Resource allocation assistance device and resource allocation assistance method
The resource allocation support device addresses uncertainty in future supply and demand by estimating distributions and optimizing allocations, ensuring appropriate resource allocation in uncertain environments.
Patent Information
- Application Number
- PCT/JP2025/017330
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-05-13
- Publication Date
- 2026-11-19
AI Technical Summary
Existing resource allocation technologies struggle to address uncertainty in future resource supply and demand, particularly in large-scale resource-sharing services like public clouds, leading to inappropriate allocations due to high uncertainty in user demands and network failures.
A resource allocation support device that estimates future resource supply and demand distributions, calculates prerequisite supply and demand considering risk tolerance, and optimizes resource allocation using single or multi-objective optimization based on these estimates.
Enables appropriate resource allocation in uncertain environments by considering risk tolerance for excessive demand and insufficient supply, reducing the risk of oversupply while maintaining demand within expected limits.
Smart Images

Figure JP2025017330_19112026_PF_FP_ABST
Abstract
Description
Resource allocation support device and resource allocation support method
[0001] This disclosure relates to a resource allocation support device and a resource allocation support method.
[0002] Resource-sharing services (e.g., public clouds) are known that are designed for multiple users to share computing and network resources. Furthermore, technologies are known for allocating computing and network resources to users in a way that optimizes predetermined metrics for resource-sharing services (e.g., Non-Patent Documents 1 and 2).
[0003] T. Cucinotta et al., "Data Center Optimization Enhanced by Software Defined Networking," in Procs. of CLOUD, 2014. D. Saxena et al., "OP-MLB: An Online VM Prediction-Based Multi-Objective Load Balancing Framework for Resource Management at Cloud Data Center," IEEE Trans. Cloud Comput., Vol.10, No.4, pp.2804-2816, Oct. 2022.
[0004] However, with prior art, including the technologies described in Non-Patent Documents 1 and 2, it is difficult to make appropriate resource allocations under circumstances where future resource supply and demand are uncertain.
[0005] This disclosure is made in view of the above points and aims to enable appropriate resource allocation even under circumstances where future resource supply and demand are uncertain.
[0006] A resource allocation support device according to one aspect of the present disclosure includes: a supply and demand distribution estimation unit that estimates the future distribution of resource supply and demand based on time-series information relating to resource supply and demand; a prerequisite supply and demand calculation unit that calculates prerequisite resource supply and demand representing the resource supply and demand assumed for resource allocation based on the future distribution of resource supply and demand and the risk tolerance for excessive demand and insufficient supply of resources; and a solution calculation unit that calculates information representing the resource allocation that optimizes the indicator based on the prerequisite resource supply and demand and the indicator to be optimized, as a solution.
[0007] It enables appropriate resource allocation even in situations where future resource supply and demand are uncertain.
[0008] This figure shows an example of the hardware configuration of a resource allocation support device according to one embodiment. This figure shows an example of the functional configuration of a resource allocation support device according to one embodiment. This flowchart shows an example of resource allocation support processing according to one embodiment.
[0009] One embodiment of the present invention will be described in detail below with reference to the drawings.
[0010] <Conventional Technology and its Problems> There are technologies that allocate computing resources and network resources to users in order to optimize predetermined metrics for resource-sharing services (for example, Non-Patent Documents 1 and 2, etc.).
[0011] For example, Non-Patent Document 1 discloses a technique for assigning physical machines to virtual machines used by users and assigning physical links to user-generated flows, based on the application's resource requirements, network topology, and the network state predicted at the time of request acceptance, in order to minimize the application's response time.
[0012] Furthermore, for example, Non-Patent Document 2 discloses a technique for allocating computing resources in such a way as to minimize an objective function that is expressed as a linear combination of the power consumption of a physical machine, the communication costs between interdependent virtual machines, and the utilization rate of computing resources.
[0013] However, with prior art, including the technologies described in Non-Patent Documents 1 and 2, it is difficult to make appropriate resource allocations under circumstances where future resource supply and demand are uncertain.
[0014] For example, the technology disclosed in Non-Patent Document 1 allocates resources such as physical machines and physical links based on the resource supply and demand between the request observed immediately before resource allocation and the network state predicted at the time the request is received (i.e., based on static resource supply and demand at a certain point in time). However, resource-sharing services, such as public clouds, are expected to have extremely high uncertainty in terms of resource demand because a large number of users utilize a wide variety of applications provided by a large number of tenants. Furthermore, in large-scale resource-sharing services, the number of physical machines and physical links constituting the service becomes enormous, and it is expected that there will be uncertainty in terms of resource sharing due to failures, etc. For this reason, technologies that allocate resources based on static resource supply and demand cannot perform appropriate resource allocation that addresses the uncertainty of resource supply and demand.
[0015] On the other hand, for example, the technology disclosed in Non-Patent Document 2 proposes a load balancing approach based on online resource forecasting to achieve flexible resource allocation in response to fluctuations in resource demand. This load balancing approach uses a neural network-based online forecaster to predict resource usage at the next point in time with high accuracy. While the load balancing approach based on online resource forecasting is considered effective when resource allocation is repeated at very short time intervals, it cannot provide appropriate resource allocation that responds to the uncertainty of resource supply and demand when resource allocation is performed at time intervals with a certain range where resource supply and demand are likely to fluctuate significantly.
[0016] Therefore, the following describes a resource allocation support device 10 that can perform appropriate resource allocation even under circumstances where future resource supply and demand are uncertain. By using this resource allocation support device 10, for example, it becomes possible to perform appropriate resource allocation that addresses uncertainty even in resource-sharing services where future resource supply and demand are uncertain.
[0017] <Example of Hardware Configuration of Resource Allocation Support Device 10> Figure 1 is a diagram showing an example of the hardware configuration of a resource allocation support device 10 according to one embodiment. As shown in Figure 1, the resource allocation support device 10 according to one embodiment includes an input device 101, a display device 102, an external I / F 103, a communication I / F 104, a RAM (Random Access Memory) 105, a ROM (Read Only Memory) 106, an auxiliary storage device 107, and a processor 108. Each of these hardware components is connected to each other so as to be able to communicate via a bus 109.
[0018] The input device 101 is, for example, a keyboard, mouse, touch panel, or physical button. The display device 102 is, for example, a display or display panel. The resource allocation support device 10 does not necessarily have to have at least one of the input device 101 and the display device 102.
[0019] The external I / F 103 is an interface with external devices such as the recording medium 103a. Examples of recording media 103a include CDs (Compact Discs), DVDs (Digital Versatile Disks), SD memory cards (Secure Digital memory cards), and USB (Universal Serial Bus) memory cards.
[0020] The communication interface 104 is an interface for connecting to a communication network. The RAM 105 is a volatile semiconductor memory (storage device) that temporarily holds programs and data. The ROM 106 is a non-volatile semiconductor memory (storage device) that can retain programs and data even when the power is turned off. The auxiliary storage device 107 is a non-volatile storage device such as an HDD (Hard Disk Drive), SSD (Solid State Drive), or flash memory. The processor 108 is a computing device such as a CPU (Central Processing Unit) or GPU (Graphic Processing Unit).
[0021] Note that the hardware configuration shown in Figure 1 is just one example, and the hardware configuration of the resource allocation support device 10 is not limited to this. For example, the resource allocation support device 10 may have multiple auxiliary storage devices 107 or multiple processors 108, it may not have some of the hardware shown, or it may have various hardware other than the hardware shown.
[0022] <Example of Functional Configuration of Resource Allocation Support Device 10> Figure 2 is a diagram showing an example of the functional configuration of a resource allocation support device 10 according to one embodiment. As shown in Figure 2, the resource allocation support device 10 according to one embodiment includes a time-series input unit 201, a supply and demand distribution estimation unit 202, a prerequisite supply and demand calculation unit 203, an objective function input unit 204, a solution calculation unit 205, and an output unit 206. Each of these units is realized, for example, by processing that one or more programs installed in the resource allocation support device 10 are executed by a processor 108 or the like.
[0023] The time-series input unit 201 inputs time-series data of user demand for resources and resource supply for user demand at each point in time (hereinafter also referred to as "past point in time") included in a predetermined past period (hereinafter also referred to as "past period") as resource supply and demand time-series information. That is, the time-series input unit 201 inputs time-series data of user demand and resource supply at each past point in time included in the past period as resource supply and demand time-series information. Resources refer to computing resources (e.g., physical machines, etc.) or network resources (e.g., physical links, etc.) or both. Hereinafter, the past period is a period that includes a predetermined number of past points in time up to the latest resource allocation point, and it is assumed that resource allocation will be performed at the latest resource allocation point. The resource allocation point is the point in time when resources are allocated to a user or the point in time when resources are allocated to a user.
[0024] Here, resource supply can include, for example, the number and usage of CPUs and GPUs in a physical machine, as well as the capacity of memory and storage, if the resource is a computing resource, and the bandwidth of the physical link and the capacity of the buffers of physical nodes (e.g., routers, switches, etc.) if the resource is a network resource. On the other hand, user demand can include, for example, the actual amount of resources used relative to the resource supply.
[0025] The supply and demand distribution estimation unit 202 estimates the distribution of user demand and resource supply during the period until the next resource allocation time (hereinafter also referred to as the "resource allocation period") based on the resource supply and demand time series information input by the time series input unit 201. That is, the supply and demand distribution estimation unit 202 estimates the distribution of user demand and resource supply from the resource supply and demand time series information input by the time series input unit 201, and then uses this distribution of user demand and resource supply as the distribution of user demand and resource supply during the resource allocation period.
[0026] For example, the past period is defined as a period containing K past time points up to the latest resource allocation time point t, with the past time points being t-K+1, t-K+2, ..., t. The next resource allocation time point is defined as t+r. In this case, the supply and demand distribution estimation unit 202 estimates the distribution of user demand and resource supply during the resource allocation period [t+1, t+r] based on the user demand and resource supply at past time points t-K+1, t-K+2, ..., t. For example, the supply and demand distribution estimation unit 202 sequentially estimates the distribution of user demand and resource supply from the user demand and resource supply at past time points t-K+1, t-K+2, ..., t, thereby using the distribution of user demand and resource supply at past time point t as the distribution of user demand and resource supply during the resource allocation period.
[0027] Here, the supply and demand distribution estimation unit 202 may, for example, assume a parametric distribution for user demand and resource supply, and then estimate the parameters of that distribution using maximum likelihood estimation or Bayesian estimation to estimate the distribution of user demand and resource supply during the resource allocation period. Alternatively, the supply and demand distribution estimation unit 202 may, for example, estimate a nonparametric distribution for user demand and resource supply using kernel density estimation to estimate the distribution of user demand and resource supply during the resource allocation period. Furthermore, the supply and demand distribution estimation unit 202 may, for example, estimate the underlying state and distribution behind the observed resource supply and demand time series information using a state-space model or a Kalman filter to estimate the distribution of user demand and resource supply during the resource allocation period.
[0028] Hereinafter, the distribution of user demand and resource supply during the resource allocation period will also be referred to as the "estimated user demand distribution" and the "estimated resource supply distribution," respectively. Furthermore, the estimated user demand distribution and the estimated resource supply distribution will be collectively referred to as the "estimated resource supply and demand distribution."
[0029] The prerequisite supply and demand calculation unit 203 calculates the user demand amount assumed at the time of the latest resource allocation (hereinafter also referred to as "prerequisite user demand") and the resource supply amount assumed at the time of the latest resource allocation (hereinafter also referred to as "prerequisite resource supply") based on the estimated resource supply and demand distribution estimated by the supply and demand distribution estimation unit 202. At this time, the prerequisite supply and demand calculation unit 203 calculates the prerequisite user demand amount and the prerequisite resource supply amount while considering the risk tolerance for the risk of resource over-demand and under-supply. Hereinafter, the prerequisite user demand amount and the prerequisite resource supply amount will be collectively referred to as "prerequisite resource supply and demand."
[0030] Here, one possible method for calculating the assumed user demand and assumed resource supply, taking into account the risk tolerance for the risk of resource over-demand and under-supply, is as follows:
[0031] First, as risk tolerance, we determine a risk tolerance p% for the risk of user demand exceeding prior expectations (i.e., the risk of excessive demand) and a risk tolerance q% for the risk of resource supply falling below prior expectations (i.e., the risk of supply shortage). One way to determine the risk tolerance p is to clarify the relationship between the risk tolerance for excessive resource demand and the risk of SLA violations across the entire service based on past performance data, and then determine the risk tolerance p so that the risk of Service Level Agreement (SLA) violations remains within a predetermined upper limit. Similarly, one way to determine the risk tolerance q is to clarify the relationship between the risk tolerance for resource shortages and the risk of SLA violations across the entire service based on past performance data, and then determine the risk tolerance q so that the risk of SLA violations remains within a predetermined upper limit.
[0032] Next, when the risk tolerance for the risk of excessive resource demand is set to p%, that is, when p% of the risk of user demand exceeding the preliminary assumption is allowed, the upper p% point of the estimated user demand distribution estimated by the supply-demand distribution estimation unit 202 is defined as the prerequisite user demand. Similarly, when the risk tolerance for the risk of resource supply shortage is set to q%, that is, when q% of the risk of resource supply falling below the preliminary assumption is allowed, the lower q percent point of the estimated resource supply distribution estimated by the supply-demand distribution estimation unit 202 is defined as the prerequisite resource supply amount.
[0033] Note that the method of determining the above risk tolerances p and q is merely an example, and other methods such as a method of determining by back-calculating from the upper limit of the excess demand risk in the entire service are also conceivable. For example, consider a service configured by user u i , user u j , and physical machine h k . At this time, the risk tolerance for the risk that the user demand amount of user u i exceeds the assumption is set to p i %, the risk tolerance for the risk that the user demand amount of user u j exceeds the assumption is set to p j %, and the risk tolerance for the risk that the supply amount of physical machine h k falls below the assumption is set to q k , respectively. The probability that the sum of the user demand amount of user u i and the user demand amount of user u j falls within the resource supply amount as assumed for physical machine h k is (1-p i )(1-p j )(1-q k ). Therefore, the probability that excess demand occurs is 1-(1-p i )(1-p j )(1-q k ). Accordingly, using this probability, in view of the excess demand risk in the entire service, the respective risk tolerances p of user u i , user u j , and physical machine h k i , pj , q k You just need to decide that.
[0034] In addition to the above, for example, risk tolerance levels p and q can also be determined based on past performance data.
[0035] The objective function input unit 204 receives as an objective function one or more functions representing one or more optimization metrics to be optimized when allocating at least one of computing resources and network resources to a user (i.e., allocating resource supply in relation to user demand). Examples of objective functions include functions that represent metrics similar to those in conventional technology (e.g., power consumption of physical machines, overload of physical machines, communication costs between virtual machines, utilization rate of computing resources, etc.).
[0036] The solution calculation unit 205 calculates the optimal solution for resource allocation to each user (hereinafter also referred to as the "resource allocation solution") by single-objective optimization or multi-objective optimization, based on the assumed user demand and assumed resource supply calculated by the assumed supply and demand calculation unit 203 and the single or multiple objective functions input by the objective function input unit 204. At this time, the solution calculation unit 205 may calculate the resource allocation solution using any method used for single-objective optimization or multi-objective optimization (e.g., mathematical programming, evolutionary algorithms, deep learning, etc.). The resource allocation solution includes, for example, at least one of the following: information representing the allocation of computing resources to each user, and information representing the allocation of network resources to each user that constitute the path between each user and the computing resources to which that user is allocated.
[0037] Furthermore, when performing multi-objective optimization based on multiple objective functions, the set of Pareto optimal solutions calculated by the multi-objective optimization method may be used as the optimal solution, or a set of unique or minority solutions among the Pareto optimal solutions that satisfy a predetermined first condition may be used as the optimal solution. Here, a Pareto optimal solution is an optimal solution in which improving the value of one objective function requires worsening the value of at least one other objective function.
[0038] The output unit 206 outputs the resource allocation solution calculated by the solution calculation unit 205 to a predetermined output destination (e.g., a resource allocation device that allocates resources to each user).
[0039] In the example shown in Figure 2, the resource allocation support device 10, implemented on a single information processing device (computer), has all the functional units, but this is just one example. For example, the resource allocation support device 10 may be implemented on multiple information processing devices, with each functional unit distributed among multiple information processing devices. In this case, the resource allocation support device 10 may be called a resource allocation support system or the like.
[0040] <Example of Resource Allocation Support Processing> Figure 3 is a flowchart showing an example of resource allocation support processing according to one embodiment. The resource allocation support processing shown in Figure 3 is executed repeatedly each time a predetermined second condition is met. Specific examples of the predetermined second condition include, for example, a change in user demand for resources, or a change in the amount of resources supplied to user demand due to the addition or failure of resources (physical machines, physical links). In addition, various arbitrary conditions can be used as the predetermined second condition, such as a predetermined amount of time (e.g., 1 hour, 1 day, 1 week, etc.) having elapsed since the previous resource allocation support processing was executed.
[0041] The time-series input unit 201 inputs time-series data of user demand and resource supply at each past point in time included in the past period as resource supply and demand time-series information (step S101). As an example, the following is the CPU usage D at K past points in time t-K+1, t-K+2, ..., t up to resource allocation time t, as time-series data of user demand for a physical machine. u (t-K+1), D u (t-K+2),...,D u Assume that (t) is input. Similarly, as time-series data of resource supply in relation to user demand, consider the amount of CPU available at K past time points t-K+1, t-K+2, ..., t up to resource allocation time t, S h (t - K + 1), S h(t-K+2),...,S h Assume that (t) is entered.
[0042] The supply and demand distribution estimation unit 202 estimates the estimated resource supply and demand distribution based on the resource supply and demand time series information input in step S101 (step S102). As an example, CPU usage D u (t-K+1), D u (t-K+2),...,D u CPU usage d during the future resource allocation period [t+1, t+r] estimated from (t) u The distribution of (i.e., the estimated user demand distribution) is f u (d u ) Similarly, the amount of CPU available S h (t - K + 1), S h (t-K+2),...,S h The amount of CPU available during the future resource allocation period [t+1, t+r] estimated from (t) is s. h The distribution of (i.e., the estimated resource supply distribution) is f h (s h )
[0043] The prerequisite supply and demand calculation unit 203 calculates the prerequisite resource supply and demand based on the estimated resource supply and demand distribution estimated in step S102 above, taking into account the risk tolerance for the risk of resource over-demand and under-supply (step S103). As an example, CPU usage d u Distribution f u (d u The assumed CPU usage (an example of assumed user demand) calculated from ) is D u * , CPU available amount s h Distribution f h (s h The assumed CPU usage amount (an example of assumed resource supply amount) calculated from ) is S h * Let's assume that.
[0044] Here, as a method for calculating the assumed CPU usage and assumed usable CPU amount, taking into account the risk tolerance for excess demand and supply shortage, for example, if the risk tolerance for the risk of CPU usage exceeding prior expectations is p%, then the CPU usage d u Distribution f u (d u ) Top p percentile D u * This can be used as the assumed CPU usage. Similarly, if the risk tolerance for the risk of the available CPU amount falling below the prior assumption is q%, then the available CPU amount s h Distribution f h (s h ) The lower q percentile point S h * This should be considered the available CPU capacity.
[0045] The objective function input unit 204 receives as an objective function one or more functions representing one or more optimization metrics to be optimized when allocating at least one of computing resources and network resources to a user (step S104). Hereinafter, let τ = t+1, t+2, ..., t+r. Also, the total overload of all physical machines at time τ is f. 1 (τ), the total power consumption of all physical machines is f 2 Let (τ) be the objective function. For example, the objective function is the total overload F of all physical machines during the future resource allocation period [t+1, t+r]. 1 (t) = f 1 (t+1)+f 1 (t+2)+...+f 1 (t+r) and the total power consumption F of all physical machines 2 (t) = f 2 (t+1)+f 2 (t+2)+...+f 2 Let (t+r) be the case.
[0046] The solution calculation unit 205 calculates a resource allocation solution by single-objective optimization or multi-objective optimization based on the prerequisite resource supply and demand amounts calculated in step S103 above and the single or plurality of objective functions input in step S104 above (step S105). For example, the solution calculation unit 205 uses the prerequisite CPU usage amount D u * and the prerequisite available CPU amount S h * and the total overload F 1 (t) and the total power consumption F 2 (t), to calculate a resource allocation solution by multi-objective optimization.
[0047] Note that in step S105 above, for example, N input in step S104 above obj objective functions F i (t) (i=1, ..., N obj ) are simultaneously optimized, and a Pareto optimal solution set S is calculated as a set of information representing the allocation of each physical machine h to each user u such that pareto (t) = {X pareto j (t) | j=1, ..., M pareto} is calculated. Here, M pareto is the number of Pareto optimal solutions. Further, X pareto j (t) = (x u,h (t)) is the j-th Pareto optimal solution at the resource allocation time point t, and takes the form of a matrix of the number of users × the number of physical machines. x u,h (t) is a decision variable representing the allocation of the physical machine h to the user u at the resource allocation time point t. For example, when x u,h (t) = 0, it represents that the physical machine h is not allocated to the user u at the resource allocation time point t, and when x u,h (t) = 1, it represents that the physical machine h is allocated to the user u at the resource allocation time point t.
[0048] Note that the relationship between the decision variable x representing the allocation of each physical machine h to each user u u,h (t) and the objective function F i (t) may, for example, be written down as a mathematical expression as follows, and the decision variable xu,h (t) and the objective function F i (t), may be represented by a model constructed by a method such as statistics / machine learning / deep learning based on observation data related thereto.
[0049] Total overload of all physical machines in the future resource allocation target period [t+1, t+r]: F 1 (t) = f 1 (t+1) + f 1 (t+2) + ... + f 1 (t+r) Total overload of all physical machines at the future time point τ∈[t+1, t+r]: f 1 (τ) = Σ h max{D h (τ) - S h (τ), 0} Here, D h (τ) is the load (CPU usage) of the physical machine h at the future time point τ, S h (τ) is the capacity (available CPU usage) of the physical machine h at the future time point τ.
[0050] Total power consumption of all physical machines in the future resource allocation target period [t+1, t+r]: F 2 (t) = f 2 (t+1) + f 2 (t+2) + ... + f 2 (t+r) Total power consumption of all physical machines at the future time point τ∈[t+1, t+r]: f 2 (τ) = Σ h P h (τ) Here, P h (τ) is the power consumption of the physical machine h at the future time point τ, and is represented by the following equation.
[0051] Power consumption of the physical machine h at the future time point τ: P h (τ) = max{P idle + (P busy - P idle )U h (τ), P busy} Here, P idle is the power consumption of the physical machine h at idle, P busy is the peak power consumption of the physical machine h, U h(τ) is the CPU utilization of the physical machine h at a future time point τ, and is expressed by the following formula:
[0052] CPU utilization of physical machine h at future time point τ: U h (τ) = min{D} h (τ) / S h (τ), 1) Note that D is the CPU usage of the physical machine h at a future time point τ. h (τ) is the amount of CPU usage D by each user u at a future point in time τ, as shown in the following equation. u (τ) and the decision variable x at resource allocation time t. u,h (t) is determined by and
[0053] CPU usage of physical machine h at future point in time τ: D h (τ) = Σ u D u (τ)x u,h (t) Also, CPU usage D by each user u u (τ) and the amount of CPU available to each physical machine h S h (τ) is a predicted value at a future time τ, and therefore the assumed CPU usage D is as follows: u * and prerequisite CPU capacity S h * Use D. u (τ) = D u * S h (τ) = S h * Let's assume that.
[0054] As mentioned above, there are multiple objective functions F i (t) (i=1,...,N obj When multi-objective optimization is performed using ), the set of Pareto optimal solutions S calculated by the multi-objective optimization method is pareto (t) = {X pareto j (t)}(j=1,...,M pareto ) can be used as the optimal solution, or a unique solution X that satisfies a predetermined first condition from among the Pareto optimal solutions included in the Pareto optimal solution set. unique (t) or minority solution set S sub (t) may be considered the optimal solution.
[0055] For example, regarding the first condition, "Minimize the changes (migration) of the physical machine h assigned to each user u", the resource allocation time t-T is the time immediately preceding the most recent resource allocation time. intr (However, T intr The unique solution X calculated using the formula (where r is the interval between resource allocation points) unique (t-T intr ) and the set of Pareto optimal solutions S calculated at the latest resource allocation time t. pareto Each Pareto optimal solution X included in (t) pareto j (t) (j=1,...,M pareto The Pareto optimal solution X that minimizes the number of migrations when compared with the other options. pareto j (t) is the unique solution X at the latest resource allocation time t. unique It may also be calculated as (t).
[0056] For example, if there are multiple Pareto optimal solutions that satisfy the first condition, then the set of minority solutions S may exist. sub (t) is calculated. However, this is just one example; for instance, the set of a predetermined number of Pareto optimal solutions that satisfy the first condition is the set of minority solutions S. sub It may also be calculated as (t).
[0057] Furthermore, by performing single-objective optimization using a single objective function F(t), the unique solution X can be directly obtained. unique (t) can be calculated and considered the optimal solution.
[0058] The output unit 206 outputs the resource allocation solution (Pareto optimal solution set S) calculated in step S105 above. pareto (t), unique solution X unique (t), or a minority solution set S sub (t)) is output to a predetermined output destination (step S106). Note that the output destination is not limited to a specific destination, but specific examples of such output destinations include a display device 102 such as a display, a resource allocation device that allocates resources to each user based on the resource allocation solution, etc.
[0059] <Summary> As described above, the resource allocation support device 10 according to one embodiment can support appropriate resource allocation in response to the uncertainty of resource supply and demand in resource-sharing services such as public clouds, when allocating resources at time intervals with a certain range in which large fluctuations in resource supply and demand are likely to be observed. For this reason, the resource allocation support device 10 according to one embodiment enables resource-sharing service providers to allocate resources while considering their risk tolerance for excessive demand and insufficient supply of resources, even in situations where future resource supply and demand are uncertain. As a result, it becomes possible to avoid oversupply while keeping the risk of excess demand within an expected range.
[0060] <Supplement> The functions of the elements disclosed herein may be implemented using circuits or processing circuits that include general-purpose processors, special-purpose processors, integrated circuits, ASICs (Application Specific Integrated Circuits), FPGAs (Field Programmable Gate Arrays), conventional circuits, and / or combinations thereof that are programmed using one or more programs stored in one or more memories, or otherwise configured to perform the disclosed functions. A processor is considered a circuit or processing circuit because it includes transistors and other circuits. A processor may be a programmed processor that executes programs stored in memory. In this disclosure, a circuit, unit, or means is hardware that performs the enumerated functions, or hardware programmed to perform the enumerated functions. Hardware may be any hardware disclosed herein that is programmed or configured to perform the enumerated functions.
[0061] There is a memory for storing a computer program that includes computer instructions. These computer instructions provide logic and routines that enable hardware (e.g., circuits or processing circuits) to perform the methods disclosed herein. This computer program can be implemented in commonly known forms as a computer-readable storage medium, a computer program product, a memory device, a recording medium such as a CD-ROM or DVD, and / or memory of an FPGA or ASIC.
[0062] The following additional information is disclosed regarding the embodiments described above.
[0063] (Note 1) A resource allocation support device comprising a memory and at least one processor connected to the memory, wherein the processor estimates the future distribution of resource supply and demand based on time-series information relating to resource supply and demand, calculates a prerequisite resource supply and demand representing the resource supply and demand assumed for resource allocation based on the future distribution of resource supply and demand and a risk tolerance for excessive demand and insufficient supply of resources, and calculates information representing the resource allocation that optimizes the indicator as a solution based on the prerequisite resource supply and demand and an indicator to be optimized. (Note 2) The resource allocation support device according to Note 1, wherein the processor estimates the future distribution of resource supply and demand by sequentially estimating the distribution of resource supply and demand at each point in time included in a predetermined past period based on the time-series information. (Note 3) The resource allocation support device according to Note 1, wherein the processor calculates the minimum or maximum resource supply and demand that satisfies the risk tolerance among the resource supply and demand included in the future distribution of resource supply and demand as the prerequisite resource supply and demand. (Note 4) The resource allocation support device according to any one of Notes 1 to 3, wherein the processor calculates the premise resource supply and demand based on the risk tolerance determined in relation to the risk of violation of service quality standards in the service supplying the resources. (Note 5) The resource allocation support device according to Note 1, wherein the index is a value calculated from the resource supply and demand in the future period, and the processor calculates the solution using the premise resource supply and demand as the resource supply and demand in the future period. (Note 6) A non-temporary storage medium storing a program capable of executing resource allocation support processing by a computer, wherein the resource allocation support processing estimates the distribution of future resource supply and demand based on time-series information relating to resource supply and demand, calculates premise resource supply and demand representing the resource supply and demand assumed for resource allocation based on the distribution of future resource supply and demand and the risk tolerance for excessive demand and insufficient supply of resources, and calculates information representing the resource allocation that optimizes the index as a solution based on the premise resource supply and demand and the index to be optimized.
[0064] The present invention is not limited to the embodiments specifically disclosed above, and various modifications, changes, and combinations with known technologies are possible without departing from the spirit of the claims.
[0065] 10 Resource allocation support device 101 Input device 102 Display device 103 External I / F 103a Recording medium 104 Communication I / F 105 RAM 106 ROM 107 Auxiliary storage device 108 Processor 109 Bus 201 Time series input unit 202 Supply and demand distribution estimation unit 203 Premise supply and demand calculation unit 204 Objective function input unit 205 Solution calculation unit 206 Output unit
Claims
1. A resource allocation support device comprising: a supply and demand distribution estimation unit that estimates the future distribution of resource supply and demand based on time-series information relating to resource supply and demand; a prerequisite supply and demand calculation unit that calculates prerequisite resource supply and demand representing the resource supply and demand assumed for resource allocation based on the future distribution of resource supply and demand and the risk tolerance for excessive demand and insufficient supply of resources; and a solution calculation unit that calculates information representing the resource allocation that optimizes the indicator based on the prerequisite resource supply and demand and the indicator to be optimized, as a solution.
2. The resource allocation support device according to claim 1, wherein the supply and demand distribution estimation unit estimates the future resource supply and demand distribution by sequentially estimating the resource supply and demand distribution at each point in time included in a predetermined period in the past, based on the time-series information.
3. The resource allocation support device according to claim 1, wherein the premise supply and demand calculation unit calculates the minimum or maximum resource supply and demand that satisfies the risk tolerance among the resource supply and demand included in the future resource supply and demand distribution as the premise resource supply and demand.
4. The resource allocation support device according to any one of claims 1 to 3, wherein the prerequisite supply and demand calculation unit calculates the prerequisite resource supply and demand based on the risk tolerance determined in relation to the risk of violation of service quality standards in the service that supplies the resources.
5. The resource allocation support device according to claim 1, wherein the index is a value calculated from the supply and demand of resources in the future period, and the solution calculation unit calculates the solution using the assumed resource supply and demand as the resource supply and demand in the future period.
6. A resource allocation support method in which a computer performs the following steps: a supply and demand distribution estimation procedure for estimating the future distribution of resource supply and demand based on time-series information on resource supply and demand; a premise supply and demand calculation procedure for calculating premise resource supply and demand that represents the resource supply and demand assumed for resource allocation, based on the future distribution of resource supply and demand and the risk tolerance for excessive demand and insufficient supply of resources; and a solution calculation procedure for calculating information representing the resource allocation that optimizes the indicator, based on the premise resource supply and demand and the indicator to be optimized.
