JP2017091001A - Virtual instance arrangement position determination device, virtual instance arrangement position determination method, and virtual instance arrangement position determination program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent a physical sever from falling into an overload state unable to respond to a computer resource request.SOLUTION: The present invention involves: predicting, on the basis of a resource use amount till a prescribed time before the prescribed time of day of a virtual instance operating in a physical server, a first resource use amount at each time of day till a prescribed time after the prescribed time of day of the virtual instance; predicting, on the basis of a resource use amount of the virtual instance at a prescribed time and a migration condition, a second resource use amount at each time of day till a prescribed time after the prescribed time of day arising due to migration of the virtual instance, a remaining time in which the migration is carried out, and a remaining stop time at which the virtual instance is stopped by the migration; and calculating, for each physical server, an index for evaluating suitability as a migration destination for the virtual instance at the prescribed time of day on the basis of the first resource use amount, the second resource use amount, the remaining time, and the remaining stop time.SELECTED DRAWING: Figure 3

Description

  The present invention relates to a virtual instance arrangement position determination device, a virtual instance arrangement position determination method, and a virtual instance arrangement position determination program.

  Conventionally, when providing virtual instances such as virtual machines and Linux (registered trademark) containers, while satisfying computer resource requirements from virtual instances, more virtual instances are placed on fewer physical servers to improve the equipment consolidation rate It is demanded to make it. As a technique for improving the device aggregation rate, a technique is known in which virtual instances are rearranged in accordance with the usage status of physical server computer resources.

  As an example of a technique for rearranging virtual instances, VMware DRS / DPM is known (for example, Non-Patent Document 1). VMware DRS / DPM is a technology that determines the placement of virtual instances so that the load on the physical server is leveled based on the computer resource usage of the physical server at a certain time, while taking into account the effects of live migration.

  In addition, a technology for predicting the physical server computer resource usage at a future time and determining the placement of virtual instances so as to suppress the load imbalance among physical servers in consideration of the predicted usage is known. (For example, patent document 1).

Japanese Patent Laying-Open No. 2015-152984

Gulati, A et al, “VMware distributed resource management design, implementation, and lessons learned.” VMware Tech. J. 1 (1), 45-64 (2012)

  However, the conventional technology for rearranging virtual instances has a problem that the physical server may fall into an overload state where it cannot respond to computer resource requests.

  For example, VMware DRS / DPM does not consider future fluctuations in computer resource usage of virtual instances. For this reason, when the computer resource usage of the virtual instance increases after the relocation, the physical server may fall into an overload state where it cannot respond to the computer resource request.

  Further, for example, the method described in Patent Document 1 does not consider the influences such as the consumption of computer resources of the physical server consumed by the live migration process and the stop time of the virtual instance accompanying the live migration process. Therefore, there is a possibility that the physical server may fall into an overload state where it cannot respond to the computer resource request due to the influence of live migration.

  The virtual instance arrangement position determination device according to the present invention provides each time from the predetermined time of the virtual instance to a predetermined time later, based on the resource usage from the predetermined time of the virtual instance operating on the physical server to a predetermined time. Based on the resource usage at the predetermined time and the status of migration of the virtual instance at each time from the predetermined time to the predetermined time later, An impact prediction unit that predicts a second resource usage amount generated by migration of a virtual instance, a remaining time during which the migration is performed, and a remaining stop time during which the virtual instance is stopped by the migration; and the usage amount prediction unit The first resource usage predicted by Based on the second resource usage, the remaining time, and the remaining stop time predicted by the impact prediction unit, an index for evaluating the suitability of the virtual instance as the migration destination at the predetermined time is used as the physical server. And a calculation unit for calculating each.

  Use of predicting the first resource usage at each time from the predetermined time to the predetermined time after the virtual instance based on the resource usage from the predetermined time to the predetermined time before the virtual instance operating on the physical server A second resource usage caused by the migration of the virtual instance at each time from the predetermined time to the predetermined time based on the amount prediction step, the resource usage of the virtual instance at the predetermined time, and the status of migration An impact prediction step of predicting an amount, a remaining time during which the migration is performed, and a remaining stop time during which the virtual instance is stopped by the migration, and the first resource usage amount predicted by the usage amount prediction step And the predicted by the impact prediction step A calculation step of calculating, for each physical server, an index for evaluating the suitability of the virtual instance as the migration destination at the predetermined time based on the resource usage amount of 2, the remaining time, and the remaining stop time. It is characterized by that.

  ADVANTAGE OF THE INVENTION According to this invention, it can prevent that a physical server falls into the overload state which cannot respond to a computer resource request | requirement.

FIG. 1 is a diagram illustrating an example of a configuration of a virtual instance arrangement system according to the first embodiment. FIG. 2 is a block diagram illustrating a configuration of the physical server control device according to the first embodiment. FIG. 3 is a block diagram illustrating a configuration of the virtual instance arrangement position determining apparatus according to the first embodiment. FIG. 4 is a diagram illustrating an example of data of the physical server basic information DB. FIG. 5 is a diagram illustrating an example of data in the virtual instance resource usage DB. FIG. 6 is a diagram illustrating an example of data in the virtual instance resource usage DB. FIG. 7 is a diagram illustrating an example of data in the virtual instance arrangement position DB. FIG. 8 is a diagram illustrating an example of data in the live migration state DB. FIG. 9 is a diagram illustrating an example of data in the usage amount prediction DB. FIG. 10 is a diagram illustrating an example of data in the usage amount prediction DB. FIG. 11 is a diagram illustrating an example of data in the impact prediction DB. FIG. 12 is a diagram illustrating an example of data in the index DB. FIG. 13 is a flowchart illustrating an example of processing of the virtual instance arrangement position determination device according to the first embodiment. FIG. 14 is a flowchart illustrating an example of processing of the virtual instance arrangement position determination device according to the first embodiment. FIG. 15 is a diagram illustrating an example of a computer in which a virtual instance arrangement position determining device is realized by executing a program.

  Hereinafter, embodiments of a virtual instance arrangement position determining apparatus, a virtual instance arrangement position determining method, and a virtual instance arrangement position determining program according to the present application will be described in detail with reference to the drawings. Note that the virtual instance arrangement position determining device, the virtual instance arrangement position determining method, and the virtual instance arrangement position determining program according to the present application are not limited to the embodiment.

[First Embodiment]
First, the configuration, process flow, and effects of the virtual instance arrangement system according to the first embodiment will be described.

[Configuration of First Embodiment]
(Configuration of virtual instance placement system)
First, the configuration of the virtual instance arrangement system will be described with reference to FIG. FIG. 1 is a diagram illustrating an example of a configuration of a virtual instance arrangement system according to the first embodiment. As shown in FIG. 1, virtual instances are arranged on the physical server. At this time, the virtual instance operates using the computer resources of the physical server.

  In addition, the number of virtual instances arranged on one physical server may be 0, or 1 or more. Furthermore, live migration of virtual servers may be performed between physical servers. In live migration, the operation of a virtual instance may stop for a predetermined time.

  As shown in FIG. 1, the virtual instance placement system 1 is connected to each physical server. The virtual instance placement system 1 includes a physical server control device 10, a virtual instance placement location determination device 20, and a time clock 30.

  The physical server control device 10 collects information related to the physical server and the virtual instance and provides the information to the virtual instance arrangement position determination device 20. The virtual instance arrangement position determination device 20 determines the arrangement position of the virtual instance. Further, the physical server control device 10 arranges the virtual instance on the physical server at the position determined by the virtual instance arrangement position determination device 20. The time clock 30 makes a processing request to the physical server control device 10 and the virtual instance arrangement position determination device 20 at a predetermined timing.

(Configuration of physical server controller)
Next, the configuration of the physical server control device 10 will be described with reference to FIG. FIG. 2 is a block diagram illustrating a configuration of the physical server control device according to the first embodiment. As illustrated in FIG. 2, the physical server control device 10 includes a control unit 11 and a storage unit 12.

  The control unit 11 includes a virtual instance resource usage collection unit 111, a virtual instance arrangement position collection unit 112, and a virtual instance arrangement change unit 113. The storage unit 12 includes a physical server basic information DB 121, a virtual instance resource usage DB 122, a virtual instance arrangement position DB 123, and a live migration state DB 124.

  The virtual instance resource usage collection unit 111 collects the resource usage of each virtual instance each time a processing request is received from the time clock 30, and stores the collected resource usage in the virtual instance usage DB 122. For example, when the time clock 30 makes a processing request every second, the virtual instance resource usage collection unit 111 updates the virtual instance usage DB 122 every second.

  The virtual instance arrangement position collection unit 112 collects the arrangement position of each virtual instance each time a processing request is received from the time clock 30 and stores the collected arrangement position in the virtual instance arrangement position DB 123. For example, when the time clock 30 makes a processing request every second, the virtual instance arrangement position collection unit 112 updates the virtual instance arrangement position DB 123 every second.

  When the virtual instance migration destination is determined, the virtual instance arrangement changing unit 113 arranges the virtual instance on the physical server of the arrangement destination. For example, the virtual instance arrangement changing unit 113 refers to information stored in the live migration state DB 124 at regular time intervals to acquire an arrangement destination physical server.

  The physical server basic information DB 121 stores the CPU specifications and memory capacity of each physical server. Here, the data of the physical server basic information DB 121 will be described with reference to FIG. FIG. 4 is a diagram illustrating an example of data of the physical server basic information DB. As shown in FIG. 4, the physical server basic information DB 121 stores a CPU specification (unit: GHz) and a memory capacity (unit MB) for each server ID of the physical server.

  The physical server ID is an ID for identifying the physical server. Hereinafter, a physical server whose physical server ID is i is represented as a physical server k. Further, the CPU spec indicates the maximum value of the usable CPU spec. The memory capacity indicates the maximum value of usable memory capacity. For example, the physical server 1 indicates that a CPU specification of maximum 3.0 GHz and a memory capacity of maximum 2000 MB can be used. It is assumed that the data in the physical server basic information DB 121 is input manually or automatically in advance.

  The virtual instance resource usage DB 122 stores the resource usage at each time of each virtual instance. Here, the data in the virtual instance resource usage DB 122 will be described with reference to FIGS. 5 and 6. 5 and 6 are diagrams illustrating an example of data in the virtual instance resource usage DB.

  As shown in FIG. 5, the virtual instance resource usage DB 122 stores the usage amount of the CPU spec at each time of each virtual instance. The virtual instance ID is an ID for identifying a virtual instance. Hereinafter, a virtual instance whose virtual instance is i is represented as a virtual instance i. For example, the usage amount of the CPU spec at the time t of the virtual instance 1 is 0.33 GHz.

  As illustrated in FIG. 6, the virtual instance resource usage DB 122 stores the memory capacity of each virtual instance at each time. For example, the usage amount of the memory capacity of virtual instance 1 at time t is 30 MB.

  The virtual instance arrangement position DB 123 stores a physical server in which each virtual instance is arranged. Here, the data in the virtual instance arrangement position DB 123 will be described with reference to FIG. FIG. 7 is a diagram illustrating an example of data in the virtual instance arrangement position DB. For example, the virtual instance 1 is arranged on the physical server 3.

  The live migration status DB 124 stores the live migration status of each virtual instance. Here, the data of the live migration state DB 124 will be described with reference to FIG. FIG. 8 is a diagram illustrating an example of data in the live migration state DB. The migration status indicates whether migration is in progress. The migration source physical server ID indicates the ID of the physical server currently arranged. The migration destination physical server ID indicates the ID of the physical server that is the placement destination.

  For example, the virtual instance 1 is not in live migration because the migration state is 0. In addition, since the migration state of the virtual instance 2 is 1, live migration is being performed, and the arrangement of the physical instance 2 is changed from the physical server 2 to the physical server 1.

(Configuration of virtual instance placement position determination device)
Next, the configuration of the virtual instance arrangement position determination device 20 will be described with reference to FIG. FIG. 3 is a block diagram illustrating a configuration of the virtual instance arrangement position determining apparatus according to the first embodiment. As illustrated in FIG. 3, the virtual instance arrangement position determination device 20 includes a control unit 21 and a storage unit 22.

  The control unit 21 includes a determination unit 211, a usage amount prediction unit 212, an influence prediction unit 213, and a calculation unit 214. The storage unit 22 includes a usage amount prediction DB 221, an impact prediction DB 222, and an index DB 223.

  Each time the determination unit 211 receives a processing request from the time clock 30, the determination unit 211 causes each unit of the control unit 21 to execute processing, and determines the physical server with the maximum index calculated by the calculation unit 214 as the migration destination of the virtual instance. To do. In addition, the determination unit 211 updates the migration state of the live migration state DB 124 of the virtual instance whose placement destination has been determined to 1, and sets the migration source physical server ID to the physical server ID of the physical server in which the virtual instance is placed. The migration destination physical server ID is updated and updated to the physical server ID of the placement destination physical server.

  The usage amount predicting unit 212 is predetermined from a predetermined time of the virtual instance based on a resource usage amount from a predetermined time (for example, time t) of the virtual instance operating on the physical server to a predetermined time before (for example, time t-A). The first resource usage amount at each time after time (for example, time t + B) is predicted. Further, the usage amount prediction unit 212 stores the predicted resource usage amount in the usage amount prediction DB 221.

  The usage amount prediction DB 221 stores the predicted value of the resource usage amount at each time of each virtual instance. Here, data of the usage amount prediction DB 221 will be described with reference to FIGS. 9 and 10. 9 and 10 are diagrams illustrating an example of data in the usage amount prediction DB.

  As illustrated in FIG. 9, the usage amount prediction DB 221 stores a predicted value of the usage amount of the CPU spec at each time of each virtual instance. For example, the predicted value of the usage amount of the CPU spec at time t + B of the virtual instance whose virtual instance ID is 1 is 0.33 GHz.

  As illustrated in FIG. 10, the usage amount prediction DB 221 stores the memory capacity of each virtual instance at each time. For example, the predicted value of the memory capacity usage of virtual instance 1 at time t + B is 30 MB.

  A procedure for the usage amount prediction unit 212 to predict the resource usage amount will be described. First, how to represent each data used in prediction will be described.

A set of virtual instances is expressed as shown in Equation (1). For example, the i-th virtual instance is represented as CO _i or the like.

A set of physical servers is expressed as shown in Equation (2). For example, the k-th physical server is expressed as S _k or the like.

The maximum value of the CPU specifications of each physical server is represented by a 1 × _ns vector shown in Expression (3). Expression (3) represents the CPU specifications of the physical server basic information DB 121 of FIG. 4 as a vector.

_{Incidentally,} MX c _{(S k)} represents the maximum value of the CPU specification of the physical server _{S k.} Further, if all of the physical servers with the largest value of the same CPU specification, the maximum value of the CPU specification of the physical server S _k may be simply expressed as MX _c.

The maximum value of the memory capacity of each physical server is represented by a 1 × _ns vector shown in Equation (4). Expression (4) represents the memory capacity of the physical server basic information DB 121 of FIG. 4 as a vector.

MX _m (S _k ) represents the maximum value of the memory capacity of the physical server S _k . Further, if all of the physical servers with the largest value of the same memory capacity, the maximum value of the memory capacity of the physical server S _k may be simply expressed as MX _m.

The CPU spec consumption of each virtual instance at time t is represented by a 1 × n _co vector shown in Equation (5). Expression (5) represents the CPU spec usage at time t in the virtual instance resource usage DB 122 of FIG. 5 as a vector.

Note that c _i (t) represents the CPU spec consumption of the virtual instance co _i at time t.

The consumption of the memory capacity of each virtual instance at time t is represented by a 1 × n _co vector shown in Expression (6). Expression (6) represents the amount of memory capacity used at time t in the virtual instance resource usage DB 122 of FIG. 6 as a vector.

Note that m _i (t) represents the memory capacity consumption of the virtual instance co _i at time t.

A physical server in which each virtual instance at time t is arranged is represented by a 1 × n _co vector shown in Expression (7). Expression (7) represents the ID of the physical server arranged at the time t in the virtual instance arrangement position DB 123 of FIG. 7 as a vector.

However, p _i (t) is a function that returns the physical server ID of the physical server in which the virtual instance whose virtual instance ID is i at time t is arranged, as shown in Expression (8).

In the prediction, a 1 × n _co vector that takes 1 only for the component of the virtual instance ID of the virtual instance placed in the physical server k at time t as shown in Expression (9) is also used.

In the prediction, a 1 × n _co vector in which all components are 1 as shown in Expression (10) is also used.

  Next, a specific calculation method when the usage amount prediction unit 212 predicts the resource usage amount will be described. The usage amount prediction unit 212 assumes a stochastic process model for the resource usage amount, estimates model parameters using data of the resource usage amount up to time A before the current time t, and determines the current time t The resource usage from B to the time after B is predicted.

  The usage amount prediction unit 212 can use, for example, a random walk, a Poisson process, an AR (Auto Regressive) model, an MA (Moving Average) model, an ARMA model, an ARIMA model, or the like as a stochastic process model. Further, the usage amount prediction unit 212 can use, for example, the Yule Walker method for the AR model and the maximum likelihood method for the Poisson process as the parameter estimation method.

  In the present embodiment, the usage amount prediction unit 212 uses the AR model as the stochastic process model and uses the Yule Walker method as the parameter estimation method. Note that the usage amount prediction unit 212 formulates that the average value of the resource usage so far continues indefinitely after time t + B.

  First, the usage amount prediction unit 212 acquires CPU spec usage data up to time t from the virtual instance resource usage DB 122, and based on the acquired CPU spec usage amount, the virtual instance i is expressed by Expression (11). CPU usage at time t + 1 is predicted.

  Then, the usage amount prediction unit 212 places the sample average as shown in Equation (12) and places the sample autocorrelation function as shown in Equation (13).

Further, the usage amount prediction unit 212 obtains the maximum likelihood estimated value of the parameters a ₁ , a ₂ ,..., A _M by solving the Yule Walker equation shown in Expression (14).

In addition, the usage amount predicting unit 212 obtains an estimated value of σ _i ² using Equation (15).

Here, when the obtained estimated value of σ _i ² is expressed as an estimated value of variance when the order is M, the AIC of the order M is expressed as Expression (16).

The usage amount prediction unit 212 determines the prediction model by selecting the order M ^* that minimizes the AIC _M in Expression (16). Then, the usage amount prediction unit 212 predicts the usage amount of the CPU spec of the virtual instance i at the time t + τ by Expression (17) using the determined prediction model.

  In the above description, the usage amount of the CPU spec is used as an example of the resource usage amount. However, the usage amount prediction unit 212 can predict the usage amount of the memory capacity in the same manner, for example.

  Based on the resource usage and migration status of the virtual instance at a predetermined time, the impact prediction unit 213 determines the resource usage generated by the migration of the virtual instance at each time from the predetermined time to a predetermined time and the migration. And the remaining stop time during which the virtual instance stops due to the migration. Further, the impact prediction unit 213 stores the impact due to the predicted migration in the impact prediction DB 222.

  The impact prediction DB 222 stores observed values and predicted values of the impact of migration at each time of each virtual instance. Here, the data of the impact prediction DB 222 will be described with reference to FIG. FIG. 11 is a diagram illustrating an example of data in the impact prediction DB.

  As shown in FIG. 11, the impact prediction DB 222 includes the remaining migration time at each time of each virtual instance, the stop time due to migration, the average CPU spec usage of the transmission source server, the average CPU spec usage of the destination server, and the transmission source server. The average memory capacity use amount, the average memory capacity use amount of the destination server, and the observed value and predicted value of the stop flag are stored.

  For example, the virtual instance 1 is in a state immediately before the start of migration, the remaining migration time is 50, the stop time due to migration, the average CPU spec usage of the transmission source server, the average CPU spec usage of the destination server, The average memory capacity usage of the destination server, the average memory capacity usage of the destination server, and the stop flag are 0.

  Further, for example, the virtual instance 2 is in the state of being migrated and the virtual instance is not stopped, the remaining migration time is 100, the migration stop time is 50, the average CPU spec usage of the transmission source server is 0.01, the destination The average CPU spec usage of the server is 0.02, the average memory capacity usage of the transmission source server is 30, the average memory usage of the destination server is 50, and the stop flag is 0.

  Further, for example, the virtual instance 3 is in a state of being migrated and the virtual instance is stopped, the migration remaining time is 40, the migration stop time is 40, the average CPU spec usage of the transmission source server is 0.02, The destination server's average CPU spec usage is 0.01, the source server's average memory capacity usage is 20, the destination server's average memory capacity usage is 20, and the stop flag is 1.

  When time t is the current time, the impact prediction DB 222 stores an observed value for time t, and stores a predicted value predicted by a prediction method described later for time t + 1 and thereafter. Further, when the prediction is performed up to time t + B, the impact prediction DB 222 stores a table as shown in FIG. 11 for each time, and therefore stores B + 1 tables. When migration is not performed, each value in the impact prediction DB 222 is 0. Further, when migration is being performed and the virtual instance has stopped operating, the stop flag becomes 1.

  A procedure for the influence prediction unit 213 to predict the influence of migration will be described. First, how to represent each data used in prediction will be described.

The migration state of each virtual instance at time t is represented by a 1 × n _co vector shown in Expression (18). Expression (18) represents the migration state of the live migration state DB 124 of FIG. 8 as a vector.

The remaining migration time of each virtual instance at time t is represented by a 1 × n _co vector shown in Expression (19). Expression (19) represents the remaining migration time in the impact prediction DB 222 of FIG. 11 as a vector.

Here, for example, when the time changes from t to t + 1, the remaining migration time is subtracted by 1. Therefore, when mgt _i (t)> 0, mgt _i (t + 1) = mgt _i (t) −1 expressed.

The migration source physical server of each virtual instance at time t is represented by a 1 × n _co vector shown in Expression (20). Expression (20) represents the migration source physical server ID of the live migration state DB 124 of FIG. 8 as a vector.

The migration destination physical server of each virtual instance at time t is represented by a 1 × n _co vector shown in Expression (21). Expression (21) represents the migration destination physical server ID of the live migration state DB 124 of FIG. 8 as a vector.

The stop time due to migration of each virtual instance at time t is represented by a 1 × n _co vector shown in Expression (22). Expression (22) represents the stop time due to migration of the influence prediction DB 222 of FIG. 11 as a vector.

Here, for example, when the time changes from t to t + 1, if ft _i (t) ≦ mgt _i (t), the stop time due to migration is subtracted by 1, so ft _i (t + 1) = ft _i It is expressed as (t) -1.

The average CPU spec usage of the transmission source server of each virtual instance at time t is represented by a 1 × n _co vector shown in Expression (23). Expression (23) represents the average CPU spec usage of the transmission source server in the influence prediction DB 222 of FIG. 11 as a vector.

The average CPU spec usage of the destination server of each virtual instance at time t is represented by a 1 × n _co vector shown in Expression (24). Expression (24) represents the average CPU spec usage of the destination server in the impact prediction DB 222 of FIG. 11 as a vector.

The average memory capacity usage of the transmission source server of each virtual instance at time t is represented by a 1 × n _co vector shown in Expression (25). Expression (25) represents the average memory capacity usage of the transmission source server in the influence prediction DB 222 of FIG. 11 as a vector.

The average memory capacity usage of the destination server of each virtual instance at time t is represented by a 1 × n _co vector shown in Equation (26). Expression (26) represents the average memory capacity usage of the destination server in the impact prediction DB 222 of FIG. 11 as a vector.

The stop flag of each virtual instance at time t is represented by a 1 × n _co vector shown in Expression (27). Expression (27) represents the stop flag of the influence prediction DB 222 of FIG. 11 as a vector.

Next, each vector ^MgT described so far ^{(t), FT (t)} , MgC s (t), MgC d (t), MgM s (t), MgM d (t), effects the prediction of FF (t) A calculation method by the unit 213 will be described. Note that each of the prediction coefficients α _mgt , α _ft , α ^s _mgc , α ^d _mgc , α ^s _mgm , and α ^d _mgm is a positive real number and is a constant obtained in advance by a measurement experiment.

Since the remaining virtual instance migration time is proportional to the memory capacity usage m _i (t), the impact prediction unit 213 obtains each element of the vector MgT (t) of Equation (19) by Equation (28).

Since the stop time due to the migration of the virtual instance is proportional to the memory capacity use amount m _i (t), the impact prediction unit 213 obtains each element of the vector FT (t) of Expression (22) by Expression (29).

Since the usage amount of the CPU specification of the transmission source physical server due to the migration of the virtual instance is hardly affected by the usage amount m _i (t) of the memory capacity, the impact prediction unit 213 uses the vector MgC ^s ( Each element of t) is obtained by equation (30).

Since the usage amount of the CPU specification of the destination physical server due to the migration of the virtual instance is gently proportional to the usage amount m _i (t) of the memory capacity, the influence prediction unit 213 uses the vector MgC ^d (t) in Expression (24). Each element of is calculated | required by Formula (31).

Also, assuming that the usage amount of the CPU specification of the destination physical server due to the migration of the virtual instance is proportional to the usage amount m _i (t) of the memory capacity, the impact prediction unit 213 determines the vector MgC ^d (t) in Expression (24). ) May be obtained by equation (32).

Since the memory capacity usage of the transmission source physical server due to the virtual instance migration is hardly affected by the memory capacity usage m _i (t), the impact prediction unit 213 uses the vector MgM ^s ( Each element of t) is obtained by Expression (33).

Since the usage amount of the memory capacity of the destination physical server due to the migration of the virtual instance is proportional to the usage amount m _i (t) of the memory capacity, the impact prediction unit 213 calculates each of the vectors MgM ^d (t) in Expression (26). An element is calculated | required by Formula (34).

  The calculation unit 214 migrates the virtual instance at a predetermined time based on the resource usage predicted by the usage prediction unit 212 and the resource usage, remaining time, and remaining stop time predicted by the impact prediction unit 213. An index for evaluating suitability as a destination is calculated for each physical server.

  The calculation unit 214 requests a resource usage amount predicted by the usage amount prediction unit 212 and an excess penalty that increases as the resource usage amount predicted by the influence prediction unit 213 decreases and a virtual instance stop based on the remaining stop time. The usage request impossibility penalty that increases as the resource usage that cannot be answered is calculated, and an index is calculated based on the sum of the excess penalty and the use request impossibility penalty.

  Further, the calculation unit 214 calculates a used physical server number penalty that increases as the number of physical servers on which virtual instances are arranged decreases, and a value obtained by multiplying the sum of the excess penalty and the use request impossible penalty by the first coefficient The index may be calculated based on the sum of the third value multiplied by the second coefficient. Note that the sum of the first coefficient and the second coefficient is a constant (for example, 1).

Here, the calculation unit 214 first defines a constant N _c for normalizing the usage amount of the CPU specification and a constant N _m for normalizing the usage amount of the memory capacity. For example, the calculation unit 214 can set N _c as the average value E [MX _c ] of the CPU specifications of the physical server. In addition, the calculation unit 214 can set N _m to, for example, the average value E [MX _m ] of the memory capacity of the physical server.

The calculation unit 214 calculates the usage amount of the CPU specifications of all the physical servers due to the influence of migration, using Expression (35). Note that the usage amount of the CPU specifications of the physical server k due to the influence of migration is expressed as MgCA _k (t).

The calculation unit 214 calculates the usage amount of the memory capacity of all the physical servers due to the influence of migration, using Expression (36). Note that the usage amount of the memory capacity of the physical server k due to the influence of migration is expressed as MgMA _k (t).

  The calculation unit 214 calculates the total of all physical servers with CPU specifications that cannot respond to the request due to the suspension of the virtual instance due to migration, using Expression (37).

  The calculation unit 214 calculates the sum of all physical servers having a memory capacity that cannot respond to the request due to the suspension of the virtual instance due to the migration, using Expression (38).

  The calculation unit 214 represents the usage amount of the CPU specifications of each physical server at time t by Expression (39).

  In addition, the calculation unit 214 represents the usage amount of the CPU specifications of the physical server k at time t by Expression (40).

  The calculation unit 214 represents the usage amount of the memory capacity of each physical server at the time t by Expression (41).

  Further, the calculation unit 214 represents the usage amount of the memory capacity of the physical server k at the time t by Expression (42).

  The calculation unit 214 calculates the total sum of excess penalties for CPU specifications of all physical servers at time t using equation (43). The excess penalty increases as the resource amount of the physical server increases.

  The calculation unit 214 calculates the sum of excess penalties for memory capacity of all the physical servers at time t using equation (44).

  The calculation unit 214 calculates the sum of the use request impossibility penalty of the CPU specifications of all physical servers at time t by the equation (45). The use request impossibility penalty increases as the amount of resources used by the stopped virtual instance decreases.

  The calculation unit 214 calculates the sum of the use request impossible penalties for the CPU specifications of all the physical servers at time t according to the equation (46).

  The calculation unit 214 calculates the number of physical servers on which virtual instances at time t are arranged according to equation (47).

  Then, the calculation unit 214 calculates the number of physical server use penalties at time t by using the equation (48-1) or the equation (48-2). The number of used physical servers penalty increases as the number of physical servers on which virtual instances are arranged decreases.

  Then, the calculation unit 214 calculates the total penalty at the time t according to the equation (49).

  Here, the excess penalty, the unusable penalty, and the used physical server number penalty become larger as migration is appropriate. For example, when the migration destination is a physical server on which no virtual instance is arranged, the excess penalty and the unusable penalty are often increased. On the other hand, in this case, the penalty for the number of used physical servers is small.

Thus, the excess penalty and disabling penalty between the use physical number of servers penalty, since the trade-off relationship is established, the calculation unit 214 has set a condition that _w s ₊ w h = _1. As a result, virtual instances can be arranged on as few physical servers as possible.

The calculation unit 214 calculates an arrangement position index Q _i (P (t), mgd _i (t) = k) when the arrangement of the virtual instance i is changed to the physical server k, using Expression (50).

  Here, as an index calculation method, for example, a method of calculating as an index a total sum of penalties from the current time to an infinite future time, which is exponentially discounted with time 0 <γ ≦ 1, can be considered. Here, the calculation unit 214 calculates, as an index, the total penalty up to a finite future time in order to shorten the calculation time.

For example, as indicated by τ _p in the equation (50), the calculation unit 214 calculates the future time by the remaining time of migration required when it is assumed that the virtual instance using the memory capacity of the physical server having the largest memory capacity has migrated. Is a finite future time. Then, the calculation unit 214 stores the calculated index in the index DB 223. As shown in FIG. 12, the index DB 223 stores an index when each physical server is a migration destination. FIG. 12 is a diagram illustrating an example of data in the index DB.

The determination unit 211 refers to the index DB 223 and determines the physical server k ′ having the largest allocation index as the allocation position. At this time, in the case of k ′ ≠ mgd _i (t), the migration is determined. Therefore, the determination unit 211 uses the expressions (51-1), (51-2), and (51-3). The migration status, migration source physical server ID, and migration destination physical server ID in the live migration status DB 124 are updated.

[Process of First Embodiment]
The process of the virtual instance arrangement position determination device 20 will be described with reference to FIGS. 13 and 14. FIG. 13 and FIG. 14 are flowcharts illustrating an example of processing of the virtual instance arrangement position determination device according to the first embodiment.

  First, as illustrated in FIG. 13, the determination unit 211 acquires the current time t from the time clock 30 (step S21). Then, the determination unit 211 extracts the resource usage from time t-A to time t of all virtual instances from the virtual instance resource usage DB 122 (step S22). Next, the determination unit 211 extracts the arrangement positions at the time t of all virtual instances from the virtual instance arrangement position DB 123 (step S23).

  Then, the determination unit 211 transmits the current time t, the data extracted from each DB, and the predicted time t + B to the usage amount prediction unit 212, and requests processing (step S24). When receiving the processing request, the usage amount prediction unit 212 predicts the resource usage amount from time t to time t + B for all virtual instances (step S25). Here, the calculation unit 214 sets i to 1 (step S26).

  The calculation unit 214 refers to the live migration state DB 124 and determines whether or not the virtual instance i is in the migration state (step S27). When the virtual instance i is in the migration state (step S27, Yes), the calculation unit 214 proceeds to step S30. When the virtual instance i is not in the migration state (step S27, No), the calculation unit 214 calculates an index for the virtual instance i (step S28).

  The process in which the calculation unit 214 calculates the index will be described with reference to FIG. First, as illustrated in FIG. 14, the calculation unit 214 sets k to 1 (step S281). Here, k indicates the ID of a physical server that is a candidate for the migration destination of the virtual instance i. When the placement position of the virtual instance i is k (Yes in step S282), migration is not performed, and the calculation unit 214 proceeds to step S286 without calculating the index.

  When the placement position of the virtual instance i is not k (step S282, No), the calculation unit 214 transmits the resource usage amount and the placement position of the virtual instance i at the current time t to the influence prediction unit 213 and requests processing. (Step S283). When receiving the processing request, the impact prediction unit 213 assumes that the migration of the virtual instance i to the physical server k is performed, and predicts the impact of the migration from time t to time t + B on the physical server k (step S284). Then, the calculation unit 214 calculates an index based on the resource usage amount predicted by the usage amount prediction unit 212 and the migration influence predicted by the influence prediction unit 213 (step S285).

  Here, if k matches the number of all physical servers (Yes in step S286), the index is calculated for all physical servers, and thus the calculation unit 214 ends the process. If k does not match the number of physical servers (No at Step S286), the calculation unit 214 sets k to k + 1, returns to Step S282, and repeats the processing.

  When the calculation of the index ends, the determination unit 211 determines the placement position of i of the virtual instance based on the index (step S29). At this time, the determination unit 211 updates the live migration state DB 124 based on the determined arrangement position.

  Here, when i matches the number of all virtual instances (step S30, Yes), since the placement positions have been examined for all virtual instances, the determination unit 211 ends the process. If i does not match the number of all virtual instances (No in step S30), the determination unit 211 sets i to i + 1 (step S31), returns to step S27, and repeats the process.

[Effect of the first embodiment]
The usage amount predicting unit 212 uses the first resource usage at each time from the predetermined time of the virtual instance to the predetermined time later, based on the resource usage from the predetermined time of the virtual instance operating on the physical server to the predetermined time before. Predict the amount. In addition, the impact prediction unit 213 uses the second resource usage amount generated by the migration of the virtual instance at each time from the predetermined time to a predetermined time later, based on the resource usage amount and the migration status of the virtual instance at the predetermined time. The remaining time during which the migration is performed and the remaining stop time during which the virtual instance is stopped by the migration are predicted. Then, the calculation unit 214 is based on the first resource usage amount predicted by the usage amount prediction unit 212, the second resource usage amount predicted by the influence prediction unit 213, the remaining time, and the remaining stop time. An index for evaluating the suitability of the virtual instance as a migration destination at a predetermined time is calculated for each physical server.

  This makes it possible to relocate virtual instances taking into account both the increase in resource usage after relocation of virtual instances and the increase in load due to the effect of live migration, and the physical server responds to computer resource requests. It is possible to prevent an overload condition that is not possible.

  In addition, the determination unit 211 determines the physical server having the maximum index calculated by the calculation unit 214 as the migration destination of the virtual instance.

  As described above, by determining the relocation destination of the virtual instance according to a predetermined rule, it is possible to perform stable relocation of the virtual instance.

  In addition, the calculation unit 214 uses the first value that increases as the first resource usage amount and the second resource usage amount decrease, and the resource usage that cannot be answered by the virtual instance stop based on the remaining stop time. The second value that increases as the amount decreases may be calculated, and the index may be calculated based on the sum of the first value and the second value.

  Thus, by calculating the index based on both the plurality of resource usage amounts and the required time for migration, it is possible to evaluate the placement destination more accurately.

  Further, the calculation unit 214 calculates a third value that increases as the number of physical servers in which virtual instances are arranged decreases, and multiplies the sum of the first value and the second value by the first coefficient. The index may be calculated based on the sum of the value and the value obtained by multiplying the third value by the second coefficient. Note that the sum of the first coefficient and the second coefficient is a constant.

  As a result, virtual instances can be arranged on as few physical servers as possible.

[Other Embodiments]
In the first embodiment, CPU specifications and memory capacity are given as examples of physical server resources used by virtual instances. However, the resources of the physical server are not limited to these. For example, it is conceivable that the network bandwidth used by the virtual instance is predicted and reflected in the index.

[System configuration, etc.]
Further, each component of each illustrated apparatus is functionally conceptual, and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution / integration of each device is not limited to that shown in the figure, and all or a part thereof may be functionally or physically distributed or arbitrarily distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured. Further, all or any part of each processing function performed in each device is realized by a CPU (Central Processing Unit) and a program analyzed and executed by the CPU, or hardware by wired logic. Can be realized as

  In addition, among the processes described in the present embodiment, all or part of the processes described as being automatically performed can be manually performed, or the processes described as being manually performed can be performed. All or a part can be automatically performed by a known method. In addition, the processing procedure, control procedure, specific name, and information including various data and parameters shown in the above-described document and drawings can be arbitrarily changed unless otherwise specified.

[program]
As one embodiment, the virtual instance arrangement position determination device can be implemented by installing a virtual instance arrangement position determination program that executes the above-described virtual instance arrangement position determination as package software or online software on a desired computer. For example, by causing the information processing apparatus to execute the virtual instance arrangement position determination program, the information processing apparatus can function as a virtual instance arrangement position determination apparatus. The information processing apparatus referred to here includes a desktop or notebook personal computer. In addition, the information processing apparatus includes mobile communication terminals such as smartphones, mobile phones and PHS (Personal Handyphone System), and slate terminals such as PDA (Personal Digital Assistants).

  In addition, the virtual instance arrangement position determination device can be implemented as a server device that uses the terminal device used by the user as a client and provides the client with the service related to the virtual instance arrangement position determination. For example, the virtual instance arrangement position determination apparatus is implemented as a server apparatus that provides a virtual instance arrangement position determination service that receives the resource usage status of a virtual instance and outputs the virtual instance arrangement position. In this case, the virtual instance arrangement position determination device may be implemented as a Web server, or may be implemented as a cloud that provides a service related to the above-described virtual instance arrangement position determination by outsourcing.

  FIG. 15 is a diagram illustrating an example of a computer in which a virtual instance arrangement position determining device is realized by executing a program. The computer 1000 includes a memory 1010 and a CPU 1020, for example. The computer 1000 also includes a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. These units are connected by a bus 1080.

  The memory 1010 includes a ROM (Read Only Memory) 1011 and a RAM (Random Access Memory) 1012. The ROM 1011 stores a boot program such as BIOS (Basic Input Output System). The hard disk drive interface 1030 is connected to the hard disk drive 1090. The disk drive interface 1040 is connected to the disk drive 1100. For example, a removable storage medium such as a magnetic disk or an optical disk is inserted into the disk drive 1100. The serial port interface 1050 is connected to a mouse 1110 and a keyboard 1120, for example. The video adapter 1060 is connected to the display 1130, for example.

  The hard disk drive 1090 stores, for example, an OS 1091, an application program 1092, a program module 1093, and program data 1094. That is, the program that defines each process of the virtual instance arrangement position determination device is implemented as a program module 1093 in which a code executable by a computer is described. The program module 1093 is stored in the hard disk drive 1090, for example. For example, a program module 1093 for executing the same processing as the functional configuration in the virtual instance arrangement position determining apparatus is stored in the hard disk drive 1090. The hard disk drive 1090 may be replaced by an SSD (Solid State Drive).

  The setting data used in the processing of the above-described embodiment is stored as program data 1094 in, for example, the memory 1010 or the hard disk drive 1090. Then, the CPU 1020 reads the program module 1093 and the program data 1094 stored in the memory 1010 and the hard disk drive 1090 to the RAM 1012 and executes them as necessary.

  The program module 1093 and the program data 1094 are not limited to being stored in the hard disk drive 1090, but may be stored in, for example, a removable storage medium and read out by the CPU 1020 via the disk drive 1100 or the like. Alternatively, the program module 1093 and the program data 1094 may be stored in another computer connected via a network (LAN (Local Area Network), WAN (Wide Area Network), etc.). The program module 1093 and the program data 1094 may be read by the CPU 1020 from another computer via the network interface 1070.

DESCRIPTION OF SYMBOLS 1 Virtual instance arrangement system 10 Physical server control apparatus 20 Virtual instance arrangement position determination apparatus 30 Time clock 11, 21 Control part 12, 22 Storage part 111 Virtual instance resource usage collection part 112 Virtual instance arrangement position collection part 113 Virtual instance arrangement change Part 121 Physical Server Basic Information DB
122 Virtual Instance Resource Usage DB
123 Virtual instance location DB
124 Live migration status DB
211 Determination Unit 212 Usage Prediction Unit 213 Impact Prediction Unit 214 Calculation Unit 221 Usage Prediction DB
222 Impact prediction DB
223 Index DB

Claims (6)

Use of predicting the first resource usage at each time from the predetermined time to the predetermined time after the virtual instance based on the resource usage from the predetermined time to the predetermined time before the virtual instance operating on the physical server A quantity prediction unit;
Based on the resource usage and migration status of the virtual instance at the predetermined time, the second resource usage generated by the migration of the virtual instance at each time from the predetermined time to the time after the predetermined time, and the migration An impact prediction unit that predicts a remaining time during which the virtual instance is stopped due to the migration, and
Based on the first resource usage predicted by the usage prediction unit, the second resource usage predicted by the influence prediction unit, the remaining time, and the remaining stop time, the predetermined amount A calculation unit that calculates an index for evaluating the suitability of the virtual instance at the time as the migration destination for each physical server;
A virtual instance arrangement position determining apparatus comprising:   The virtual instance arrangement position determining apparatus according to claim 1, further comprising a determining unit that determines a physical server having the maximum index calculated by the calculating unit as a migration destination of a virtual instance.   The calculation unit is configured such that the first value that increases as the first resource usage amount and the second resource usage amount decrease, and a resource that has not been answered by a virtual instance stop based on the remaining stop time. The second value that increases as the usage amount decreases, and the index is calculated based on a sum of the first value and the second value. Virtual instance arrangement position determination device. The calculation unit further calculates a third value that increases as the number of the physical servers on which the virtual instances are arranged decreases.
The index is calculated based on a sum of a value obtained by multiplying a sum of the first value and the second value by a first coefficient and a value obtained by multiplying the third value by a second coefficient. ,
The virtual instance arrangement position determining apparatus according to claim 3, wherein a sum of the first coefficient and the second coefficient is a constant. Use of predicting the first resource usage at each time from the predetermined time to the predetermined time after the virtual instance based on the resource usage from the predetermined time to the predetermined time before the virtual instance operating on the physical server A quantity prediction process;
Based on the resource usage and migration status of the virtual instance at the predetermined time, the second resource usage generated by the migration of the virtual instance at each time from the predetermined time to the time after the predetermined time, and the migration An impact prediction step of predicting a remaining time during which the virtual instance is stopped due to the migration,
Based on the first resource usage predicted by the usage prediction step, the second resource usage estimated by the influence prediction step, the remaining time, and the remaining stop time, the predetermined amount A calculation step for calculating, for each physical server, an index for evaluating the suitability of the virtual instance as a migration destination at the time;
A virtual instance arrangement position determining method including:   A virtual instance arrangement position determination program for causing a computer to function as the virtual instance arrangement position determination apparatus according to any one of claims 1 to 4.

Images