JP5101140B2 - System resource control apparatus and control method - Google Patents

Description

  The present invention relates to a system resource control apparatus and control method, and in particular, a system resource that controls a non-uniformity representing a degree of variation in resource utilization of a system as a control target having a plurality of system resources to a predetermined neighborhood value. The present invention relates to a control device and a control method.

  In recent years, with the advancement of Internet-related technologies such as web services, it has become possible to remotely monitor the operation of a computer system at a practical cost. For this reason, it is becoming possible to collect system operation monitoring performed for each site having a computer system and to remotely monitor a plurality of sites collectively. By doing so, it is possible to reduce the operation monitoring personnel deployed for each site and anticipate a reduction in operation costs, so that the integration of sites is in the direction of progress.

  However, if a simple personnel reduction is performed, the monitoring load per operation monitoring person increases, and the possibility of overlooking a malfunction of the computer system increases. Therefore, a method is conceivable in which monitor values obtained from the computer system to be monitored are integrated in some form, an index indicating the operation state is derived, and automatic control is performed so that this index is kept in the vicinity of a predetermined value. In order to realize this method, it is necessary to calculate an index and a control value indicating an operation state at high speed. That is, it is a technical problem to provide a method for calculating the index and the control value in a short time.

  Examples of the monitoring / control target of the computer system include utilization rates of resources such as a CPU, a memory, and a disk. Usually, the resource utilization rate can be handled by a normalized numerical value from 0.0 to 1.0. The degree of variation in these values is a kind of index indicating the state of the computer system.

Usually, dispersion is used as the degree of variation. When calculating the variance, if parallel calculation can be performed, the processing time can be shortened. For this purpose, as described in Patent Document 1, it is necessary that the definition of the variance is given by the recurrence formula, but the variance cannot be given by the recurrence formula. Further, the dispersion, as disclosed in Non-Patent Document 1, since represented by a quadratic polynomial is indicative of essentially linear. For this reason, it is inevitable to calculate the control value in a short time because it is inevitable to calculate the resource utilization rate of the computer system that realizes a predetermined distributed value and to obtain the control value from it. .
JP 11-39272 A Haruhiko Okumura, "The latest algorithm encyclopedia in C language" 1991

  As described above, even if the non-uniformity indicating the degree of variation in the system resource utilization rate is controlled to a predetermined neighborhood value using the conventional technology, iterative calculation is inevitable, and the control value is reduced for a short time. Thus, it is difficult to control the computer system.

  In view of the above-described points, the object of the present invention is to enable the non-uniformity representing the degree of variation to be calculated in a short time from the resource usage rates of a plurality of computer systems, and to reduce the resource usage rates of the computer systems. It is an object of the present invention to provide a system resource control device and a control method capable of controlling the uniformity to a predetermined neighborhood value.

According to the present invention, an object of the present invention is to provide a system resource control device that controls the nonuniformity of the system resource utilization rate of a plurality of control target devices having system resources to a predetermined value. means for collecting, inhomogeneity representing the degree of variation of the collected plurality of system resources utilization, sequentially obtains a difference from the outside of the plurality of system resource utilization for the data string arranged in ascending or descending order, Summing up the differences, dividing the total value by the number of pairs, and changing the values of the plurality of system resource utilization rates does not change the order of the data strings in which the utilization rates are arranged. in the range, the system re-determining said change in the non-uniformity in the case of changing the values of a plurality of system resource utilization as a predetermined displacement of the non-uniformity Using means for calculating the inhomogeneity over scan utilization, inhomogeneity, the predetermined control priority rule definition file for a single record with Resource give inhomogeneity displacement and displacement of the target, When the resource usage rate is rearranged in ascending or descending order, the displacement value that can change the resource usage rate of the control target is calculated so that the order of adjacent data does not change. The upper and lower limit values of the nonuniformity displacement when the system resource utilization value is changed are obtained, and it is determined whether or not the predetermined displacement of the nonuniformity is within the upper and lower limit values of the nonuniformity displacement. a determination is predetermined displacement realization that, when a predetermined displacement was in the range of upper and lower limit values, and means for calculating the system resource utilization control target value by calculating a resource utilization displacement, the sheet The system resource utilization rate non-uniformity is calculated by rearranging the collected system resource utilization rates in ascending or descending order, and when the total number of system resource utilizations is a power of 2, 2 is smaller than the total number. each as a group by dividing a sequence of resource utilization a sequence of power-number of resource utilization units perform parallel processing by group, and calculates the degree of non-uniformity to the system resource utilization by integrating the results as sorted also by changing the values of said plurality of system resources utilization by arranging the utilization data string does not change, when obtaining by inverse calculation multiple system resource utilization for a given non-uniformity within a range not changing the order of arrangement of the system resource utilization, when reducing the inhomogeneity predetermined displacement, its the group In each of the above groups, if the first half of the system resource utilization is decreased by a certain displacement, the subsequent half of the system resource utilization is increased by a certain displacement , and the non-uniformity is increased by a predetermined displacement, only the first is a system resource utilization of the half displacement is reduced, by a displacement in a system resource utilization of the subsequent half increased, a combination of linear sum of the displacement of the system resource utilization by combining a predetermined displacement of the non-uniformity system This is achieved by calculating the non-uniformity of resource utilization .

  According to the present invention, the nonuniformity value of the resource usage rate of the computer system to be controlled can be calculated at high speed by parallel processing, and the nonuniformity value of the resource usage rate in the controlled system can be calculated as a predetermined neighborhood value. Can be controlled.

  Embodiments of a system resource control apparatus and control method according to the present invention will be described below in detail with reference to the drawings. The embodiment of the present invention described below calculates the nonuniformity of the utilization rate of resources used by an information system composed of a plurality of servers and network devices connected by a network such as a LAN or WAN. It is possible to control once to a predetermined neighborhood value.

  FIG. 1 is a block diagram showing a system configuration including a system resource control apparatus according to an embodiment of the present invention.

  The system shown in FIG. 1 includes a system resource control device 110 according to an embodiment of the present invention and a plurality of control target devices 102 and 108 such as servers connected to a network 101.

  The control target device 102 acquires information regarding the network interface 103 that transmits resource data and receives control target values of resource utilization, and system resources 106 such as a CPU, memory, and disk in the control target device such as a server. A resource data providing agent 104 represented by an SNMP agent, a system resource 160 such as a CPU, a memory, and a disk in a control target device, and a resource pool 107 in which an unused resource is secured for each type of the system resource 106 The system resource 106 and the resource pool 107 are configured to include a system resource utilization rate control agent 105 for exchanging the same type of resources.

  Although FIG. 1 shows that five control target devices are connected to the network 101 as the control target device 102 and the other control target device 108, the number is not limited, and In the embodiment, the entire control target apparatus is controlled as a control target system 109. The other control target device 108 is the same as the control target device 102 although its internal configuration is not shown.

  The system resource control device 110 includes a network interface 111 that receives resource data and transmits a control target value for resource utilization, a system resource data collection program 112, a system resource utilization nonuniformity calculation program 113, a system A resource utilization rate control target value calculation program 114 and a system resource utilization rate control program 115 are provided. The system resource control apparatus 110 includes a virtual resource 116 and a virtual resource pool 117, which are virtual resources obtained by integrating all the system resources 106 and the resource pool 107 in the control target system 109. It is controlled by the system resource utilization rate control program 115. The controlled result is immediately transmitted to the control agents 105 of all system resource usage rates in the control target system 109 and reflected in the system resource 106 and the resource pool 107.

  The system resource control device 110 also includes a dummy insertion rule definition file 118 that records information that is referred to by the system resource utilization rate non-uniformity calculation program 113 and information that is referenced by the control target value calculation program 114 of the system resource utilization rate. The control priority rule definition file 119 in which the system resource data is recorded and the system resource data storage database 120 for storing the data collected by the system resource data collection program 112 are included.

  Further, a system resource monitoring apparatus 121 is connected to the system resource control apparatus 110, and the nonuniformity calculated by the system resource utilization ratio nonuniformity calculation program 113 and a control target value calculation program for the system resource utilization ratio are calculated. The control target value calculated by 114 and the result of controlling the system resource control target system 109 using this control target value are transferred to the system resource monitoring device 121, and the system resource utilization non-uniformity display program 122 Is displayed. Note that the system resource monitoring apparatus 121 may be integrally configured in the system resource control apparatus 110.

  In the above description, the control target devices 102 and 108, the system resource control device 110, and the system resource monitoring device 121 are so-called information processing devices, which are not illustrated, but are input devices such as a CPU, a main memory, a disk device, a keyboard, and a mouse, It is configured with hardware such as a display device and a communication device. The network interface 103 in the control target devices 102 and 108, the agents 104 and 105, the network interface 111 in the system resource control device 110, the programs 112 to 115, the virtual resource 116, the virtual resource pool 117, the files 118 and 119, The database 120 and the program 122 in the system resource monitoring device 121 may be stored in a disk device included in each device. The programs and agents necessary for processing are loaded into the main memory and executed by the CPU at the time of processing operation, thereby realizing their functions.

  FIG. 2 is a diagram showing a configuration of system resource data stored in the system resource data storage database 120. The system resource data has a data format in which a resource ID 201 for identifying each resource, a resource name 202, a device ID 203 for identifying a control target device to which the resource belongs, a utilization value 204 at an observation time, and an observation time 205 are columns. It consists of multiple data strings.

  FIG. 3 is a flowchart for explaining the outline of the processing operation in the system resource control apparatus according to the embodiment of the present invention. Next, this will be described. Details of each process will be described later individually.

(1) First, the system resource data collection program 112 collects the resource utilization rate of each device from the control target system 109, normalizes the collected utilization rate to 0 and the maximum value to 1, and calculates the result. Is stored in the system resource data storage database 120 (steps 301 and 302).

(2) Next, the resource utilization rate non-uniformity calculation program 113 calculates the resource utilization rate non-uniformity and determines whether or not the non-uniformity is within a predetermined range (steps 303 and 304). ).

(3) If it is determined in step 304 that the non-uniformity is within a predetermined range, it is determined whether there is a monitoring end command from the monitoring staff. If there is a monitoring end command, If not, the process returns from step 301 to continue the process (step 307).

(4) If it is determined in step 304 that the non-uniformity is not within the predetermined range, the system resource usage rate control target value calculation program 114 calculates the resource usage rate control target value (step 305).

(5) Next, the system resource utilization rate control program 115 determines and controls the resource utilization rate for the virtual resource 116, and returns to the processing from step 301 to continue the processing (step 306).

  Next, the definition and nature of the non-uniformity will be described, and a specific example of a non-uniformity calculation method and a control device will be described. Note that the non-uniformity calculation method can be applied to other than the resource utilization rate as long as it is a numerical value normalized between the minimum value 0 and the maximum value 1. Therefore, in the following, simply the data or Expressed as a data string.

  FIG. 4 is a diagram for explaining an example of a method for determining the non-uniformity when the number of data is eight. In the case of the example shown in FIG. 4, for the data string 401 (x1 to x8) in which 8 pieces of data are arranged in ascending order, differences are taken sequentially from the outside, and finally those differences are summed, and the total value Is divided by the number of pairs, and the obtained value is considered as non-uniformity. The reason for dividing by the number of pairs is to normalize the non-uniformity value to a range from 0 to 1. This non-uniformity value indicates how scattered the data is, and may be considered as an index indicating the degree of variation of a kind of data.

  FIG. 5 is a diagram for explaining the method of determining the non-uniformity in the example shown in FIG. 4 from another viewpoint. In the example shown in FIG. 5, the value of the data shown in FIG. 4 is represented by the height of the rectangle, and the rectangle is arranged in the order of height as shown by 501, and the difference in height between the rightmost rectangle and the leftmost rectangle is taken, The same calculation is sequentially performed on the inner rectangle. Even in such a method, the value as the nonuniformity can be obtained as in the case of FIG. Assuming that the data strings x1 to x8 are data strings in increments of 0.1 from 0.1 to 0.8, the value s1 as the non-uniformity can be obtained by calculation using an expression such as 502. Can do.

  FIG. 6 is a diagram for explaining the method of determining the non-uniformity in the example shown in FIG. 5 from another viewpoint. The example shown in FIG. 5 is equivalent to performing the operation shown in FIG. That is, the data values in the initial state are represented by the height of the rectangle, the rectangles are arranged in the order of height as indicated by 601, and the first four data strings (parts indicated by dots) are centered on the midpoint of the data string. It is reversed and overlapped with the subsequent four data strings, and an operation of cutting the overlapping portion (the portion shown in gray) from the original rectangular portion is performed. By performing the above operation, the heights of the remaining portions (portions shown in white) are summed, and the obtained value is divided by the number of pairs to obtain a value as the nonuniformity. Also in this case, the value s1 as the non-uniformity is the same as that shown in FIG.

  FIG. 7 is a diagram for explaining a method of determining the non-uniformity when the number of data is eight by repeating the operation described with reference to FIG. The same operation as described with reference to FIG. 6 can be performed twice more as shown in FIG.

  That is, first, similarly to the case shown in FIG. 5, as shown in the initial state 701 in FIG. 7A, the data values are represented by the heights of the rectangles, and the rectangles are arranged in the height order. Next, similarly to the case shown in FIG. 6, as shown as the first difference 702 in FIG. 7B, the first four data strings (portions indicated by dots) are inverted and the subsequent four data strings are inverted. An operation is performed in which the overlapping portion (the portion shown in gray) is cut out from the original rectangular portion. For the remaining four portions (portions shown in white), the first two data strings (portions shown by dots) are inverted as shown in FIG. Are overlapped with the two data strings, and an operation of cutting the overlapped portion (the portion shown in gray) from the original rectangular portion is performed. Further, for the remaining two portions (portions shown in white), as shown by the third difference 704 in FIG. An operation of superposing one piece of data and cutting the overlapped portion (shown in gray) from the original rectangular portion is performed. Thereafter, the height as the last remaining portion is summed and divided by the number of pairs, thereby obtaining a value as the nonuniformity.

  In the present invention, the numerical value obtained by imposing the difference as described above on the data string is considered as non-uniformity. This way of thinking is to measure the degree of variation based on how many rectangular parts remain after performing the above-mentioned difference-taking operation k times when there are 2k data. The idea is to consider how much the data string will hold for a kind of additional operation called difference and the value will not become 0 as an index.

  Next, based on the method of repeating the difference as described above, a specific method for calculating the non-uniformity and a definition formula will be described. Note that the data may be in descending order. In that case, since the obtained value is negative, the sign is inverted to make it nonuniformity. In the following description, it is assumed that the order is ascending.

  FIG. 8 is a diagram for explaining a non-uniformity calculation method and a definition formula when the number of data is 16 according to the method described with reference to FIG.

  According to the method described with reference to FIG. 7, the difference between the maximum data and the minimum data is calculated for the data string 801 (x1 to u4) arranged in ascending order from the top, and then the second largest data and the minimum data are obtained. The process of taking the difference with the next smaller data is continued until there are no more pairs. Then, the differences are arranged in ascending order in the second column. This data string is also paired in the same manner, and the differences are arranged in ascending order in the third column. Further, the third data column is similarly paired, and the differences are arranged in ascending order in the fourth column. These values are summed, and the value obtained by dividing by the number of pairs is defined as the non-uniformity. The reason for dividing by the number of pairs is to normalize the non-uniformity as a numerical value from 0 to 1. Each difference is always positive, and the non-uniformity is never negative. This is clear from the operation of FIG. 7 if the data string is in ascending order. Then, the non-uniformity definition formula S2 (16) when the number of data is 16 is as shown as 803.

9, FIG. 10 and FIG. 11 are diagrams showing the non-uniformity calculation method and the definition formula when the number of data is 8, 4, and 2, respectively, according to the method described with reference to FIG. The calculation method is the same as that when the number of data shown in FIG. 8, 9, and 10, the number of data as shown as Expression (1) from Expressions 803, 903, and 1003 of the non-uniformity when the number of data is 16, 8, and 4, It is possible to derive a relational expression that holds between the nonuniformities in the case of eight or four. This relational expression indicates that the non-uniformity of the number of data of 16 is the average value of the non-uniformity of the first 8 data and the non-uniformity of the subsequent 8 data. It is shown that the data is divided into four pieces, and is equal to the average value of the non-uniformity of the four groups. This relationship can be generalized when the number of data is 2 to the k-th power (k: natural number), and the result is the result of the fact that the number of data is 2 to the k-th power as shown in Equation (2). The degree of uniformity indicates that the number of data is an average value of 2 (k−1) power non-uniformities. In equation (2), the subscripts U and D indicate the first half and the subsequent half of the data string, respectively.

  Further, from the nonuniformity definition formulas 803, 903, and 1003 shown in FIG. 8, FIG. 9, and FIG. 10, when the number of data is four or more, the nonuniformity is the average value of the even-numbered data and the odd-numbered data. It is also shown that the difference is from the average value of the data. From this result, it can be said that the non-uniformity is also an index indicating the degree of variation.

  Now, in FIG. 7, at the stage where the second difference 703 has been taken, a numerical value obtained by summing the results and dividing by the number of pairs is also obtained as the non-uniformity.

  12, 13, 14, and 15, in the example shown in FIG. 7, when the second difference 703 has been obtained, the results are totaled and divided by the number of pairs to obtain a value as the nonuniformity. It is a figure which shows the method to obtain | require and a definition formula about the example in case the number of data is 16, 8, 4, and 2 respectively.

12, 13, and 14, when the number of data is four or more, the non-uniformity is 3 in the multiples of 4 and the multiples of 4. The average value of the data with the number obtained by adding 1 is obtained, the average value of the data with the number obtained by adding 1 to a multiple of 4 and the data of the number obtained by adding 2 to the multiple of 4 is obtained, and the difference between these two average values is obtained. It can also be obtained as Note that when the number of data is two, such a remainder system modulo 4 cannot be considered, so this rule does not hold. From this result, it can be said that the non-uniformity defined as described above is also an index indicating the degree of variation. Further, similarly to the expressions (1) and (2), Expression (3) indicating the relationship that holds between the non-uniformities when the number of data is 16, 8, and 4, and the number of data is 2 The relationship shown as the equation (4) indicating that the k-th nonuniformity is the average value of the (k-1) th nonuniformity of the number of data 2 is also established. Here, considering the non-uniformity as shown by 1503 in FIG. 15, even if the number of data is two, the formula (4) is established, but the non-uniformity when the number of data is two. Since it is unreasonable that at least four pieces of data are required to define one, we will remove it.

  The results described in FIGS. 4 to 15 are summarized as shown in Table 1, which is a nonuniformity classification table based on the basic data length and the number of data. In Table 1, * represents data and is arranged in ascending order from the left side. The symbol in the table shows the average value between the data with arrows on the + side (upper side), the average value between the data with arrows on the-side (lower side), and the average value from the + side It shows that the average value on the minus side is subtracted. This is a schematic representation of definition formulas 803, 903, and 1003 for non-uniformity and definition formulas 1203, 1303, and 1403. The difference 1 time, the difference 2 times, the difference 3 times, etc. are respectively obtained from the results obtained at the stage of completing the first difference, the second difference, and the third difference in FIG. , Meaning to define non-uniformity. Table 1 also shows the direction of increasing the number of differences using a large arrow. The basic data length can be reduced to the shortest data string (basic data string) by recursively using the relational expressions shown in equations (2) and (4), but the minimum data Refers to the length of the column.

In this Table 1, when the basic data length is 2, what calculation is performed when the number of data is 2, 4, 8, 16 and when the basic data length is 4, What calculation is performed when the number is 4, 8, 16 and what calculation is performed when the number of data is 8, 16 when the basic data length is 8 When the basic data length is 16, the calculation is performed in the horizontal direction when the number of data is 16. When viewing Table 1 in the horizontal direction, the number of differences described in the table is ignored.

  Looking at Table 1 in the vertical direction, it can be seen that as the number of differences increases, the non-uniformity shifts from the difference between the average data values at both ends to the difference between the average values for data at more dispersed positions. From this, it can be said that a larger number of differences is more appropriate as a definition of non-uniformity. Further, looking at Table 1 in the horizontal direction, it can be seen that this reaches the basic data string by sequentially using the relational expressions shown in Expression (2) and Expression (4).

  Here, it should be noted that the definition of the non-uniformity changes when the basic data length is changed. Therefore, when paying attention to the non-uniformity change with time, the basic data length determined at the beginning should not be changed halfway. As a criterion for determining the basic data length, there is a method of using a ratio of a minimum value to a maximum value of data included in a data string. If this ratio is close to 1, the minimum value and the maximum value are close, and the degree of variation is small. Therefore, the basic data length is increased, that is, the difference operation as shown in FIG. In this case, since the degree of variation is small, it is not necessary to perform many differential operations, and the non-uniformity becomes a small value, indicating that the degree of variation is small. On the contrary, if the ratio is close to 0, the minimum value and the maximum value of the data string are far from each other, so it is appropriate to calculate the non-uniformity by reducing the basic data length.

  The relational expressions shown in Expression (2) and Expression (4) are the grounds that parallel calculation is possible as calculation of non-uniformity in the present invention. When performing parallel calculation, there is an arbitrary choice as to whether a basic data string or a longer data string is used. In any case, the relational expressions of Equations (2) and (4) claim that the non-uniformity is not calculated directly from the given data sequence, but is inhomogeneous with a shorter data sequence. If we calculate once and take the average of them, it is equal to the non-uniformity of the original data.

  FIG. 16 is a diagram for explaining the efficiency of parallel computation when the number of data is 16. In the example shown in FIG. 16, it is assumed that 16 data 1601 arranged in ascending order are subjected to parallel calculation based on the relational expression shown in Expression (4) with a basic data length of 4. At this time, the first group 1602 has the same value for all data. In this case, the nonuniformity of the first group 1602 is zero. Therefore, if all the data is equal in a certain group, the non-uniformity of that group becomes zero, and there is no need to calculate. As a result, more efficient parallel computation becomes possible. Actually, when all the data in the group is within a predetermined allowable range, they are regarded as equal.

  Next, the nonuniformity definition formulas 1203, 1303, 1403, and 1503 shown in FIGS. 12, 13, 14, and 15 will be described as representative examples of the case where the basic data length is 4, as representative examples. To do.

  First, the relationship between this nonuniformity and dispersion will be described. For this purpose, it is convenient to consider separately when the number of data is an even number and an odd number. First, a case where the number of data is four as an even number of representatives, and a case where the number of data is three as an odd number of representatives will be described.

  FIG. 17 is a diagram for explaining the non-uniformity calculation method and the definition formula in the case where the number of data is four from a different viewpoint.

  Up to this point, in order to normalize the non-uniformity value to a numerical value from 0 to 1, it has been divided by the number of pairs, but this will be reviewed from another point of view. When the number of data is 4, the numerator of nonuniformity is minimized when all the data are equal, and the value is zero. The numerator with the non-uniformity is maximized when the first two data are 0 and the following two data are 1, and the maximum value of the numerator is 2. Therefore, it is sufficient to divide by this maximum value for normalization. Thus, the number of pairs can be regarded as the maximum value of the numerator in the calculation formula of the non-uniformity.

  FIG. 18 is a diagram showing a nonuniformity calculation method and definition formula when the number of data is three. In this example, the normalization of the nonuniformity is also reviewed from another viewpoint, and the normalization in this case also uses the maximum value of the numerator of the nonuniformity. However, in this example, one piece of data that cannot be paired remains, but it is paired with an average value of three pieces of data. The reason for this will be described later. The numerator of non-uniformity is minimized when all the data are equal and its value is zero. The numerator of non-uniformity is maximized when the first two data are 0 and the following one data is 1, or the first one data is 0 and the following two data are 1. This is when the data is divided into 0 and 1 and the difference between the numbers is 1. At this time, the maximum value of the non-uniformity numerator is a value obtained by adding the number of pairs to the value obtained by dividing the number of pairs by the number of data. Note that the number of pairs in this example is the number excluding one set paired with the average value. When the number of data is 3, the number of pairs is defined in this way.

Now, the degree of variation is defined separately from the non-uniformity. In other words, for the data arranged in ascending order or descending order, the difference between the largest data and the smallest data is taken, and the difference between the second largest data and the smallest data is taken next to the smallest data. This is performed until there are no more pairs. If the number of data is an even number, the difference is defined as the sum of the differences divided by the number of pairs. If the number of data is an odd number, one piece of data that cannot be paired remains. Divided by the value added with is defined as variation. When the number of data is an odd number, the number of pairs is the number excluding one set paired with the average value. Expressions for defining the degree of variation in the case of even number (2m) and odd number (2m + 1) data are as shown in Expression (5) and Expression (6), respectively. In equations (5) and (6), m represents the number of pairs.

  Looking at the definition formulas shown in equations (5) and (6), the non-uniformity when the number of data is 4 is equal to the variation when the number of data is 4, and the number of data is 3 Is equal to the degree of variation when the number of data is three.

However, the nonuniformity when the number of data shown in FIGS. 12 and 13 is 16 or 8 is not the same as the degree of variation when the number of data is 16 or 8. This can also be understood from the fact that the non-uniformity on the diagonal line in Table 1 is equal to the variation degree. Here, using FIG. 17 and FIG. 18, when calculating the relationship between the degree of variation when the number of data is four, the degree of variation when the number of data is three, and the respective variances, Equation (7) is obtained. Obtainable. In the example illustrated in FIG. 18, the reason why the data remaining without being paired is paired with the average value is to derive the relationship with such variance. This relational expression can be generalized when the number of data is n, and the relational expression between the lower limit value of the variation degree and the variance is as shown in Expression (8). Furthermore, as shown in Expression (9) including Expression (9.1) to Expression (9.5), a relational expression between the upper and lower limit values of the variation degree and the variance can also be obtained.

  The non-uniformity and the degree of dispersion when the number of data is 16 are expressed by the equation (10) including the equations (10.1) to (10.4) with respect to the variance ( 10.2) and formula (10.3). It can be seen that the variation width with respect to the dispersion is smaller in the nonuniformity than in the variation. Formula (10.4) is generalized when the number of data is 2 to the power of k. From this formula, the fluctuation range for the variance is constant regardless of the number of data. It can be seen that the degree of variation increases as the number of data increases. Accordingly, it can be said that the non-uniformity behaves closer to the dispersion than the variation degree.

When considering the non-uniformity by the method described with reference to FIGS. 4 and 5, the degree of variation is considered. If the degree of variation is used, the degree of variation in the data string is evaluated with a large dispersion value as described above, which is not preferable. As described with reference to FIG. 7, the non-uniformity is a method of performing an evaluation suitable for the number of data by increasing the number of differences with respect to the data as the number of data increases. That is, the number of differences is increased so that the fluctuation range with respect to dispersion is constant. On the other hand, the variation range for the variance increases as the number of data increases because the number of differences does not increase even if the number of data increases. Thus, the non-uniformity and the variation degree have contrasting properties. And if the method shown to Formula (10) is used, the result of having evaluated the upper and lower limits of the nonuniformity by dispersion | distribution can be calculated with respect to various nonuniformities. This is shown in Table 2, which is a table showing the upper and lower limit values of variance and non-uniformity for each basic data length.

  From Table 2, the upper and lower limit values of the non-uniformity do not change even if the number of data increases if the basic data length is the same, and the range of the upper and lower limit values gradually expands as the basic data length increases. I understand. From this, it can be seen that it is more preferable to use the non-uniformity where the basic data length is not increased as much as possible.

  FIG. 19 is a diagram for explaining a non-uniformity calculation method when the number of data is 16, and a mechanism for explaining the non-uniformity parallel calculation. Next, this will be described.

  FIG. 19 shows a process 1901 for calculating the non-uniformity when the number of data is 16. In this process, first, a data string having 16 data is divided into 4 groups 1902 having 4 data, and then a non-uniformity 1904 is obtained by parallel calculation 1903 for each group. This is a process for obtaining another non-uniformity average value 1905 to obtain the non-uniformity 1906 for the original data string. In the example shown in FIG. 19, the basic data length is 4, and the data string having 4 data is the lowest layer. When all the data in a certain group is within the predetermined allowable range, these data are regarded as equal values, the non-uniformity is set to 0, and the parallel calculation of that portion is not performed. There is also a method of performing parallel calculation with a data string of 8 data as the lowest hierarchy without changing the basic data length.

  In general, when the original data is 2k power, data having the power of 2 with the number of data being (k-1) power or less can be the lowest layer. Which data column of the lowest hierarchy is to be determined is determined while viewing the table in the horizontal direction after designating the basic data length in Table 1 presented above. If the number of data in the lowest hierarchy is made larger than the basic data length, the degree of parallelism in parallel computation is lowered. The degree of parallelism is highest when the number of data in the lowest layer is made to match the basic data length, and this is desirable depending on the implementation method. Moreover, the parallel degree of parallel calculation changes also with basic data length.

Now, in order to see the process of parallel computation described above from another viewpoint, for example, consider the case where the number of data shown in FIG. Referring to FIG. 19, it can be seen that the codes for data are changed as the processing proceeds. This is shown in Table 3, which is a table for explaining the rules for inserting dummy data.

  The 16 pieces of data shown in Table 3 are considered to have a coefficient of −1 to the 0th power applied to the data in each initial state. The index of the initial coefficient changes its value according to the progress of the first operation, the second operation, the third operation and processing, and finally becomes a numerical value as shown in Table 3. This is equivalent to a process of multiplying the initial coefficient by −1 to the 0th power or the first power three times.

  FIG. 20 is a diagram for explaining the outline of the nonuniformity calculation process when the number of data is 2 to the power of k. If the above-described case of 16 data is expanded to the case where the number of data is 2 to the kth power, as shown in FIG. 20, the initial coefficient 2001 is set to −1 to the 0th power or the first power to (k− 1) Multiply (k-1) operations 2002 to obtain a final coefficient 2003 and take the sum of the calculated results to obtain an inequalities when the number of data is 2 to the kth power We will ask for 2004 once.

  From the distribution of exponents of −1 in the output shown in Table 3, for example, if the first data string is viewed, if the data for which the exponent is 0 and the value of the data for which 1 is equal, the two data are offset. It turns out that they do not contribute to the non-uniformity. From this, even when the number of given data strings is not 16, for example, 9 data, the number of data can be 16 and parallel calculation can be performed as described above. The reason for this will be described with reference to FIG.

  FIG. 21 is a diagram for explaining where dummy data should be inserted when the number of given data is 9 and the number of data is 16. In FIG. 21, mean value data mu is inserted, and dummy data (zd, yd, zd) having the same value is inserted as a pair at the position to be canceled. By doing so, the number of data can be reduced to 16, and the influence of dummy data can be prevented from appearing in non-uniformity.

  Since it is assumed that the data sequence is rearranged in ascending or descending order, the dummy data may be combined with the data adjacent to the position where the dummy data is inserted. Since they are canceled in the calculation process, the magnitude relationship between the dummy data and the adjacent data may be ignored and 0 may be substituted for all dummy data. However, the insertion position must be a fixed position as described above.

As described above, when a given data string is not 2 k power data, dummy data can be inserted into the data string so that it can be handled equivalently to 2 k data data strings. The dummy data insertion position has a certain rule as described above, and 0 may be substituted for all dummy data. Table 4 shows specific examples of data arrangement as dummy data insertion patterns. Table 4 shows an example in which the basic data length is 4 and the number of data in the lowest hierarchy is 4.

  The insertion pattern shown in Table 4 can be expanded in the case of 2 to the power of k, and this expanded one is provided as the dummy insertion rule definition file 118 in FIG. In Table 4, black circles are given data, white circles are dummy data, m is average value data, and parentheses are different patterns of the immediately preceding data string. Table 4 also shows the position where the average value m shown in FIG. 21 is inserted.

  Next, details of each processing of collection of the resource usage rate at step 301, normalization of the resource usage rate at step 302, and calculation of the non-uniformity at step 303 in the processing flow of the system resource control described with reference to FIG. Will be explained. Here, the data and data string described above are re-read as resource utilization.

  The resource utilization rate collection process in step 301 is a process in which the system resource data collection program 112 collects resource utilization rates at different time periods for each system resource 106 of the control target system 109. Then, the process proceeds to a resource utilization rate normalization process in step 302. The resource utilization rate normalization process in step 303 uses the minimum value and the maximum value for each resource utilization rate obtained by the system resource data collection program 112, and sets the minimum value to 0 and the maximum value to 1. This is a process of normalizing by linear calculation. When the stored data exceeds a predetermined capacity, the system resource data storage database 120 is deleted from the oldest one.

  FIG. 22 is a flowchart for explaining the details of the processing for calculating the non-uniformity in step 303 shown in FIG. 3, which will be described next. Note that the process of calculating the non-uniformity is executed at a constant time period.

(1) When this process is started, the resource utilization rate non-uniformity calculation program 113 is collected by the system resource data collection program 112, normalized, and stored in the system resource data storage database 120. The normalized resource usage rate is read from the system resource data storage database 120 and rearranged in ascending order or descending order (step 2201).

(2) Next, the nonuniformity calculation program 113 selects a predetermined basic data length. This determines the method of calculating the non-uniformity. The resource utilization rate is not always 2k (step 220 2 ).

(3) Therefore, whether or not dummy insertion is necessary is determined by determining whether or not the number of data is 2k. If the number of data is not 2 to the power of k in this determination, the dummy insertion rule definition file 118 as described with reference to Table 4 is read, and dummy data is inserted into the resource utilization data string according to the rules in the file. (Steps 2203 and 2204).

(4) If the number of data is 2 to the power of k in the determination at step 2203, or after the processing at step 2204, the non-uniformity is calculated by performing parallel calculation. In this calculation, the degree of parallelism is highest when the four (4-2) power groups are calculated in parallel by grouping the four resource utilization rates in the order of arrangement. On the other hand, the resource usage rate is divided into two blocks in order so that the number of data is halved, and the process proceeds separately for each block when the degree of parallelism is the lowest. There is this arbitrary degree of parallelism. This is appropriately determined by the mounting specification (step 2205).

(5) After the parallel calculation process in step 2205, the non-uniformity calculation program 113 performs an integration process for obtaining an average value of the calculation results of the blocks that have been subjected to the parallel calculation to obtain the non-uniformity (step 2206). ).

  Thus far, the description of the first embodiment of the present invention has been completed, and then the second embodiment of the present invention will be described.

  FIG. 23 is a diagram for explaining that the calculation of non-uniformity maintains linearity unless the order of arrangement with the adjacent data is changed.

In the first embodiment of the present invention described above, according to the definition of the nonuniformity when the basic data length is 2, as shown in FIG. 23, the values of the data string 2301 arranged in ascending order are displaced. In this case, the non-uniformity calculation 2303 can maintain linearity as long as the order of arrangement of the adjacent data does not change. It is also clear that there is no change in the magnitude relationship before and after the change in non-uniformity. However, if the data in the data string is displaced so that the order of arrangement with the adjacent data changes, it becomes unclear whether the non-uniformity after the change is larger or smaller than the non-uniformity before the change, It is also unclear how much the value has changed. An example of this is equation (11).

  This equation (11) is an equation showing how the non-uniformity changes before and after the change when a certain data is displaced so that the arrangement of the data string is changed when the number of data is four, When the number of data is 4, the value of the second data is decreased and made smaller than the first data. This can be understood by looking at the top row of the basic data length 2 shown in Table 1. If the basic data length is 2, the average value of the even-numbered data minus the average value of the odd-numbered data is non-uniformity. Is invariant and the linearity of non-uniformity is maintained. Similarly, looking at the case where the basic data length in Table 1 is 4 or more, the linearity of the non-uniformity can be maintained unless the data is exchanged between the first half and the subsequent half in the basic data string. I understand.

  As described above, under the constraint that data replacement does not occur even if the data value is changed, the inverse problem for obtaining the value of the data string from the predetermined non-uniformity is linear, and the solution is easy. If variance is used as the non-uniformity, the variance is essentially non-linear because it is a second-order polynomial, there is no such linear condition, and it cannot be solved as easily.

  Next, consider an example in which a value obtained by reducing the non-uniformity by a positive displacement is considered as a predetermined value, and the value of the data string corresponding thereto is calculated backward. If this displacement is large, this predetermined value may not be realized by changing the value of only one of the data strings. Therefore, in general, a method for changing a plurality of data values with respect to a predetermined displacement of non-uniformity is necessary. First, how much displacement should be given to data will be described as an example where the basic data length is 4.

  FIG. 24 is a diagram for explaining a combination of data displacements for realizing a predetermined displacement with nonuniformity when the basic data length is 4 and the number of data is 4.

  Referring to FIG. 24, the displacement may contribute positively or negatively to the non-uniformity depending on the position of the data. In the case of the example 2401 in which the non-uniformity is decreased by a predetermined displacement as shown in FIG. 24A, the sum of the displacements is less than or equal to the difference of the data for only the central two data of the given four data. Like that. For example, the two displacements may be equal to be half the difference between the data. The displacement of the data at the head and tail is the difference from the adjacent data.

  Considering the above-mentioned conditions for maintaining the linearity, the displacement of the data at the head and tail should be able to bring all four given data to the position 2402 where the displacement was given to the two central data. . However, in this case, a large displacement is allowed, and such a large displacement may not be realized in the control device. Therefore, a method is adopted in which the displacement is not increased as much as possible. This is because control is considered to be more feasible.

  As shown in FIG. 24B, in the case of Example 2403 in which the non-uniformity is increased by a predetermined displacement, the displacement is given only to the data at the beginning and the tail of the given four data. The displacement of the central two data is the difference from the adjacent data. In FIG. 24B, it can be considered that the first data of the two pieces of data in the center is equal to the head data, and the second data is equal to the tail data. However, increasing the non-uniformity leads to disturbing the control target system 109. Therefore, it is desirable to allow only a small displacement even if given, so that it cannot be easily disturbed. In this sense, the latter is avoided and the former is taken.

  FIG. 25 is a diagram for explaining a combination of data displacements for realizing a predetermined displacement with non-uniformity when the basic data length is 8 and the number of data is 8. 25A is an example 2501 in which the non-uniformity is decreased by a predetermined displacement, as in FIG. 24A, and FIG. 25B is the same as in FIG. 24B. The example 2502 increases the non-uniformity by a predetermined displacement. Even when the basic data length is 8 or more, the displacement can be given in the same manner as described above, and the displacement is made as small as possible.

  As described above, each data is subjected to a constraint that the displacement should not be such that the adjacent data can be replaced.

  FIG. 26 is a diagram for explaining a combination of data displacements for realizing a predetermined displacement of nonuniformity. The example shown in FIG. 26 is an example in which four pieces of data are provided when the basic data length is 4 and the number of data in the lowest layer is 4.

  In FIG. 26, consider a situation 2601 where the non-uniformity is reduced by a predetermined displacement. From the definition of non-uniformity, considering that the linear multiple of each data contributes to the non-uniformity, to achieve this state, the sum of the combinations of the linear displacements of each displacement of the data becomes the non-uniformity. It is only necessary that the predetermined displacement can be made approximately equal. The same applies when the non-uniformity is increased by a positive displacement.

  When three pieces of data are given, as described in the first embodiment of the present invention, an average value is calculated, inserted into the original data, and treated as if there are four pieces of data. However, at this time, the displacement with respect to the average value is determined depending on the displacement of the three data. In addition, when two pieces of data are given and two pieces of dummy data are inserted to make a total of four pieces, it can be considered that there are two pieces of data without considering the displacement of the dummy data. These points are different from the case of net 4 data, but the same as described above.

  Considering the above, the problem of finding a combination of displacements of data values equal to a predetermined displacement of non-uniformity can be regarded as a problem of packing different volume objects into a given constant volume container. . This problem is equivalent to the “0-1 knapsack problem” described in “Shuji Tsukiyama,“ Algorithm and Data Structure Design Method ”2003”. The problem with the 0-1 knapsack is that there are one item of various volumes, and the price of the item varies depending on the volume, and when the item is packed into a knapsack having a certain volume, the total price becomes the maximum. It is a combinatorial optimization problem of doing so, and it is known that there is a solution by a branch and bound method, a dynamic programming method, or the like, but its contents are not mentioned because they are known.

Now, in order to apply the 0-1 knapsack problem to the problem of obtaining a combination of displacements of a plurality of data values with respect to a predetermined displacement of non-uniformity, both the displacements of the data values can be regarded as volume and price. That's fine. Formula (12) shows this formula, and this formula (12) is a combination optimization problem when obtaining the displacement value of the input data string with respect to the predetermined displacement of non-uniformity by inverse calculation. The formulation is shown.

  In equation (12), the coefficient a may be considered as a coefficient for converting displacement into price. The coefficient c is a coefficient representing sensitivity to nonuniformity. The coefficient a and the coefficient c may be set to the same value, and any price may be proportional to the displacement at a constant rate, or may be set to different values and unevenly weighted. When uniform weighting is performed, there is a problem of obtaining a combination by preferentially collecting data having a large displacement. This is equivalent to minimizing the number of data to be displaced. Then, it can be solved by a branch and bound method, a dynamic programming method, or the like. In practice, it is not always possible to obtain a combination of displacements of data values that completely match a predetermined displacement of non-uniformity, but a combination with the smallest difference can be obtained, which is sufficient in practice. is there.

  By the way, if the predetermined nonuniformity displacement, the data displacement, the coefficient a, and the coefficient c are converted into integers, the solution of the 0-1 knapsack problem becomes simple. Therefore, in the following, all of these will be described as integers. However, in the implementation, conversion to integerization before solution is performed and conversion to real number after solution is performed, and the solution of data displacement is set to a value of 0.0 to 1.0 as in the case of data.

  Next, a system resource control method will be described by replacing the aforementioned data string or data with a resource utilization rate.

  In general, it is often impossible to control all resource utilization rates and only a part of them. For example, there may be cases where the CPU utilization of a certain server cannot be controlled due to operational limitations. On the other hand, there is a case where it is necessary to control the memory usage rate of a certain server from the viewpoint of operational efficiency. Therefore, the present invention provides a method for selecting a resource to be adopted as a control target and limiting variables to be incorporated into the combination optimization problem described above. At this time, which system resource is to be preferentially controlled is pre-defined as a control priority rule in accordance with the non-uniformity value and the set of displacement. It is assumed that there is a priority between priority rules, and the combinatorial optimization problem is solved with the highest priority, but if there is no solution, the priority is lowered sequentially. In this way, the system resources can be used based on the experience and know-how of monitoring personnel, which system resources are appropriate to be controlled, depending on how many nonuniformity values are given and how much displacement is given. It can be used for degree control. Here, since it is not forbidden to set all system resources to be controlled, a rule for setting all system resources to be controlled may be adopted.

  FIG. 27 shows the data structure of the control priority rule definition file 119 shown in FIG. The control priority rule includes a priority rule number 2701, a non-uniformity 2702, a target non-uniformity displacement 2703, and a plurality of records in which the resource ID 2704 column is one record. The smaller the number 2701, the higher the priority.

  As already explained, if a too large displacement is given to the non-uniformity, there may be no resource utilization displacement that can be achieved. Therefore, it is desirable that upper and lower limit values of a predetermined displacement that can be realized are obtained for the non-uniformity before solving the above-described combination optimization problem. Such a predetermined displacement can be obtained from a linear sum of displacements as shown in FIG. As a result, the upper and lower limit values are obtained for the displacement of the system resource controlled by the control priority rule. If this upper and lower limit value is known, as long as the value within the upper and lower limit value is used as the nonuniformity displacement, the solution can be obtained. Use it. In addition, if the predetermined displacement of the non-uniformity is outside the range of the upper and lower limit values obtained, it will be found before the solution that the control to realize such non-uniformity is impossible. It can also be used as a trigger to issue a warning indicating that the condition is not good.

  FIG. 28 is a flowchart for explaining the details of the control target value calculation processing in step 305 shown in FIG. 3, which will be described next. In this process, when dummy data is inserted, only the resource utilization rate actually measured excluding the dummy data is used. It is assumed that a predetermined displacement with respect to the nonuniformity is given. In other words, it is assumed that the determination criterion in the determination 304 of the degree of non-uniformity in FIG. 3 is given.

(1) Since the non-uniformity has already been calculated and given a predetermined displacement at the start of this process, the control target value calculation program 114 for the system resource utilization rate refers to the control priority rule definition file 119 Then, the resource ID 2704 to be controlled is specified (step 2801).

(2) Next, the upper and lower limit values of the non-uniformity displacement are calculated for the resource utilization rate of the control target. That is, as described with reference to FIG. 24, when the resource usage rates are rearranged in ascending order or descending order, the displacement is determined so that the order of adjacent data does not change. This process is a process for shifting the resource utilization rate for each resource while keeping the displacement small. That is, the strategy is to increase the number of resources to be controlled to approximate a predetermined displacement of nonuniformity. Then, a block is created for each of the four data in order from the head of the resource utilization rate sequence, and the displacement is determined in consideration of the difference with the adjacent data within a range not exceeding the block. However, when dummy data is included, the displacement with respect to them is not considered, and when the average value is included, a point determined depending on the data to which the displacement with respect to the average value is given is considered. Corresponding to the number of data in the lowest hierarchy, the coefficients for the displacement of these resource utilization rates are determined, and the sum is obtained by multiplying the determined displacement by the coefficient to obtain the upper and lower limit values for the displacement of the non-uniformity ( Step 2802).

(3) Next, determination processing for determining whether or not the predetermined displacement of the non-uniformity is within the above-described upper and lower limit range is executed, and as a result of the determination, the predetermined displacement is increased. If it is within the lower limit range, the resource utilization rate displacement is calculated. In the calculation process of the resource utilization rate displacement, the resource utilization rate displacement is calculated backward with respect to the predetermined displacement of the nonuniformity. Therefore, the combinatorial optimization problem called 0-1 knapsack problem is solved. As this preparation, the resource usage rate, the predetermined displacement of the non-uniformity, and the various coefficients described above are converted into integers. The obtained solution is converted back to a real number from 0.0 to 1.0 by performing the inverse transformation of the integerization. After this processing, the processing here ends (steps 280 3 and 2804).

(4) If it is determined in step 2803 that the predetermined displacement is not within the range of the upper and lower limit values, it is determined whether or not a rule having a lower priority than the current stage remains in the control priority rule definition file 119. If it remains, the process returns to the process of selecting the control priority rule in Step 2801, and the control object of the next priority is specified and the process is continued (Step 2805).

(5) If no rule with a lower priority than the current stage remains in the determination at step 2805, a warning flag is set to indicate that the predetermined displacement has entered an uncontrollable range. A process issued as a warning from the monitoring device 121 is executed, and the process here is terminated (step 2806).

  Next, control of the resource utilization rate in step 306 shown in FIG. 3 will be described. This control is related to the system resource utilization rate control agent 105, the system resource 106, the resource pool 107, the system resource utilization rate control program 115, the virtual resource 116, and the virtual resource pool 117 in FIG. Increasing the resource utilization rate corresponds to control for reducing the upper limit value of the system resource. Taking the memory as an example, this control means that some memory is still available, so it will return somewhat. Conversely, reducing the resource utilization rate corresponds to control for increasing the upper limit value of the resource. Speaking of memory, it is almost exhausted, meaning adding memory. Therefore, a storage device that separately prepares system resources to be added or stores them separately when they are returned is necessary. A virtual resource pool 117 is used as a storage device for such system resources. The virtual resource pool 117 is prepared for each type of system resource, and stores a predetermined amount of resources in the initial state.

  When the resource usage rate of the control target resource ID 2704 shown in the control priority rule definition is changed to change the resource allocation in the virtual resource 116, the control agent 105 of the system resource usage rate operates in the control target device 102. Then, the resource is exchanged between the corresponding system resource 106 and the resource pool 107. The virtual resource 116 and the virtual resource pool 117 conceal the real resource and the real resource pool, and control the resource usage rate even if the system resource usage rate control program 115 does not know the details of the control target system 109. be able to.

  FIG. 29 is a flowchart for explaining the details of the resource utilization rate control process in step 306 shown in FIG. 3, which will be described next.

(1) When this process is started, the system resource utilization rate control program 115 first determines whether or not there is a warning depending on whether or not a warning occurrence flag is set in the process of step 2806 of the flow shown in FIG. If the warning flag is set, the processing here is terminated without doing anything (step 2901).

(2) If the warning flag is not raised in the determination at step 2901, the resource ID 2704 shown in FIG. 27 and the resource ID 201 shown in FIG. They are collected and taken into the system resource utilization rate control program 115 for each type (step 2902).

(3) Thereafter, it is determined whether or not the displacement of the resource usage rate of the control target resource obtained in the calculation of the resource usage rate displacement in the processing of step 2804 in the flow shown in FIG. 28 is positive (step 2903).

(4) If the displacement of the resource utilization rate of the control target resource is positive in the determination at step 2903, the system is configured to realize the displacement of the given resource utilization rate in order to lower the upper limit value of the virtual resource 116. The resource utilization rate control program 115 collects surplus resources from the virtual resources 116 and performs processing for collecting virtual resources to be returned to the virtual resource pool 117. As a result of this processing, the control agent 105 of the system resource utilization rate in the control target apparatus 102 executes the control of the real resource collection that collects surplus resources from the corresponding system resource 106 according to the resource ID 201 and returns them to the resource pool 107. (Steps 2904, 2905).

(5) When the displacement of the resource utilization rate of the control target resource is negative in the determination of step 2903, the system is configured to realize the displacement of the given resource utilization rate in order to increase the upper limit value of the virtual resource 116. The resource utilization rate control program 115 performs a virtual resource addition process for adding a necessary amount of resources from the virtual resource pool 117 to the virtual resource 116. As a result of this processing, the control agent 105 of the system resource utilization rate in the control target device 102 acquires a necessary amount of resources from the resource pool 107 according to the resource ID 201 and adds them to the corresponding system resource 106. Are executed (steps 2906 and 2907).

  Each process in each embodiment of the present invention described above is configured by a program and can be executed by a CPU included in the present invention. These programs are stored in a recording medium such as an FD, a CDROM, or a DVD. It can also be provided by digital information via a network.

  The present invention quantifies the degree of variation in the measured value and controls it to a predetermined neighborhood value, which is related to object shape check, hardware and software performance inspection, manpower evaluation, etc. It can be applied not only to industrial fields related to science and engineering such as management, performance evaluation, process monitoring, resource management, management strategy, but also to industrial fields related to business administration.

It is a block diagram which shows the system configuration | structure containing the system resource control apparatus by one Embodiment of this invention. It is a figure which shows the structure of the system resource data preserve | saved at the system resource data storage database. It is a flowchart explaining the outline of the processing operation in the system resource control apparatus by embodiment of this invention. It is a figure explaining an example of the method of determining the nonuniformity in case the number of data is eight. It is a figure explaining the method to determine the nonuniformity in the example shown in FIG. 4 from another viewpoint. It is a figure explaining the method to determine the nonuniformity in the example shown in FIG. 5 from another viewpoint. It is a figure explaining the method of determining the nonuniformity in case the number of data is 8 by repeating operation demonstrated by FIG. It is a figure explaining the calculation method and definition formula of nonuniformity in case the number of data is 16 according to the method demonstrated by FIG. It is a figure which shows the calculation method and definition formula of nonuniformity in case the number of data is 8 according to the method demonstrated by FIG. It is a figure which shows the calculation method and definition formula of nonuniformity in case the number of data is four according to the method demonstrated by FIG. It is a figure which shows the calculation method and definition formula of nonuniformity in case the number of data is 2 according to the method demonstrated by FIG. It is a figure explaining the calculation method and definition formula of another nonuniformity in case the number of data is 16. It is a figure explaining the calculation method and definition formula of another nonuniformity in case the number of data is eight. It is a figure explaining the calculation method and definition formula of another nonuniformity in case the number of data is four. It is a figure explaining the calculation method and definition formula of another nonuniformity in case the number of data is two. It is a figure explaining efficiency improvement of parallel calculation in case the number of data is 16. It is a figure explaining the normalization of a nonuniformity from another viewpoint, and the calculation method and definition formula of a nonuniformity in case the number of data is four. It is a figure which shows the calculation method and definition formula of nonuniformity in case the number of data is three. It is a figure explaining the calculation method of the nonuniformity in case the number of data is 16, and the structure of the parallel calculation of nonuniformity. It is a figure explaining the outline of the calculation process of the nonuniformity in case the number of data is 2k power. It is a figure explaining where dummy data should be inserted when the number of given data is nine and the number of data is 16. It is a flowchart explaining the detail of the process of the nonuniformity calculation in step 303 shown in FIG. It is a figure explaining that calculation of nonuniformity keeps linearity, unless the order of arrangement with the next data changes. It is a figure explaining the combination of the displacement of the data for implement | achieving the predetermined displacement of nonuniformity when the basic data length is 4 and the number of data is 4. It is a figure explaining the combination of the displacement of the data for implement | achieving the predetermined displacement of nonuniformity when the basic data length is 8 and the number of data is 8. It is a figure explaining the combination of the displacement of the data for implement | achieving the predetermined displacement of nonuniformity. It is a figure which shows the data structure of the control priority rule definition file 119 shown in FIG. It is a flowchart explaining the detail of the process of calculation of the control target value in step 305 shown in FIG. 4 is a flowchart illustrating details of a resource usage rate control process in step 306 shown in FIG. 3.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 Network 102,108 Control object apparatus 103,111 Network interface 104 Resource data provision agent 105 System resource utilization control agent 106 System resource 107 Resource pool 109 Control target system 110 System resource control apparatus 112 System resource data collection program 113 System Resource utilization rate nonuniformity calculation program 114 System resource utilization rate control target value calculation program 115 System resource utilization rate control program 116 Virtual resource 117 Virtual resource pool 118 Dummy insertion rule definition file 119 Control priority rule definition file 120 System Resource data storage database 121 System resource monitoring device 122 System resource utilization rate non-uniformity display program

Claims (13)

In the system resource control device that controls the non-uniformity of the system resource usage rate of a plurality of control target devices having system resources to a predetermined value,
Means for collecting system resource utilization from controlled devices;
Inhomogeneity representing the degree of variation of the collected plurality of system resources utilization, sequentially obtains a difference from the outside of the plurality of system resource utilization for the data string arranged in ascending or descending order, sum the difference between them The total value is divided by the number of pairs, and a plurality of system resource utilization ratio values are changed, and the plurality of system resource utilization ratio values are changed within a range in which the arrangement order of the data strings in which the utilization ratios are arranged does not change. Means for calculating the non-uniformity of the system resource usage rate for obtaining the change in the non-uniformity when the value of the system resource usage rate is changed as a predetermined displacement of the non-uniformity;
When the resource usage rate is rearranged in ascending or descending order using a control priority rule definition file defined in advance for one record with non-uniformity, target non-uniformity displacement, and the resource giving the displacement In addition, the displacement value that can change the resource usage rate of the control target is obtained so that the order of adjacent data does not change, and the disparity when the system resource usage rate value is changed by this displacement value. Once the upper and lower limit values of the displacement are obtained, it is determined whether or not the predetermined displacement of the non-uniformity is within the range of the upper and lower limit values of the non-uniformity displacement. Means for calculating a resource utilization rate displacement and calculating a system resource utilization rate control target value if it falls within the range of values ,
The means for calculating the non-uniformity of the system resource utilization rate rearranges the collected system resource utilization rates in ascending order or descending order, and the total number of system resource utilization rates is a power of 2, which is smaller than the total number 2 Divide the resource utilization sequence into units of power resource utilization sequences as a unit, group each , perform parallel processing by group, and calculate the heterogeneity for system resource utilization by integrating the results A system resource control device. In the system resource control device that controls the non-uniformity of the system resource usage rate of a plurality of control target devices having system resources to a predetermined value,
Means for collecting system resource utilization from controlled devices;
Inhomogeneity representing the degree of variation of the collected plurality of system resources utilization, sequentially obtains a difference from the outside of the plurality of system resource utilization for the data string arranged in ascending or descending order, sum the difference between them The total value is divided by the number of pairs, and a plurality of system resource utilization ratio values are changed, and the plurality of system resource utilization ratio values are changed within a range in which the arrangement order of the data strings in which the utilization ratios are arranged does not change. Means for calculating the non-uniformity of the system resource usage rate for obtaining the change in the non-uniformity when the value of the system resource usage rate is changed as a predetermined displacement of the non-uniformity;
When the resource usage rate is rearranged in ascending or descending order using a control priority rule definition file defined in advance for one record with non-uniformity, target non-uniformity displacement, and the resource giving the displacement In addition, the displacement value that can change the resource usage rate of the control target is obtained so that the order of adjacent data does not change, and the disparity when the system resource usage rate value is changed by this displacement value. Once the upper and lower limit values of the displacement are obtained, it is determined whether or not the predetermined displacement of the non-uniformity is within the range of the upper and lower limit values of the non-uniformity displacement. Means for calculating a resource utilization rate displacement and calculating a system resource utilization rate control target value if it falls within the range of values ,
The means for calculating the non-uniformity of the system resource utilization rate rearranges the collected system resource utilization rates in ascending order or descending order, and the total number of system resource utilization rates is a power of 2, which is smaller than the total number 2 Divide the resource utilization sequence into units of power resource utilization sequences as a unit, group each , perform parallel processing by group, and calculate the heterogeneity for system resource utilization by integrating the results Then, a plurality of system resource utilization rates for a predetermined non-uniformity are calculated by back-calculation so that the arrangement order of the data strings in which the utilization rates are arranged does not change even if the values of the plurality of system resource utilization rates are changed. when, within a range that does not alter the order of the arrangement of the system resource utilization, when reducing the inhomogeneity predetermined displacement, the Group Displaced reduces with the system resource utilization of the top half in each flop, by a displacement in a system resource utilization of the subsequent half increased, also when increasing the inhomogeneity predetermined displacement, in each of the groups only the first is a system resource utilization of the half displacement is reduced, by a displacement in a system resource utilization of the subsequent half increased, a combination of linear sum of the displacement of the system resource utilization by combining a predetermined displacement of the non-uniformity system A system resource control apparatus characterized by calculating a nonuniformity of a resource utilization rate . The means for calculating the non-uniformity of the system resource utilization rate rearranges the collected system resource utilization rates in ascending order or descending order, and when the total number of system resource utilization rates is not a power of 2, the ascending order or descending order Dummy data is inserted at a position where the non-uniformity value calculated with respect to the arranged system resource utilization rate is not changed, so that the total number of all system resource utilization rates including the dummy data is a power of 2. by the system resource control apparatus according to claim 1, wherein the calculating the parallel processing inhomogeneity to system resource utilization.   The means for calculating the non-uniformity of the system resource utilization rate selects system resources according to a priority rule given in advance for a plurality of system resources, and calculates the non-uniformity for the system resource utilization levels. The system resource control apparatus according to claim 1, wherein The means for calculating the non-uniformity of the system resource utilization rate includes the control priority rule definition file in the process of calculating the resource utilization rates of a plurality of system resources from a predetermined displacement of the non-uniformity with respect to the resource utilization rate. 3. The system according to claim 2 , wherein one rule indicating that the number of resources to which displacement is given is selected from the minimum is selected so that the number of system resources to which displacement is to be given is minimized. Resource control unit.   The means for calculating the non-uniformity of the system resource utilization rate, when all the resource utilization rates in any of the groups are within a predetermined allowable range, regards those resource utilization rates as equal, The system resource control apparatus according to claim 1, wherein the non-uniformity with respect to the system resource utilization is calculated without executing the non-uniformity calculation for the group. In a system resource control method for controlling the non-uniformity of the system resource utilization rate of a plurality of control target devices having system resources to a predetermined value,
Means for collecting system resource utilization from controlled devices;
Inhomogeneity representing the degree of variation of the collected plurality of system resources utilization, sequentially obtains a difference from the outside of the plurality of system resource utilization for the data string arranged in ascending or descending order, sum the difference between them The total value is divided by the number of pairs, and a plurality of system resource utilization ratio values are changed, and the plurality of system resource utilization ratio values are changed within a range in which the arrangement order of the data strings in which the utilization ratios are arranged does not change. Means for calculating the non-uniformity of the system resource usage rate for obtaining the change in the non-uniformity when the value of the system resource usage rate is changed as a predetermined displacement of the non-uniformity;
When the resource usage rate is rearranged in ascending or descending order using a control priority rule definition file defined in advance for one record with non-uniformity, target non-uniformity displacement, and the resource giving the displacement In addition, the displacement value that can change the resource usage rate of the control target is obtained so that the order of adjacent data does not change, and the disparity when the system resource usage rate value is changed by this displacement value. Once the upper and lower limit values of the displacement are obtained, it is determined whether or not the predetermined displacement of the non-uniformity is within the range of the upper and lower limit values of the non-uniformity displacement. Means for calculating a resource utilization rate displacement and calculating a system resource utilization rate control target value if it falls within the range of values ,
The means for calculating the non-uniformity of the system resource utilization rate rearranges the collected system resource utilization rates in ascending order or descending order, and the total number of system resource utilization rates is a power of 2, which is smaller than the total number 2 Divide the resource utilization sequence into units of power resource utilization sequences as a unit, group each , perform parallel processing by group, and calculate the heterogeneity for system resource utilization by integrating the results A system resource control method. In a system resource control method for controlling the non-uniformity of the system resource utilization rate of a plurality of control target devices having system resources to a predetermined value,
Means for collecting system resource utilization from controlled devices;
Inhomogeneity representing the degree of variation of the collected plurality of system resources utilization, sequentially obtains a difference from the outside of the plurality of system resource utilization for the data string arranged in ascending or descending order, sum the difference between them The total value is divided by the number of pairs, and a plurality of system resource utilization ratio values are changed, and the plurality of system resource utilization ratio values are changed within a range in which the arrangement order of the data strings in which the utilization ratios are arranged does not change. Means for calculating the non-uniformity of the system resource usage rate for obtaining the change in the non-uniformity when the value of the system resource usage rate is changed as a predetermined displacement of the non-uniformity;
When the resource usage rate is rearranged in ascending or descending order using a control priority rule definition file defined in advance for one record with non-uniformity, target non-uniformity displacement, and the resource giving the displacement In addition, the displacement value that can change the resource usage rate of the control target is obtained so that the order of adjacent data does not change, and the disparity when the system resource usage rate value is changed by this displacement value. Once the upper and lower limit values of the displacement are obtained, it is determined whether or not the predetermined displacement of the non-uniformity is within the range of the upper and lower limit values of the non-uniformity displacement. Means for calculating a resource utilization rate displacement and calculating a system resource utilization rate control target value if it falls within the range of values ,
The means for calculating the non-uniformity of the system resource utilization rate rearranges the collected system resource utilization rates in ascending order or descending order, and the total number of system resource utilization rates is a power of 2, which is smaller than the total number 2 Divide the resource utilization sequence into units of power resource utilization sequences as a unit, group each , perform parallel processing by group, and calculate the heterogeneity for system resource utilization by integrating the results Then, a plurality of system resource utilization rates for a predetermined non-uniformity are calculated by back-calculation so that the arrangement order of the data strings in which the utilization rates are arranged does not change even if the values of the plurality of system resource utilization rates are changed. when, within a range that does not alter the order of the arrangement of the system resource utilization, when reducing the inhomogeneity predetermined displacement, the Group Displaced reduces with the system resource utilization of the top half in each flop, by a displacement in a system resource utilization of the subsequent half increased, also when increasing the inhomogeneity predetermined displacement, in each of the groups only the first is a system resource utilization of the half displacement is reduced, by a displacement in a system resource utilization of the subsequent half increased, a combination of linear sum of the displacement of the system resource utilization by combining a predetermined displacement of the non-uniformity system A system resource control method characterized by calculating a non-uniformity of a resource utilization rate . The means for calculating the non-uniformity of the system resource utilization rate rearranges the collected system resource utilization rates in ascending order or descending order, and when the total number of system resource utilization rates is not a power of 2, the ascending order or descending order Dummy data is inserted at a position where the non-uniformity value calculated with respect to the arranged system resource utilization rate is not changed, so that the total number of all system resource utilization rates including the dummy data is a power of 2. it allows the system resource control method according to claim 7, wherein the calculating the parallel processing inhomogeneity to system resource utilization to.   The means for calculating the non-uniformity of the system resource utilization rate selects system resources according to a priority rule given in advance for a plurality of system resources, and calculates the non-uniformity for the system resource utilization levels. 8. The system resource control method according to claim 7, wherein: The means for calculating the non-uniformity of the system resource utilization rate includes the control priority rule definition file in the process of calculating the resource utilization rates of a plurality of system resources from a predetermined displacement of the non-uniformity with respect to the resource utilization rate. 9. The system according to claim 8 , wherein one rule indicating that the number of resources to which displacement is given is selected from the minimum is selected so that the number of system resources to be given displacement is minimized. Resource control method.   The means for calculating the non-uniformity of the system resource utilization rate, when all the resource utilization rates in any of the groups are within a predetermined allowable range, regards those resource utilization rates as equal, 8. The system resource control method according to claim 7, wherein the non-uniformity with respect to the system resource utilization is calculated without executing the non-uniformity calculation for the group. In a system resource control program for controlling the nonuniformity of the system resource utilization rate of a plurality of control target devices having system resources to a predetermined value,
Collecting system resource utilization from controlled devices; and
Inhomogeneity representing the degree of variation of the collected plurality of system resources utilization, sequentially obtains a difference from the outside of the plurality of system resource utilization for the data string arranged in ascending or descending order, sum the difference between them The total value is divided by the number of pairs, and a plurality of system resource utilization ratio values are changed, and the plurality of system resource utilization ratio values are changed within a range in which the arrangement order of the data strings in which the utilization ratios are arranged does not change. Calculating the non-uniformity of the system resource usage rate, which obtains the change in the non-uniformity when the value of the system resource usage rate is changed as a predetermined displacement of the non-uniformity;
When the resource usage rate is rearranged in ascending or descending order using a control priority rule definition file defined in advance for one record with non-uniformity, target non-uniformity displacement, and the resource giving the displacement In addition, the displacement value that can change the resource usage rate of the control target is obtained so that the order of adjacent data does not change, and the disparity when the system resource usage rate value is changed by this displacement value. Once the upper and lower limit values of the displacement are obtained, it is determined whether or not the predetermined displacement of the non-uniformity is within the range of the upper and lower limit values of the non-uniformity displacement. If it is within the range of values, the CPU having the control device executes a step of calculating a resource utilization rate displacement and calculating a system resource utilization rate control target value,
The step of calculating the non-uniformity of the system resource utilization rate further includes the step of rearranging the collected system resource utilization rates in ascending order or descending order, and the total number of system resource utilization rates is a power of 2. System resource usage by dividing the resource usage rate sequence into groups, each of which is a power of 2 less than the resource usage rate, and grouping the results into groups. Calculating a non-uniformity with respect to a rate; and a plurality of predetermined non-uniformities with respect to a predetermined non-uniformity so that the arrangement order of the data strings in which the utilization rates are arranged does not change even if the values of the plurality of system resource usage rates are changed . when obtaining by inverse calculation system resource utilization, within a range that does not alter the order of the arrangement of the system resource utilization When reducing the inhomogeneity predetermined displacement, the steps leading by a displacement in a system resource utilization half reduced is increased by a displacement in a system resource utilization of the subsequent half in each of the groups, the inhomogeneity If it is increased by a predetermined displacement, displaced by reducing a certain system resource utilization of the top half in each of the groups, a step of increasing by a displacement in a system resource utilization of the subsequent half linear system resource utilization of the displacement A system resource control program comprising: calculating a non-uniformity of the system resource utilization rate by combining a predetermined displacement of the non-uniformity with a combination of sums.

Images