processing requests using queuing model theory. Disclosed is a performance degradation factor detection system that checks whether or not file lock contention occurs when the difference between the waiting time and the actual processing waiting time is large.

  In the technique of Patent Document 5, when one database fails, if the other database is not yet active, it is immediately reassigned to the active state and continues to process subsequent queries and updates during that failure. When the failed database is repaired and restarted, all records are flushed from it, and the failed database is reconstructed using interleaved copy and update operations in a single pass through the active database. The Subsequent transactions and copy operations are interleaved according to a queue threshold method to suppress copy operations in response to an increase in the number of subsequent transactions. The transaction processing system remains operational both during failures and recovery activities, and obtains full recovery without downtime.

JP 2002-351726 A Japanese Patent Laid-Open No. 10-200574 Japanese Patent Laid-Open No. 11-234269 JP 2006-252176 A JP 7-262068 A

  By the way, a major factor that determines the throughput of transaction processing is competition for specific resources that hinder the parallelism of processing. In particular, waiting for lock by exclusive control when updating a database is a problem. For example, when locks are concentrated on a database record of a specific popular product, the throughput is controlled by a lock contention state on the database record. Further, when locks are concentrated on a plurality of records instead of one record in one transaction, it becomes more difficult to improve the throughput.

In a system based on such severe lock contention, it is not easy to obtain the transaction processing time for satisfying the target throughput. If lock contention occurs in one record, the limit value of the processing time can be found from the reciprocal of the throughput without considering the lock waiting time. For example, in order to achieve a throughput of 50 transactions per second, the processing time excluding the lock waiting time should not exceed 20 milliseconds.
On the other hand, when lock contention occurs between two records, it must be assumed that the lock of one record is held while the lock of the other record is held. Therefore, the limit value of the processing time cannot be obtained by simple calculation ignoring the lock waiting time.

Here, as described above, a queuing model has already been used in the field of information processing.
However, any of the techniques disclosed in Patent Documents 1 to 5 discloses a method for obtaining a processing time excluding a transaction lock waiting time for satisfying a target throughput when two records are locked in one transaction. Not what you want.

  An object of the present invention is to obtain a processing time excluding a transaction lock waiting time for achieving a target performance when two records need to be locked in one transaction.

  For this purpose, the present invention is an apparatus for calculating a processing time excluding a transaction lock waiting time for achieving a target transaction performance, and is a transaction for updating at least two records including a first record. A first rate that is a rate of the second record, a second rate that is a rate of a transaction that updates at least two records that include the second record, and at least two records that include the first record and the second record. A rate acquisition unit that acquires a third rate that is the rate of the transaction to be updated, a response time acquisition unit that acquires a target response time of the transaction, and a first of the response times acquired by the response time acquisition unit The first waiting time which is the lock waiting time of the record of the first record and the lock keeping of the first record By applying the first stage of the M / M / 1 queuing model with the first holding time being the time as the service waiting time and service time of the M / M / 1 queuing model, respectively, the rate acquisition unit Based on the acquired first rate and the response time acquired by the response time acquisition unit, a first calculation unit that calculates the first holding time, and the first calculation unit calculated by the first calculation unit The second waiting time that is the lock waiting time of the second record and the second holding time that is the lock holding time of the second record in the holding time of the M / M / 1 queue model As the time and service time, the second rate and the third rate acquired by the rate acquisition unit and the first calculation unit are calculated by applying the second stage of the M / M / 1 queuing model. A second calculating unit that calculates a second holding time based on the first holding time and an output unit that outputs the second holding time calculated by the second calculating unit, Providing the device.

を用いて、第１の保持時間を算出し、第２の算出部は、式

を用いて、第２の保持時間を算出する、ものであってよい（Raは第１のレートであり、Rbは第２のレートであり、Rabは第３のレートであり、Trespは応答時間であり、Tsaは第１の保持時間であり、Tsbは第２の保持時間である）。 Here, the first calculation unit has the formula

Is used to calculate the first holding time, and the second calculating unit

To calculate the second holding time (Ra is the first rate, Rb is the second rate, Rab is the third rate, and Tresp is the response time. Tsa is the first holding time and Tsb is the second holding time).

  Further, the present invention is an apparatus for calculating a processing time excluding a transaction lock waiting time for achieving a target transaction performance, wherein a first rate that is a transaction rate for updating a first record; A rate acquisition unit that acquires a second rate that is a transaction rate for updating the second record, and a third rate that is a transaction rate for updating the first record and the second record; A response time acquisition unit that acquires a target response time, a first waiting time that is a lock waiting time of the first record, and a first holding time that is a lock holding time of the first record are M / M / The first stage of the M / M / 1 queuing model is applied as the service waiting time and service time of the one queuing model. The second waiting time that is the lock waiting time of the second record and the second holding time that is the lock holding time of the second record are the service waiting time and service time of the M / M / 1 queue model, respectively. By applying the second stage of the M / M / 1 queuing model, the first rate, the second rate, and the third rate acquired by the rate acquisition unit, and the response time acquisition unit An apparatus is also provided that includes a calculating unit that calculates a second holding time based on the response time, and an output unit that outputs the second holding time calculated by the calculating unit.

を用いて、第２の保持時間を算出する、ものであってよい（Raは第１のレートであり、Rbは第２のレートであり、Rabは第３のレートであり、Trespは応答時間であり、Tsbは第２の保持時間である）。 Here, the calculation unit uses the formula

To calculate the second holding time (Ra is the first rate, Rb is the second rate, Rab is the third rate, and Tresp is the response time. And Tsb is the second retention time).

  In addition, the calculation unit applies the first stage of the M / M / 1 queuing model to obtain the first rate acquired by the rate acquisition unit and the response time acquired by the response time acquisition unit. Based on the first calculation unit for calculating the first holding time and the second stage of the M / M / 1 queuing model, the second rate acquired by the rate acquisition unit and the second rate And a second calculation unit that calculates the second holding time based on the third rate and the first holding time calculated by the first calculation unit.

  Alternatively, the calculation unit acquires the first rate and the third rate acquired by the rate acquisition unit and the response time acquisition unit by applying the first stage of the M / M / 1 queue model. Based on the response time obtained, the first calculation unit that calculates the first holding time and the second stage of the M / M / 1 queuing model are applied to obtain the first acquired by the rate acquisition unit. And a second calculation unit that calculates the second holding time based on the second rate and the first holding time calculated by the first calculation unit.

  Furthermore, the present invention is a method in which a computer calculates a processing time excluding a transaction lock latency time to achieve a target transaction performance, and the computer is a transaction rate at which a first record is updated. Obtaining a first rate, a second rate that is a rate of a transaction for updating the second record, and a third rate that is a rate of a transaction for updating the first record and the second record. And a computer obtaining a target response time of the transaction, and a first hold time that is a lock waiting time of the first record and a lock holding time of the first record. Service waiting time and service time of M / M / 1 queuing model respectively Then, the first stage of the M / M / 1 queuing model is applied, the second waiting time that is the lock waiting time of the second record, and the second holding that is the lock holding time of the second record. By applying the second stage of the M / M / 1 queuing model, where time is the service latency and service time of the M / M / 1 queuing model, respectively, the first rate, the second rate, A method is also provided that includes calculating a second hold time based on the third rate and the response time, and a computer outputting the calculated second hold time.

  Furthermore, the present invention provides a program for causing a computer to function as a device for calculating a processing time excluding a transaction lock waiting time for achieving a target transaction performance, wherein the computer updates a first record. A first rate that is the rate of the second record, a second rate that is the rate of the transaction that updates the second record, and a third rate that is the rate of the transaction that updates the first record and the second record A rate acquisition unit that acquires a response time, a response time acquisition unit that acquires a target response time of a transaction, and a first waiting time that is a lock waiting time of the first record among the response times acquired by the response time acquisition unit The first holding time, which is the time and the lock holding time of the first record, is M / M By applying the first stage of the M / M / 1 queue model as the service waiting time and service time of one queue model, the first rate acquired by the rate acquiring unit and the response time acquiring unit Based on the acquired response time, the first calculation unit that calculates the first holding time, and the lock waiting time of the second record of the first holding time calculated by the first calculation unit M / M / 1 queuing model, where the second waiting time and the second holding time that is the lock holding time of the second record are respectively the service waiting time and service time of the M / M / 1 queue model By applying the second stage, the second rate and the third rate acquired by the rate acquisition unit, and the first holding time calculated by the first calculation unit, the second rate is obtained. Maintenance A second calculation unit for calculating a time to function as an output unit for outputting a second retention time calculated by the second calculation unit, also provides programs.

  According to the present invention, when it is necessary to lock two records in one transaction, it is possible to obtain a processing time excluding the transaction lock waiting time for achieving the target performance.

It is a figure for demonstrating the calculation method of the transaction processing time for satisfy | filling the target throughput in embodiment of this invention. It is the block diagram which showed the function structural example of the processing time calculation apparatus in embodiment of this invention. It is the flowchart which showed the operation example of the processing time calculation apparatus in embodiment of this invention. It is the figure which showed the specific example of the processing result of the processing time calculation apparatus in embodiment of this invention. It is the figure which showed the hardware constitutions of the computer which can apply embodiment of this invention.

Embodiments of the present invention will be described below in detail with reference to the accompanying drawings.
In this embodiment, the server, service time, and service wait time in the queue theory are used as a lock target record, lock holding time, and lock wait time in lock contention, respectively. A one-queue model (hereinafter simply referred to as “M / M / 1 model”) is applied to lock contention.
Specifically, when it is necessary to hold two record locks in one transaction, the M / M / 1 model is applied to the first lock and the second lock in two stages. Thus, the upper limit of the pure transaction processing time excluding the lock waiting time for obtaining a given transaction throughput is obtained.

Therefore, in this embodiment, first, two records updated in a transaction are assumed to be a record A and a record B, and the transactions are classified into the following three types.
The first type is a transaction type for updating record A and records other than record B. In the present embodiment, this type of transaction is referred to as “transaction [AB]”.
The second type is a transaction type for updating record A and record B. In this embodiment, this type of transaction is referred to as “transaction [A & B]”.
The third type is a transaction type for updating a record other than the record A and the record B. In this embodiment, this type of transaction is referred to as “transaction [BA]”.
In one transaction, in order to avoid the occurrence of deadlock, two records are updated from the one with the smallest record key. For example, for record A and record B, record A has a smaller record key and is updated in the order of record A and record B.

In this embodiment, the records A and B are locked as locks A and B, respectively, and the following requirements are given.
The first requirement is the transaction response time. This includes lock A latency and lock B latency. In this embodiment, this transaction response time is expressed as Tresp.
The second requirement is the transaction rate of the transaction that acquires lock A. In this embodiment, this transaction rate is expressed as Ra.
The third requirement is the transaction rate of the transaction for acquiring lock B. In this embodiment, this transaction rate is expressed as Rb.
The fourth requirement is the transaction rate of the transaction for acquiring lock A and lock B. In this embodiment, this transaction rate is expressed as Rab.
Further, from the third requirement and the fourth requirement, the ratio of transactions that do not acquire lock A among transactions that acquire lock B is obtained. In this embodiment, this ratio is expressed as r. The ratio r is calculated by the following formula.

  FIG. 1 is a diagram schematically showing a method for calculating an upper limit of a pure transaction processing time for satisfying a given transaction throughput. In this embodiment, each transaction is described as updating two records. However, three or more records may be updated. However, it is assumed that no conflict occurs in records other than record A and record B.

  In the figure, a thick arrow 201 indicates that transaction [A-B] is about to acquire lock A. The transaction rate of the transaction [A-B] is Ra-Rab as shown in the thick arrow 201. This transaction [A-B] waits for processing for the waiting time Twa of lock A as indicated by a broken line arrow 221, and then acquires lock A and holds lock A as indicated by a thin arrow 241. Process for time Tsa.

  A thick arrow 202 indicates that transaction [A & B] is about to acquire lock A. The transaction rate of the transaction [A & B] is Rab as shown in the thick arrow 202. This transaction [A & B] waits for the waiting time Twa of the lock A and then acquires the lock A and performs the processing for the holding time Tsa of the lock A, as indicated by the dashed arrow 221. At that time, the transaction [A & B] requests the lock B as indicated by a thin arrow 261 after waiting for the lock B as indicated by a thin arrow 261 after the elapse of time t after acquiring the lock A, and As indicated by the thin arrow 242, the lock B is acquired and processed.

  In the present embodiment, the first stage of the M / M / 1 model is applied to such processing by transaction [AB] and transaction [A & B]. Specifically, the lock A hold time Tsa, the lock A hold time Tsa, and the transaction rate Ra are respectively applied to the service wait time, service time, and arrival rate of the queue theory, and the lock A hold time Tsa is set. calculate. The calculation method of the lock A holding time Tsa will be described in detail later.

  On the other hand, a thick arrow 203 in the figure indicates that transaction [BA] is trying to acquire lock B. The transaction rate of the transaction [B-A] is Rb-Rab as indicated by the thick arrow 203. This transaction [B-A] waits for the lock B waiting time Twb as indicated by the dashed arrow 222, and then acquires the lock B and holds the lock B as indicated by the thin arrow 242. Processing is performed for time Tsb.

  In the present embodiment, the second stage of the M / M / 1 model is applied to such processing by transaction [BA]. Specifically, the lock B holding time Twb, the lock B holding time Tsb, and the transaction rate Rb-Rab are applied to the queuing theory service waiting time, service time, and arrival rate, respectively. Calculate Tsb. The holding time Tsb of the lock B is a processing time after acquiring the locks of the record A and the record B, and is a pure transaction processing time excluding lock waiting. The method for calculating the lock B holding time Tsb will be described in detail later.

FIG. 2 is a block diagram illustrating an example of a functional configuration of the processing time calculation apparatus 10 that calculates the processing time. In the processing time calculation device 10, the CPU 90a (see FIG. 5) of the computer reads the processing time calculation program from, for example, the magnetic disk device 90g (see FIG. 5) into the main memory 90c (see FIG. 5) and executes it. It is realized in a computer.
As illustrated, the processing time calculation device 10 includes a competition requirement acquisition unit 11, a competition analysis unit 12, a competition analysis result storage unit 13, a performance requirement acquisition unit 14, a first analysis unit 15, and a first analysis. The result storage unit 16, the second analysis unit 17, and the second analysis result output unit 18 are provided.

  The competition requirement acquisition unit 11 acquires a requirement (contention requirement) regarding lock contention. Specifically, the transaction rate Ra of the transaction for acquiring the lock A, the transaction rate Rb of the transaction for acquiring the lock B, and the transaction rate Rab of the transaction for acquiring the lock A and the lock B are acquired. These competing requirements may be obtained from information stored in the computer in the magnetic disk device 90g (see FIG. 5) or input by the user using the keyboard / mouse 90i (see FIG. 5). You may acquire from information. In the present embodiment, as an example of a first rate that is a transaction rate for updating at least two records including the first record and a first rate that is a transaction rate for updating the first record, a transaction As an example of a second rate that is a rate of a transaction that uses a rate Ra and that updates a transaction that updates at least two records including the second record, and a rate that a transaction updates the second record, A transaction rate Rb is used, and a third rate, which is a transaction rate for updating at least two records including the first record and the second record, and a transaction rate for updating the first record and the second record. As an example of a third rate that is a rate, Transaction rate Rab is used. Moreover, the competition requirement acquisition part 11 is provided as an example of the rate acquisition part which acquires a 1st rate, a 2nd rate, and a 3rd rate.

The contention analysis unit 12 receives the transaction rate acquired by the contention requirement acquisition unit 11, and of the transactions that do not acquire the lock A among the transactions that acquire the lock B based on the transaction rate, for the transaction that acquires the lock B The ratio r is calculated.
The competition analysis result storage unit 13 stores the transaction rate received by the competition analysis unit 12 and the ratio r calculated by the competition analysis unit 12 as a competition analysis result.

  The performance requirement acquisition unit 14 acquires requirements (performance requirements) related to transaction processing performance. Specifically, the transaction response time Tresp is acquired. This performance requirement may be acquired from information stored in the computer in the magnetic disk device 90g (see FIG. 5), or information input by the user using the keyboard / mouse 90i (see FIG. 5). May be obtained from In the present embodiment, a transaction response time Tresp is used as an example of a target response time of a transaction. Moreover, the performance requirement acquisition part 14 is provided as an example of the response time acquisition part which acquires a target response time.

  The first analysis unit 15 applies the M / M / 1 model as the M / M / 1 model service waiting time and service time using the lock A waiting time Twa and the lock A holding time Tsa as the first stage. M / M / 1 analysis processing is performed. Specifically, the lock A holding time Tsa including the lock B wait time is used by using the transaction rate Ra stored in the competition analysis result storage unit 13 and the transaction response time Tresp acquired by the performance requirement acquisition unit 14. To calculate. In this embodiment, the lock A wait time Twa is used as an example of the first wait time that is the lock wait time of the first record, and the first hold that is the lock hold time of the first record. As an example of the time, the lock A holding time Tsa is used. Moreover, the 1st analysis part 15 is provided as an example of the 1st calculation part which calculates 1st holding time.

  The first analysis result storage unit 16 stores the result of the first stage M / M / 1 analysis process. Specifically, the lock A holding time Tsa calculated by the first analysis unit 15 is stored.

  The second analysis unit 17 applies the M / M / 1 model as the service waiting time and service time of the M / M / 1 model using the lock B waiting time Twb and the lock B holding time Tsb as the second stage. M / M / 1 analysis processing is performed. Specifically, the lock B holding time Tsb allowed under the requirements acquired by the competition requirement acquisition unit 11 and the performance requirement acquisition unit 14 is set to the transaction rate Rb and the ratio r stored in the competition analysis result storage unit 13. And the lock A holding time Tsa stored in the first analysis result storage unit 16. When calculating the lock B holding time Tsb, it is assumed that the time t from the acquisition of the lock A to the request for the lock B is sufficiently small, and t = 0. In this embodiment, the lock B wait time Twb is used as an example of the second wait time that is the lock wait time of the second record, and the second hold that is the lock hold time of the second record. As an example of the time, the holding time Tsb of the lock B is used. Moreover, the 2nd analysis part 17 is provided as an example of the 2nd calculation part which calculates 2nd holding time.

  The second analysis result output unit 18 outputs the result of the second stage M / M / 1 analysis process to the display mechanism 90d (see FIG. 5). Specifically, the lock B holding time Tsb calculated by the second analysis unit 17 is output to the display mechanism 90d (see FIG. 5). In the present embodiment, the second analysis result output unit 18 is provided as an example of an output unit that outputs the second holding time.

FIG. 3 is a flowchart illustrating an operation example of the processing time calculation apparatus 10.
As shown in the figure, in the processing time calculation device 10, first, the competition requirement acquisition unit 11 uses, as the competition requirements, the transaction rate Ra of the transaction for acquiring the lock A, the transaction rate Rb of the transaction for acquiring the lock B, the lock A, The transaction rate Rab of the transaction for acquiring the lock B is acquired (step 101).

そして、競合分析部１２は、ステップ１０１で取得されたトランザクションレートRa、Rbと、ステップ１０２で算出された比率rとを、競合分析結果記憶部１３に記憶する（ステップ１０３）。 Then, based on the transaction rates Rb and Rab acquired in step 101, the contention analysis unit 12 calculates a ratio r of transactions that do not acquire lock A among transactions that acquire lock B to transactions that acquire lock B. Calculate (step 102). Specifically, the ratio r is calculated by performing the following calculation.

Then, the competition analysis unit 12 stores the transaction rates Ra and Rb acquired in step 101 and the ratio r calculated in step 102 in the competition analysis result storage unit 13 (step 103).

Moreover, in the processing time calculation apparatus 10, the performance requirement acquisition part 14 acquires transaction response time Tresp as a performance requirement (step 104).
Then, the first analysis unit 15 performs the lock A based on the transaction rate Ra stored in step 103 and the transaction response time Tresp acquired in step 104 as the first stage M / M / 1 analysis process. The holding time Tsa is calculated (step 105).

これにより、トランザクションレスポンスタイムTrespは、以下のように展開される。

更に式を以下のように展開し、TsaをTrespとRaとから算出する式を求め、この式を用いてロックＡの保持時間Tsaを算出する。

Here, the process of calculating the holding time Tsa of the lock A will be described in detail.
First, the following relationship is established from the M / M / 1 model (ρa is the usage rate of lock A).

As a result, the transaction response time Tresp is expanded as follows.

Further, the equation is expanded as follows, an equation for calculating Tsa from Tresp and Ra is obtained, and the holding time Tsa of lock A is calculated using this equation.

  The first analysis unit 15 stores the lock A holding time Tsa calculated in step 105 in the first analysis result storage unit 16 (step 106).

  Further, in the processing time calculation device 10, the second analysis unit 17 performs the transaction rate Rb and ratio r stored in step 103 and the lock stored in step 106 as the second stage M / M / 1 analysis processing. Based on the holding time Tsa of A, a process of calculating the holding time Tsb of lock B is performed (step 107).

これにより、ロックＡの保持時間Tsaは、以下のように展開される。

更に式を以下のように展開し、TsbをTsaとrとRbとから算出する式を求め、この式を用いてロックＢの保持時間Tsbを算出する。

Here, the process of calculating the holding time Tsb of the lock B will be described in detail.
First, the following relationship is established from the M / M / 1 model (ρb is the utilization rate of lock B).

Thereby, the holding time Tsa of the lock A is developed as follows.

Further, the equation is expanded as follows, an equation for calculating Tsb from Tsa, r, and Rb is obtained, and the holding time Tsb of lock B is calculated using this equation.

  Then, the second analysis result output unit 18 outputs the lock B holding time Tsb calculated in step 107 to the display mechanism 90d (see FIG. 5), and the display mechanism 90d (see FIG. 5) holds the lock B. The time Tsb is displayed (step 108).

Here, the above operation will be described using specific numerical values.
FIG. 4 shows specific numerical values stored in each variable at this time.
First, in step 101, it is assumed that the competition requirement acquisition unit 11 acquires “Ra = 120 transactions / second, Rb = 80 transactions / second, Rab = 30 transactions / second” as the competition requirements.
In step 104, it is assumed that the performance requirement acquisition unit 14 acquires “Tresp = 1.0 second” as the performance requirement.
Then, in step 105, by performing the following calculation, as shown in the “first stage M / M / 1 analysis process” column, from “Tresp = 1.0 seconds” and “Ra = 120 transactions / second”, “ Tsa = 8.3 milliseconds ".

Further, in step 107, by performing the following calculation, as shown in the “second stage M / M / 1 analysis process” column, “Tsa = 8.3 milliseconds”, “Rb = 80 transactions / second”, “ From “Rab = 30 transactions / second”, “Tsb = 5.8 milliseconds” is obtained (indicated by a bold line in the figure).

  From the above, it can be seen that the target transaction throughput can be achieved if the pure transaction processing time without lock waiting is within 5.8 milliseconds.

2 and 3, when calculating the holding time Tsb of the lock B from the holding time Tsa of the lock A, Rab is replaced with a ratio r of Rb-Rab to Rb, and this ratio r is used. Not limited to this. That is, as shown in FIG. 4, Rab may be used as it is.
In that case, the first analysis unit 15 stores the transaction rate Rab in the first analysis result storage unit 16 instead of the ratio r, and the second analysis unit 17 calculates the holding time Tsb of the lock B by the following equation. calculate.

Further, in the above, the first stage of the M / M / 1 model is applied to the processing by the transaction [AB] and the transaction [A & B], and the processing by the transaction [BA] is performed by the M. A second stage application of the / M / 1 model was performed. However, the first stage of the M / M / 1 model is applied to the process by the transaction [AB], and the process by the transaction [BA] and the transaction [A & B] is performed by the M / M / A modification in which the second stage of one model is applied is also conceivable.
In that case, the contention analysis unit 12 calculates the ratio r ′ of transactions that do not acquire lock B out of transactions that acquire lock A to transactions that acquire lock A in step 102 by performing the following calculation. To do.

In step 103, the competition analysis unit 12 stores the ratio r ′ in the competition analysis result storage unit 13 together with the transaction rates Ra and Rb.
In step 105, the first analysis unit 15 calculates the holding time Tsa of the lock A using the following equation.

On the other hand, in Step 107, the second analysis unit 17 calculates the holding time Tsb of the lock B using the following formula.

Furthermore, in the above description, in response to the application of the first stage of the M / M / 1 model and the application of the second stage of the M / M / 1 model, the M / M / 1 analysis process of the first stage and Although the second-stage M / M / 1 analysis process is performed, the present invention is not limited to this. That is, the following expression is obtained from the expression used in the first stage M / M / 1 analysis process and the expression used in the second stage M / M / 1 analysis process, and one M / M is performed using this expression. / 1 analysis processing may be performed.

  Furthermore, the formula used in the present embodiment is merely an example, and the formula is not necessarily used. Any expression that obtains the same output variable based on the same input variable and has similar properties can be variously replaced.

  Thus, in this embodiment, when it is necessary to hold two record locks in one transaction, the M / M / 1 model is applied to the first lock and the second lock in two stages. Applied. As a result, it is possible to obtain an upper limit of pure transaction processing time excluding lock waiting for giving a given transaction throughput.

  Finally, a hardware configuration of a computer suitable for applying this embodiment will be described. FIG. 5 is a diagram showing an example of the hardware configuration of such a computer. As shown in the figure, the computer includes a CPU (Central Processing Unit) 90a which is a calculation means, a main memory 90c connected to the CPU 90a via an M / B (motherboard) chip set 90b, and an M / B chip set 90b. And a display mechanism 90d connected to the CPU 90a. Further, a network interface 90f, a magnetic disk device (HDD) 90g, an audio mechanism 90h, a keyboard / mouse 90i, and a flexible disk drive 90j are connected to the M / B chip set 90b via a bridge circuit 90e. Has been.

  In FIG. 5, each component is connected through a bus. For example, the CPU 90a and the M / B chip set 90b, and the M / B chip set 90b and the main memory 90c are connected via a CPU bus. Further, the M / B chipset 90b and the display mechanism 90d may be connected via an AGP (Accelerated Graphics Port). However, if the display mechanism 90d includes a video card compatible with PCI Express, the M / B The chip set 90b and this video card are connected via a PCI Express (PCIe) bus. When connecting to the bridge circuit 90e, for example, PCI Express can be used for the network interface 90f. For the magnetic disk device 90g, for example, serial ATA (AT Attachment), parallel transfer ATA, or PCI (Peripheral Components Interconnect) can be used. Furthermore, USB (Universal Serial Bus) can be used for the keyboard / mouse 90i and the flexible disk drive 90j.

  Here, the present invention may be realized entirely by hardware or entirely by software. It can also be realized by both hardware and software. The present invention can be realized as a computer, a data processing system, and a computer program. This computer program may be stored and provided on a computer readable medium. Here, the medium may be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system (apparatus or equipment), or a propagation medium. Examples of computer-readable media include semiconductors, solid state storage devices, magnetic tape, removable computer diskettes, random access memory (RAM), read-only memory (ROM), rigid magnetic disks, and optical disks. The Current examples of optical disks include compact disk-read only memory (CD-ROM), compact disk-read / write (CD-R / W) and DVD.

  As mentioned above, although this invention was demonstrated using embodiment, the technical scope of this invention is not limited to the said embodiment. It will be apparent to those skilled in the art that various modifications and alternative embodiments can be made without departing from the spirit and scope of the invention.

DESCRIPTION OF SYMBOLS 10 ... Processing time calculation apparatus, 11 ... Competition requirement acquisition part, 12 ... Competition analysis part, 13 ... Competition analysis result storage part, 14 ... Performance requirement acquisition part, 15 ... 1st analysis part, 16 ... 1st analysis result storage part , 17 ... second analysis unit, 18 ... second analysis result output unit

Claims (8)

An apparatus for calculating a processing time excluding a transaction lock waiting time for achieving a target transaction performance,
A first rate is the rate of transaction that updates at least two records containing the first record, the second rate is a rate of a transaction that updates at least two records containing the second record, the first a transaction that updates at least two records that contain one record and said second record, said for the first record and the second record, the order of the first record, the second record A rate acquisition unit that acquires a third rate that is the rate of the transaction to be updated in
A response time acquisition unit for acquiring a target response time of the transaction;
Of the response times acquired by the response time acquisition unit, a first waiting time that is a lock waiting time of the first record and a first holding time that is a lock holding time of the first record, respectively. By applying the first stage of the M / M / 1 queuing model as the service waiting time and service time of the M / M / 1 queuing model, the first rate acquired by the rate acquiring unit and A first calculation unit that calculates the first holding time based on the response time acquired by the response time acquisition unit;
Of the first holding time calculated by the first calculating unit, the second waiting time that is the lock waiting time of the second record and the second holding time that is the lock holding time of the second record By applying the second stage of the M / M / 1 queuing model with the holding time as the service waiting time and service time of the M / M / 1 queuing model, respectively, the first time acquired by the rate acquisition unit A second calculating unit that calculates the second holding time based on the rate of 2 and the third rate, and the first holding time calculated by the first calculating unit;
An output unit that outputs the second holding time calculated by the second calculation unit. The first calculation unit has an equation

To calculate the first holding time,
The second calculation unit has the formula

The apparatus of claim 1, wherein Ra is the first rate, Rb is the second rate, and Rab is the third rate. Tresp is the response time, Tsa is the first holding time, and Tsb is the second holding time). An apparatus for calculating a processing time excluding a transaction lock waiting time for achieving a target transaction performance,
A first rate is the rate of transactions to update the first record, the second rate is a rate of a transaction that updates the second record, the first record and the second record, the A rate acquisition unit that acquires a first rate and a third rate that is a rate of a transaction to be updated in the order of the second record ;
A response time acquisition unit for acquiring a target response time of the transaction;
The first waiting time that is the lock waiting time of the first record and the first holding time that is the lock holding time of the first record are respectively expressed as the service waiting time and service time of the M / M / 1 queue model. The first stage of the M / M / 1 queuing model is applied, and the second waiting time that is the lock waiting time of the second record and the second holding time of the lock of the second record are By applying the second stage of the M / M / 1 queuing model with the holding time as the service waiting time and service time of the M / M / 1 queuing model, respectively, the first time acquired by the rate acquisition unit A calculation unit that calculates the second holding time based on the rate of 1, the second rate, the third rate, and the response time acquired by the response time acquisition unit ,
An output unit that outputs the second holding time calculated by the calculation unit. The calculation unit has the formula

The apparatus of claim 3, wherein Ra is the first rate, Rb is the second rate, and Rab is the third rate. Tresp is the response time, and Tsb is the second holding time). The calculation unit includes:
Based on the first rate acquired by the rate acquisition unit and the response time acquired by the response time acquisition unit by applying the first stage of the M / M / 1 queue model A first calculation unit for calculating the first holding time;
By applying the second stage of the M / M / 1 queue model, the second rate and the third rate acquired by the rate acquisition unit, and the first calculation unit are calculated. 5. The apparatus according to claim 3, further comprising: a second calculation unit that calculates the second holding time based on the first holding time. The calculation unit includes:
By applying the first stage of the M / M / 1 queuing model, the first rate and the third rate acquired by the rate acquisition unit, and the response time acquisition unit A first calculation unit for calculating the first holding time based on the response time;
By applying the second stage of the M / M / 1 queue model, the second rate acquired by the rate acquisition unit and the first holding calculated by the first calculation unit 5. The apparatus according to claim 3, further comprising: a second calculation unit configured to calculate the second holding time based on time. A method in which a computer calculates a processing time excluding a transaction lock latency to achieve a target transaction performance,
Said computer comprises a first rate is the rate of transactions to update the first record, the second rate is a rate of a transaction that updates the second record, the first record and the second Obtaining a third rate that is a rate of a transaction for updating the records in the order of the first record and the second record ;
The computer obtaining a target response time for the transaction;
The computer sets a first waiting time that is the lock waiting time of the first record and a first holding time that is the lock holding time of the first record, respectively, to the service waiting of the M / M / 1 queue model. As the time and service time, the first stage of the M / M / 1 queuing model is applied, and the second waiting time that is the lock waiting time of the second record and the lock holding time of the second record By applying the second stage of the M / M / 1 queuing model with the second holding time as the service waiting time and service time of the M / M / 1 queuing model, respectively, the first rate and Calculating the second holding time based on the second rate, the third rate, and the response time;
And outputting the calculated second hold time of the computer. A program for causing a computer to function as a device for calculating a processing time excluding a transaction lock waiting time for achieving a target transaction performance,
The computer,
A first rate is the rate of transactions to update the first record, the second rate is a rate of a transaction that updates the second record, the first record and the second record, the A rate acquisition unit that acquires a first rate and a third rate that is a rate of a transaction to be updated in the order of the second record ;
A response time acquisition unit for acquiring a target response time of the transaction;
Of the response times acquired by the response time acquisition unit, a first waiting time that is a lock waiting time of the first record and a first holding time that is a lock holding time of the first record, respectively. By applying the first stage of the M / M / 1 queuing model as the service waiting time and service time of the M / M / 1 queuing model, the first rate acquired by the rate acquiring unit and A first calculation unit that calculates the first holding time based on the response time acquired by the response time acquisition unit;
Of the first holding time calculated by the first calculating unit, the second waiting time that is the lock waiting time of the second record and the second holding time that is the lock holding time of the second record By applying the second stage of the M / M / 1 queuing model with the holding time as the service waiting time and service time of the M / M / 1 queuing model, respectively, the first time acquired by the rate acquisition unit A second calculating unit that calculates the second holding time based on the rate of 2 and the third rate, and the first holding time calculated by the first calculating unit;
A program that functions as an output unit that outputs the second holding time calculated by the second calculation unit.

Images