Resource discovery program, resource discovery method, and information processing device.

The resource discovery program optimizes logical server configurations by selecting resource combinations with minimal switch crossovers, addressing performance degradation issues in CDI systems by efficiently utilizing hardware resources.

JP2026182214APending Publication Date: 2026-11-19FSAS TECH INC
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
JP2025078248
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2025-05-08
Publication Date
2026-11-19

AI Technical Summary

Technical Problem

Conventional technologies face challenges in configuring logical servers by deploying various types of hardware resources that are physically far apart, leading to potential performance degradation due to large physical distances and switch performance differences, making it difficult for users to select resources that maintain optimal performance.

Method used

A resource discovery program that receives logical server order information, acquires unused resources, selects optimal combinations based on virtual switch crossover estimation, and determines the resource candidates with the minimum estimated number of switch crossovers to ensure efficient resource utilization.

Benefits of technology

Enables the search for optimal resources from a resource pool, minimizing switch crossovers and bus overlaps, thereby maintaining performance and efficiency in logical server configurations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2026182214000001_ABST
    Figure 2026182214000001_ABST
Patent Text Reader

Abstract

The ability to search for the optimal resource from the resource pool. [Solution] The resource discovery program causes the computer to execute the following processes: receiving, acquiring, selecting, calculating, and deciding. The receiving process receives logical server order information from a user using a computer system composed of CDI. The acquiring process acquires information on unused resources among the resources included in the tenant. The selecting process selects one or more combinations of resource candidates from the unused resource information based on the order information. The calculating process calculates the estimated number of virtual switch crossovers in the path connecting the resource candidates in the selected combination, based on the number of crossovers for each type of virtual switch between resources and the weighting. The deciding process determines the resource candidate combination with the minimum estimated number of crossovers as the resource to be used.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a resource search program, a resource search method, and an information processing apparatus.

Background Art

[0002] A computer system composed of CDI (Composable Disaggregated Infrastructure) (hereinafter referred to as a CDI system) can change the configuration of a logical server by deploying and releasing hardware resources (hereinafter referred to as resources) in accordance with workloads.

[0003] Further, the related art includes a VM placement management system suitable for performing optimal VM placement that conforms to physical network specifications when creating virtual machines (VMs: Virtual Machines) on a cloud platform on which a plurality of business systems operate.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, while the conventional technologies described above can deploy VMs to physical machines by adapting them to the specifications of the physical network, it is difficult to configure logical servers by deploying various types of hardware resources that are physically far apart. On the other hand, when configuring logical servers by deploying various types of hardware resources from a hardware resource pool, the performance of resources may degrade if the physical distance between resources is large. Furthermore, even between resources that are physically close, the performance of resources may degrade due to differences in switch performance. For this reason, for example, when a user creates a logical server, they want to select resources so that the performance of the resources does not degrade, but it can be difficult for the user to decide which resources to select.

[0006] One aspect of this is to provide a resource discovery program, a resource discovery method, and an information processing device that can search for the optimal resource from the resource pool of a CDI system. [Means for solving the problem]

[0007] In one embodiment, the resource discovery program causes the computer to execute the following processes: receiving, acquiring, selecting, calculating, and deciding. The receiving process receives logical server order information from a user using a computer system composed of CDI. The acquiring process acquires information on unused resources among the resources included in the tenant. The selecting process selects one or more combinations of resource candidates from the acquired unused resource information based on the order information. The calculating process calculates the estimated number of virtual switch crossovers in the path connecting the resource candidates in the selected combination, based on the number of crossovers and weighting for each type of virtual switch between resources. The deciding process determines the resource candidate combination with the minimum estimated number of switch crossovers as the resource to be used, corresponding to the order information. [Effects of the Invention]

[0008] The optimal resources can be searched from the resource pool. [Brief explanation of the drawing]

[0009] [Figure 1] Figure 1 shows an example of generating a logical server from the resource pool of the embodiment. [Figure 2] Figure 2 shows an example of the resource hierarchy structure in the embodiment. [Figure 3] Figure 3 is a block diagram showing an example of the configuration of the information processing system in the embodiment. [Figure 4] Figure 4 is a block diagram showing an example of the functional configuration of the information processing device in the embodiment. [Figure 5] Figure 5 shows an example of a resource table storage unit. [Figure 6] Figure 6 shows an example of a logical server table storage unit. [Figure 7] Figure 7 shows an example of a management table storage unit. [Figure 8] Figure 8 shows an example of a cross-table storage unit. [Figure 9] Figure 9 shows an example of a weighted table storage unit. [Figure 10] Figure 10 shows an example of a bus table storage unit used. [Figure 11] Figure 11 shows an example of the hierarchical structure of the virtual switch before the logical server is generated in the embodiment. [Figure 12] Figure 12 shows an example of searching for the number of switches to cross during the generation of a logical server in the embodiment. [Figure 13] Figure 13 shows an example of searching for the number of switches to cross during the generation of a logical server in the embodiment. [Figure 14] Figure 14 shows an example of the number of bus overlaps during the creation of a logical server in the embodiment. [Figure 15] Figure 15 is a flowchart showing an example of the monitoring process in the embodiment. [Figure 16]Figure 16 is a flowchart illustrating an example of search processing according to an embodiment. [Figure 17] Figure 17 is a diagram illustrating an example of a computer that executes a resource search program. [Mode for Carrying Out the Invention]

[0010] Hereinafter, embodiments of a resource search program, a resource search method, and an information processing apparatus disclosed in the present application will be described in detail with reference to the accompanying drawings. The disclosed technology is not limited by the following embodiments. Further, the following embodiments may be appropriately combined within a consistent scope without contradiction. [Embodiment]

[0011] Figure 1 shows an example of generating a logical server from the resource pool of the embodiment. As shown in Figure 1, the resource pool 200 is prepared with various types of hardware resources. This embodiment is a resource discovery technique for a CDI system. Tenant administrators A and B of the CDI system, who are users, create and manage logical servers according to order information from the unused resources in the resource pool 200 included in their respective CDI system tenants #1 and #2. A tenant is a group unit used in the CDI system, and tenant administrators can reserve resources for tenants to prevent users from competing for resources from the resource pool. Reserved resources can be freely used by administrators (users) belonging to the tenant, and users can generate logical servers from these reserved resources. Resources are, for example, one or more types of resources from CPU (Central Processing Unit), GPU (Graphics Processing Unit), and SSD (Solid State Drive). In the example in Figure 1, when tenant administrator A creates a new logical server #2, he selects and reserves CPU #0, GPU #1, and SSD #1. In other words, tenant administrator A generates logical servers from unused resources in resource pool 200 according to the order information. Alternatively, tenant administrator A may add resources to existing logical servers from unused resources included in their own tenant #1. Note that resource pool 200 is also called a fabric pool.

[0012] Figure 2 is a diagram illustrating an example of the hierarchical structure of resources according to the embodiment. As shown in Figure 2, the resource group included in the resource pool 200 is connected via buses such that they form a hierarchical structure through virtual switches. Examples of types of the virtual switch include a PCIe switch, a PCIe fabric switch, an NVMe switch, and the like. To the PCIe fabric switch 202, the respective PCIe switches 206 and 209 of the PCIe boxes 204 and 205 are connected. To the PCIe switch 206, PCIe devices 207 and 208 which are resources are connected. The PCIe device 207 is, for example, a GPU. The PCIe device 208 is, for example, a NIC (Network Interface Card).

[0013] To the PCIe switch 209, the NVMe switch 211 within the PCIe device 210 is connected. The PCIe device 210 includes the NVMe switch 211 and NVMe SSDs 212 and 213 which are resources. To the NVMe switch 211, the NVMe SSDs 212 and 213 are connected. The NVMe switch 211 is a switch within the device, and is, for example, a switch that branches to the plurality of NVMe SSDs 212 and 213 present on a single PCIe device.

[0014] To the PCIe fabric switch, servers 214 and 215 which are resources are connected. The servers 214 and 215 are, for example, CPUs. The PCIe fabric switch, the PCIe switches 206 and 209, and the NVMe switch 211 are examples of various types of virtual switches with significantly different physical distances. By controlling the PCIe fabric switch 202, the PCIe switches 206 and 209, and the NVMe switch 211, a logical server can be generated from the unused resource group in the resource pool 200 according to order information. That is, the present embodiment has a configuration different from distributed computing performed by a network switch. Note that the more virtual switches are crossed, the longer the physical distance becomes.

[0015] Figure 3 is a block diagram showing an example of the configuration of an information processing system in an embodiment. As shown in Figure 3, the information processing system 1 includes a terminal device 10, an information processing device 100, and a resource pool 200.

[0016] The terminal device 10, the information processing device 100, and the resource pool 200 are connected via network N, enabling them to communicate with each other. Similarly, the information processing device 100 and the resource pool 200 are also connected via network N, enabling them to communicate with each other. Network N can be any type of communication network, including the internet, LAN (Local Area Network), or VPN (Virtual Private Network), whether wired or wireless. Furthermore, communication via network N may be encrypted, for example, by TLS (Transport Layer Security) / SSL (Secure Sockets Layer).

[0017] Information processing system 1 is an example of a system that can provide various services by generating logical servers from a group of resources within a resource pool 200 included in a tenant, through instructions given by an administrator belonging to the tenant to the information processing device 100 via a terminal device 10. In the following description, the administrator may be referred to as a user.

[0018] Terminal device 10 is an information processing device operated by an administrator (user) belonging to a tenant. Terminal device 10 can be, for example, a PC (Personal Computer). Terminal device 10 transmits order information of the logical server to the information processing device 100. Terminal device 10 can also access management information of the logical server in the information processing device 100.

[0019] The information processing device 100 manages resources within the resource pool 200 and also manages logical servers for each tenant. For example, when the information processing device 100 receives logical server order information from a terminal device 10, it searches for a resource corresponding to the order information from the group of resources within the resource pool 200 included in the tenant to which the user of the terminal device 10 belongs. Based on the search results, the information processing device 100 generates a logical server by controlling the PCIe fabric switch 202, PCIe switches 206, 209, and NVMe switch 211. Although the information processing device 100 and the resource pool 200 are shown separately in Figure 3, they may be integrated. Furthermore, the information processing device 100 may provide the terminal device 10 with an API (Application Programming Interface) for managing logical servers and resources within the resource pool 200. In other words, the information processing device 100 may include CDI management software.

[0020] The resource pool 200 has various types of resources, as shown in Figures 1 and 2. The resource pool 200 has resources such as CPUs, GPUs, and SSDs, and virtual switches that connect each resource in a hierarchical structure.

[0021] Next, the configuration of the information processing device 100 will be described. Figure 4 is a block diagram showing an example of the functional configuration of the information processing device in this embodiment. As shown in Figure 4, the information processing device 100 has a communication unit 110, a storage unit 120, and a control unit 130. In addition to the functional units shown in Figure 4, the information processing device 100 may also have various functional units that are known to be present in computers, such as various input devices and audio output devices.

[0022] The communication unit 110 is implemented, for example, by a NIC. The communication unit 110 is connected to the terminal device 10 and the resource pool 200 via the network N by wired or wireless means, and is a communication interface that manages the communication of information between the terminal device 10 and the resource pool 200. The communication unit 110 receives logical server order information and management information from the terminal device 10 via the network N. The communication unit 110 outputs the received logical server order information to the control unit 130. The communication unit 110 also transmits the logical server management information input from the control unit 130 to the terminal device 10 via the network N. The communication unit 110 also transmits resource control information input from the control unit 130 to the resource pool 200 via the network N. The communication unit 110 also outputs resource control information from the resource pool 200 to the control unit 130 via the network N. In other words, the communication unit 110 sends and receives resource control information between the control unit 130 and the resource pool 200.

[0023] The storage unit 120 is implemented by, for example, semiconductor memory elements such as RAM (Random Access Memory) and flash memory, or storage devices such as hard disks and optical discs. The storage unit 120 includes a resource table storage unit 121, a logical server table storage unit 122, a management table storage unit 123, a crossover count table storage unit 124, a weighting table storage unit 125, and a bus usage table storage unit 126. The storage unit 120 also stores information used for processing in the control unit 130.

[0024] The resource table storage unit 121 stores a resource table that shows the usage status of each resource in the resource pool 200. Figure 5 is a diagram showing an example of the resource table storage unit. As shown in Figure 5, the resource table storage unit 121 has items such as "No.", "Resource Name", "Type", "Usage Status", and "Failure". "No." is an identifier that identifies each resource included in the resource pool 200. "Resource Name" is a name unique to each resource. For example, for a CPU, the CPU serial number is represented by # followed by a number, in the format "CPU#0". "Type" is information indicating the type of resource. Examples of resource types include CPU, GPU, SSD, NIC, etc. Alternatively, a type that can identify GPUs with different processing capabilities may be used. "Usage Status" is information indicating whether or not the resource is being used by a logical server. "Failure" is information indicating whether or not the resource is failing. For example, "Failure" can be left blank if it is normal, and "1" if it is failing.

[0025] Returning to the explanation of Figure 4, the logical server table storage unit 122 stores a logical server table that shows the name of the logical server and its constituent resources. Figure 6 shows an example of the logical server table storage unit. As shown in Figure 6, the logical server table storage unit 122 has items such as "No.", "Logical Server Name", and "Constituent Resources". "No." is an identifier that identifies the logical server. "Logical Server Name" is the name of the logical server. "Constituent Resources" lists the resource names of the resources that make up each logical server. In the example in Figure 6, it is shown that the logical server name "machine1" is composed of CPU#0 and GPU#3.

[0026] Returning to the explanation of Figure 4, the management table storage unit 123 stores a management table that manages the hierarchical structure of resources. Figure 7 shows an example of the management table storage unit. As shown in Figure 7, the management table storage unit 123 has items such as "bus", "destination 1", and "destination 2". "Bus" is an identifier that identifies the bus that connects each resource in the hierarchical structure of resources to a virtual switch, or between virtual switches. In the example in Figure 7, the buses are represented as "b1" to "b18". "Desired destination 1" is information indicating the resource to which one of the buses is connected. "Desired destination 2" is information indicating the resource to which the other of the buses is connected. In "Desired destination 1" and "Desired destination 2", the virtual switches are represented as "sw1" to "sw12". For example, bus "b1" indicates that CPU#0 is connected to virtual switch sw1.

[0027] Returning to the explanation of Figure 4, the crossover count table storage unit 124 stores a crossover count table that shows the number of virtual switches that cross over each resource, for each switch type. Figure 8 is a diagram showing an example of the crossover count table storage unit. As shown in Figure 8, if, for example, there are two types of virtual switches, switch type α and β, the crossover count table storage unit 124 has a crossover count table 124a for switch type α and a crossover count table 124b for switch type β. In the crossover count table 124a, connectable resources are listed in row-direction items and column-direction items, and the number of crossovers of switch type α between resources ("number of α" in Figure 8) is represented by a brute-force method. Similarly, in the crossover count table 124b, connectable resources are listed in row-direction items and column-direction items, and the number of crossovers of switch type β between resources ("number of β" in Figure 8) is represented by a brute-force method. For example, it can be seen that the connection between CPU#0 and GPU#3 spans three switch type α and one switch type β. The crossover count table storage unit 124 is generated and updated based on the management table storage unit 123.

[0028] Returning to the explanation of Figure 4, the weighting table storage unit 125 stores a weighting table that shows weighting parameters normalized by the physical configuration (connection distance) for each switch type of virtual switch. Figure 9 shows an example of the weighting table storage unit. As shown in Figure 9, the weighting table storage unit 125 associates weights with each switch type of virtual switch. The weight is a parameter normalized by the physical configuration, and the closer the physical connection distance, the smaller the weight value. In the example in Figure 9, the weight value for switch type α is "0.2", which is smaller than the weight value for switch type β, which is "1", indicating that the physical connection distance is closer for switch type α than for switch type β. Furthermore, the weight values ​​may also reflect the performance of the virtual switch. The weighting table storage unit 125 is pre-configured.

[0029] Returning to the explanation of Figure 4, the bus usage table storage unit 126 stores the buses used between all resources within each logical server, and also stores a bus usage table that shows the number of times each bus is used before and after the creation of a new logical server. Figure 10 is a diagram showing an example of the bus usage table storage unit. As shown in Figure 10, the bus usage table storage unit 126 has circles around the buses used by the logical server "machine0" and the buses that "machine1" is scheduled to use, for example, for each bus "b1" to "b18". Note that "machine0" is composed of CPU#1 and GPU#1 in the hierarchical structure 220 described later. Also, "machine1" is the configuration when CPU#0 and GPU#3 are selected in the hierarchical structure 220 described later. "Total bus use (before machine1 creation)" in the bus usage table storage unit 126 represents the number of times each bus is used before the creation of "machine1". In the example in Figure 10, "Total bus use (before machine1 creation)" shows that buses "b2" and "b6" to "b9" are "1", while the other buses are "0". The "Total bus use (after machine1 creation)" in the bus usage table storage unit 126 represents the number of times each bus is used from the perspective of "machine1" after "machine1" is created. In the example in Figure 10, "Total bus use (after machine1 creation)" shows that buses "b1" and "b12" to "b14" are "1", "b6" is "2", and the other buses are "0". In other words, when "machine1" is created, bus "b6" overlaps with "machine0".

[0030] Returning to the explanation of Figure 4, the control unit 130 is implemented, for example, by a CPU or MPU (Micro Processing Unit) executing a program stored in an internal memory device using RAM as the working area. Alternatively, the control unit 130 may be implemented by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or FPGA (Field Programmable Gate Array).

[0031] The control unit 130 includes a reception unit 131, an acquisition unit 132, a selection unit 133, a calculation unit 134, a determination unit 135, a generation unit 136, and a monitoring unit 137, and realizes or executes the information processing functions and operations described below. Note that the internal configuration of the control unit 130 is not limited to the configuration shown in Figure 4, and other configurations are also acceptable as long as they perform the information processing described later.

[0032] The reception unit 131 receives order information for logical servers from administrators who are users belonging to the tenant. The order information includes the desired specifications for the logical server. For example, for a GPU, the specifications may include the model number and the capacity of VRAM (Video Random Access Memory). For an SSD, the specifications may include the storage capacity. The reception unit 131 outputs the received order information to the selection unit 133. The reception unit 131 also outputs an acquisition instruction to the acquisition unit 132.

[0033] When the acquisition unit 132 receives an acquisition instruction from the reception unit 131, it refers to the resource table storage unit 121 and acquires information on unused resources assigned to the user's tenant. In other words, the acquisition unit 132 acquires information on unused resources among the resources included in the user's tenant. The acquisition unit 132 outputs the acquired unused resource information to the selection unit 133.

[0034] The selection unit 133 receives order information from the reception unit 131 and unused resource information from the acquisition unit 132. Based on the input order information, the selection unit 133 selects one or more combinations of resource candidates from the unused resource information. In other words, the selection unit 133 selects one or more combinations of resource candidates from the unused resource information for resources that correspond to the specification conditions of the order information.

[0035] The selection unit 133 assigns numbers i and j to unused CPU resources and GPU resources, respectively, among the unused resource information. i and j are counters for unused CPU resources and GPU resources, and their upper limits are the maximum number of unused resources. The selection unit 133 initializes the counters by setting i=0 and j=0.

[0036] The selection unit 133 selects the i-th CPU resource. Next, the selection unit 133 selects the j-th GPU resource. The selection unit 133 outputs combination information indicating the combination of the selected i-th CPU resource and j-th GPU resource to the calculation unit 134.

[0037] When combination information is input from the selection unit 133, the calculation unit 134 refers to the crossover count table storage unit 124 and obtains the number of crossovers for each type of virtual switch in the path between the i-th CPU resource and the j-th GPU resource included in the combination information. The calculation unit 134 also refers to the weighting table storage unit 125 and obtains the weights for each type of virtual switch in the said path. Based on the obtained number of crossovers and weights for each type of virtual switch, the calculation unit 134 calculates the estimated number of switch crossovers for the selected combination in the said path. The calculation unit 134 calculates the estimated number of switch crossovers using, for example, the following formula (1). The calculation unit 134 outputs the combination information and the calculated estimated number of switch crossovers to the determination unit 135.

[0038]

number

[0039] Here, n corresponds to the number of virtual switch types. For example, if there are two types of switches, α and β, then n=2. That is, as shown in equation (1) above, the calculation unit 134 calculates the estimated number of switch crossovers by adding a value obtained by multiplying the number of crossovers by the weight for each type of virtual switch, for each type of virtual switch.

[0040] In other words, the calculation unit 134 calculates the estimated number of virtual switch crossovers in the path connecting the candidate resources in the selected combination, based on the number of crossovers and weighting for each type of virtual switch between the resources.

[0041] When the determination unit 135 receives combination information and the estimated number of switches crossed from the calculation unit 134, it determines whether the estimated number of switches crossed is the minimum. If the determination unit 135 determines that the estimated number of switches crossed is not the minimum, it does not consider the combination of the i-th CPU resource and the j-th GPU resource included in the combination information as a combination candidate. If the determination unit 135 determines that the estimated number of switches crossed is the minimum, it determines whether there is already an estimated number of switches crossed with the same value.

[0042] If the determination unit 135 determines that there are no switch-crossing estimates with the same value, it sets the combination of the i-th CPU resource and the j-th GPU resource included in the combination information as a provisional combination. Note that, during the initial determination (i=0, j=0), the determination unit 135 determines the combination to be determined as a provisional combination (resource candidate). The determination unit 135 temporarily stores the combination information of the provisional combination in the storage unit 120.

[0043] If the determination unit 135 determines that there is already an estimated number of switches crossed with the same value, it refers to the usage bus table storage unit 126 and obtains the usage status of buses used by other logical servers. The determination unit 135 also refers to the usage bus table storage unit 126 and obtains the usage status of buses for combinations of CPU resources and GPU resources (existing provisional combinations) with the same estimated number of switches crossed. The determination unit 135 counts the number of overlapping buses based on the usage status of buses used by other logical servers and the usage status of buses for existing provisional combinations. The determination unit 135 also counts the number of overlapping buses based on the usage status of buses used by other logical servers and the usage status of buses for combinations of the i-th CPU resource and j-th GPU resource included in the combination information (the current combination).

[0044] The decision unit 135 compares the number of overlapping buses in the existing provisional combination with the number of overlapping buses in the current combination to determine whether the number of overlapping buses in the current combination is less than the number of overlapping buses in the provisional combination. In other words, the decision unit 135 determines whether the current combination has fewer overlapping buses. If the decision unit 135 determines that the number of overlapping buses in the current combination is less than the number of overlapping buses in the provisional combination, it sets the current combination as the provisional combination. If the decision unit 135 determines that the number of overlapping buses in the current combination is equal to or greater than the number of overlapping buses in the provisional combination, it continues to use the existing provisional combination as the provisional combination.

[0045] The determination unit 135, after completing the determination of the number of switches crossed and the number of overlapping buses for the combination of the i-th CPU resource and the j-th GPU resource included in the combination information, determines whether or not there is a (j+1)th GPU resource. If the determination unit 135 determines that there is a (j+1)th GPU resource, it increments j and instructs the selection unit 133 to select the next combination and repeat the determination of the number of switches crossed and the number of overlapping buses. If the determination unit 135 determines that there is no (j+1)th GPU resource, it determines whether or not there is an (i+1)th CPU resource. If the determination unit 135 determines that there is an (i+1)th CPU resource, it increments i and instructs the selection unit 133 to select the next combination and repeat the determination of the number of switches crossed and the number of overlapping buses.

[0046] If the determination unit 135 determines that there is no (i+1) CPU resource, it determines a provisional combination of CPU and GPU resources as the resources to be used corresponding to the order information. The determination unit 135 outputs the information of the determined resources to the generation unit 136. The information of the resources to be used is an example of a converted order form and describes the specific resources in the resource pool 200. In addition to the optimization using the iterative calculation described above, the determination unit 135 may also use other search methods such as a graph database.

[0047] In other words, if there are multiple combinations that result in the minimum number of switches crossed, the determination unit 135 determines, based on the bus usage status along the route, how many buses used by other logical servers are included in the route for each combination. The determination unit 135 then determines the resource candidate for the combination with the fewest such buses included in the route as the resource to be used. In short, the determination unit 135 determines the resource candidate for the combination that results in the minimum number of switches crossed as the resource to be used corresponding to the order information.

[0048] When the generation unit 136 receives information on the resources to be used from the determination unit 135, it sends and receives control information for the resources to the resource pool 200 and the resources to generate logical servers from the group of resources in the resource pool 200 based on the information on the resources to be used. In other words, the generation unit 136 generates logical servers based on the determined resources to be used. The generation unit 136 stores the information of the generated logical servers in the resource table storage unit 121, the logical server table storage unit 122, and the bus usage table storage unit 126.

[0049] The monitoring unit 137 monitors the resource group within the resource pool 200. The monitoring unit 137 starts monitoring the resource group in the resource pool 200 at predetermined intervals (approximately 1 day to 1 week). The monitoring unit 137 compares the results with the previous check to determine whether there are any failed resources. If the monitoring unit 137 determines that there are failed resources, it deletes the information of the failed resources from the management table storage unit 123. The monitoring unit 137 also updates the resource table storage unit 121, the crossover count table storage unit 124, and the usage bus table storage unit 126 to delete the information of the failed resources.

[0050] If the monitoring unit 137 determines that there are no faulty resources, it compares the current results with the previous check results to determine whether there have been any additions or deletions of resources. If the monitoring unit 137 determines that there have been additions or deletions of resources, it adds or deletes information about the added or deleted resources to the management table storage unit 123. The monitoring unit 137 also updates the resource table storage unit 121, the crossover count table storage unit 124, and the usage bus table storage unit 126 to add or delete information about the added or deleted resources. If the monitoring unit 137 determines that there have been no additions or deletions of resources, it terminates the monitoring process. In other words, the monitoring unit 137 monitors resources and updates the resource information if there are faulty resources, added resources, or deleted resources. Note that the addition and deletion of resources here refer to, for example, attaching or detaching hardware resources such as SSD cards (NVNeSSD212,213) or GPU cards (PCIe device 207) to or from PCIe boxes 204,205.

[0051] Next, an example of the search process will be explained using Figures 11 to 14. Figure 11 is a diagram showing an example of the hierarchical structure of the virtual switch before the logical server of the embodiment is generated. The hierarchical structure 220 shown in Figure 11 has CPU #0, #1 and GPU #0 to #4 as resources, and each resource is connected by virtual switches sw1 to sw12 and buses b1 to b18. The connection between each resource and each virtual switch is represented by the management table storage unit 123 shown in Figure 7. In Figure 11, the virtual switches sw1, sw4, sw5, and sw8, which are represented by rectangles, are switch type α, and the virtual switches sw2, sw3, sw6, sw7, sw9 to sw12, which are represented by circles, are switch type β.

[0052] In hierarchical structure 220, CPU#1 and GPU#1 are already assigned to logical server#0, and CPU#0 and GPU#0, GPU#2~GPU#4 are unused resources. Now, let's consider the resource search when creating logical server#1 which includes one CPU resource and one GPU resource as order information.

[0053] Figure 12 shows an example of searching for the estimated number of switches crossed during the generation of a logical server in the embodiment. As shown in Figure 12, the selection unit 133 selects each combination of CPU#0 and GPU#0, GPU#2 to GPU#4. The calculation unit 134 calculates the estimated number of switches crossed for each combination. For example, in the combination of CPU#0 and GPU#3, the calculation unit 134 calculates the estimated number of switches crossed for the path 221 shown by the dotted line, which crosses the four virtual switches sw1, sw4, sw8, and sw9 in the hierarchical structure 220. Note that the number of virtual switches crossed from CPU#0 is 3 for GPU#0, 4 for GPU#2, and 5 for GPU#4. In this case, if only the number of virtual switches crossed is considered, GPU#0, which has the minimum number of switches crossed, will be selected.

[0054] In route 221, there are three switch type αs and one switch type β. Therefore, the calculation unit 134 refers to the weighting table storage unit 125 and obtains the weight value "0.2" for switch type α and the weight value "1" for switch type β. For example, switch type α is a network switch and switch type β is a PCIe switch. Based on the above formula (1), the calculation unit 134 calculates the estimated number of switches crossed as "1.6". Similarly, the calculation unit 134 calculates the estimated number of switches crossed for each combination of CPU#0 and GPU#0, GPU#2, and GPU#4.

[0055] Figure 13 shows an example of searching for the estimated number of switches crossed during the generation of a logical server in this embodiment. As shown in Figure 13, among the estimated number of switches crossed for each combination, the combination of CPU#0 and GPU#2 and the combination of CPU#0 and GPU#3 have the same minimum value of "1.6" for the estimated number of switches crossed. Also, the estimated number of switches crossed for the combination of CPU#0 and GPU#0 is "2.2", and for the combination of CPU#0 and GPU#4 it is "3.4". If we only consider the number of virtual switches crossed, the combination of CPU#0 and GPU#0 would be selected, but in this embodiment, the switch type is taken into account. Therefore, the determination unit 135 selects two combinations as candidates: the combination of CPU#0 and GPU#2, and the combination of CPU#0 and GPU#3. Note that the combination of CPU#0 and GPU#2 results in a path 222 that crosses four virtual switches sw1, sw4, sw5, and sw7.

[0056] Figure 14 shows an example of the number of bus overlaps during the creation of a logical server in the embodiment. As shown in Figure 14, the determination unit 135 obtains the bus usage status of logical server #0 because there are the same estimated number of switch crossovers. Logical server #0 has a route 223 that crosses four virtual switches sw1, sw4, sw5, and sw6, so it uses buses b2, b6, b7, b8, and b9 (see Figures 10 and 11). On the other hand, the combination of CPU #0 and GPU #2 uses buses b1, b6, b7, b10, and b11 (see Figure 11). For the combination of CPU #0 and GPU #2, if we calculate the values ​​for buses b1 to b18 in "Total bus use (after machine1 creation)" in Figure 10 assuming they are selected, buses b1, b10, and b11 will be "1", buses b6 and b7 will be "2", and the others will be "0" (not shown in Figure 10). If the combination of CPU#0 and GPU#2 is selected, the sum of the values ​​for buses b1 to b18 in "Total bus use (after machine1 is created)" will be "7". In other words, the number of overlapping buses is "2" because buses b6 and b7 overlap. Also, the combination of CPU#0 and GPU#3 uses buses b1, b6, b12, b13, and b14 (see Figures 10 and 11). For the combination of CPU#0 and GPU#3, if we calculate the values ​​for buses b1 to b18 in "Total bus use (after machine1 is created)" in Figure 10 assuming that it has been selected, buses b1, b12 to b14 will be "1", bus b6 will be "2", and the others will be "0" (corresponding to the example in Figure 10). If the combination of CPU#0 and GPU#3 is selected, the sum of the values ​​for buses b1 to b18 in "Total bus use (after machine1 is created)" will be "6". In other words, the number of overlapping buses is "1" because bus b6 is overlapping.

[0057] The determination unit 135 compares the bus used by logical server #0 with the bus used by the combination of CPU #0 and GPU #2, and the bus used by the combination of CPU #0 and GPU #3, and counts the number of overlapping buses. The number of overlapping buses is, for example, the sum of the values ​​for buses b1 to b18 in the "Total bus use (after machine1 generation)" mentioned above. In the calculation example above, the number of overlapping buses is "7" for the combination of CPU #0 and GPU #2, and "6" for the combination of CPU #0 and GPU #3. Therefore, the determination unit 135 determines the combination of CPU #0 and GPU #3, which has the fewest overlapping buses, as the resource to be used corresponding to the order information. Alternatively, the determination unit 135 may compare the number of overlapping buses instead of the number of overlapping buses. By performing this search process, the optimal resource can be found from the resource pool 200. Subsequently, the generation unit 136 generates logical server #1 using the combination of CPU #0 and GPU #3 in the resource pool 200.

[0058] Furthermore, if a new logical server uses multiple GPU resources, the decision unit 135 may select the GPU resource that is physically closest. Also, the decision unit 135 may prioritize route combinations that result in fewer bus overlaps with other logical servers than the estimated number of switches used when determining the resources to be used.

[0059] Next, the operation of the information processing device 100 in this embodiment will be described. First, the resource monitoring process will be explained using Figure 15. Figure 15 is a flowchart showing an example of the monitoring process in this embodiment.

[0060] The monitoring unit 137 starts monitoring the resource pool 200 (step S1). The monitoring unit 137 determines whether or not there are any failed resources (step S2). If the monitoring unit 137 determines that there are failed resources (step S2: Yes), it deletes the information of the failed resources from the management table storage unit 123 (step S3). The monitoring unit 137 also updates the resource table storage unit 121, the crossover count table storage unit 124, and the usage bus table storage unit 126 to delete the information of the failed resources (step S4), and then terminates the monitoring process.

[0061] On the other hand, if the monitoring unit 137 determines that there are no failed resources (step S2: No), it determines whether there have been any additions or deletions of resources (step S5). If the monitoring unit 137 determines that there have been additions or deletions of resources (step S5: Yes), it adds or deletes information about the added or deleted resources to the management table storage unit 123 (step S6). The monitoring unit 137 also updates the resource table storage unit 121, the crossover count table storage unit 124, and the usage bus table storage unit 126 to add or delete information about the added or deleted resources (step S4), and then terminates the monitoring process. If the monitoring unit 137 determines that there have been no additions or deletions of resources (step S5: No), it terminates the monitoring process. As a result, the information processing device 100 can reflect information about failed, added, or deleted resources in the resource table storage unit 121, the management table storage unit 123, the crossover count table storage unit 124, and the usage bus table storage unit 126.

[0062] Next, we will explain the search process when generating a logical server using Figure 16. Figure 16 is a flowchart showing an example of the search process in the embodiment.

[0063] The reception unit 131 receives order information for the logical server from the administrator, who is a user belonging to the tenant (step S11). The reception unit 131 outputs the received order information to the selection unit 133. The reception unit 131 also outputs an acquisition instruction to the acquisition unit 132.

[0064] When the acquisition unit 132 receives an acquisition instruction from the reception unit 131, it refers to the resource table storage unit 121 and acquires information on unused resources (step S12). The acquisition unit 132 outputs the acquired unused resource information to the selection unit 133.

[0065] When the selection unit 133 receives unused resource information from the acquisition unit 132, it assigns the numbers i and j to the unused CPU resources and GPU resources, respectively (step S13). The selection unit 133 initializes the counters by setting i=0 and j=0 (step S14).

[0066] The selection unit 133 selects the i-th CPU resource (step S15). Next, the selection unit 133 selects the j-th GPU resource (step S16). The selection unit 133 outputs combination information indicating the combination of the selected i-th CPU resource and j-th GPU resource to the calculation unit 134.

[0067] When combination information is input from the selection unit 133, the calculation unit 134 refers to the crossover count table storage unit 124 and obtains the number of crossovers for each type of virtual switch in the path between the i-th CPU resource and the j-th GPU resource included in the combination information. The calculation unit 134 also refers to the weighting table storage unit 125 and obtains the weights for each type of virtual switch in that path. Based on the obtained number of crossovers and weights for each type of virtual switch, the calculation unit 134 calculates the estimated number of switch crossovers for the path in the selected combination (step S17). The calculation unit 134 outputs the combination information and the calculated estimated number of switch crossovers to the determination unit 135.

[0068] When the determination unit 135 receives combination information and the estimated number of switches to cross from the calculation unit 134, it determines whether the estimated number of switches to cross is the minimum (step S18). If the determination unit 135 determines that the estimated number of switches to cross is not the minimum (step S18: No), the combination of the i-th CPU resource and the j-th GPU resource included in the combination information is not considered a combination candidate (step S19), and the process proceeds to step S26.

[0069] If the determination unit 135 determines that the estimated number of switches crossing is the minimum (step S18: Yes), it determines whether there is already an estimated number of switches crossing with the same value (step S20). If the determination unit 135 determines that there is no estimated number of switches crossing with the same value (step S20: No), it sets the combination of the i-th CPU resource and the j-th GPU resource included in the combination information as a provisional combination (step S21) and proceeds to step S26.

[0070] If the determination unit 135 determines that there is already an estimated number of switches crossed with the same value (step S20: Yes), it refers to the usage bus table storage unit 126 and obtains the usage status of buses used by other logical servers. The determination unit 135 also refers to the usage bus table storage unit 126 and obtains the usage status of buses for combinations of CPU resources and GPU resources (existing provisional combinations) with the same estimated number of switches crossed (step S22).

[0071] The decision unit 135 counts the number of bus overlaps between the existing provisional combination and the current combination included in the combination information, based on the bus usage status of other logical servers (step S23).

[0072] The decision unit 135 determines whether the current combination has fewer overlapping buses (step S24). If the decision unit 135 determines that the current combination has fewer overlapping buses (step S24: Yes), it sets the current combination as the provisional combination (step S25) and proceeds to step S26. If the decision unit 135 determines that the number of overlapping buses in the current combination is greater than or equal to the number of overlapping buses in the provisional combination (step S24: No), it continues to proceed to step S26 with the existing provisional combination as the provisional combination.

[0073] The decision unit 135 determines whether or not there is a (j+1)th GPU resource (step S26). If the decision unit 135 determines that there is a (j+1)th GPU resource (step S26: Yes), it increments j (step S27) and returns to step S16. If the decision unit 135 determines that there is no (j+1)th GPU resource (step S26: No), it determines whether or not there is an (i+1)th CPU resource (step S28). If the decision unit 135 determines that there is an (i+1)th CPU resource (step S28: Yes), it increments i (step S29) and returns to step S15.

[0074] If the determination unit 135 determines that there is no i+1 CPU resource (step S28: No), it determines a provisional combination of CPU and GPU resources as the resources to be used corresponding to the order information (step S30). The determination unit 135 outputs the information of the determined resources to the generation unit 136.

[0075] When the generation unit 136 receives information on the resources to be used from the determination unit 135, it generates a logical server based on the information on the resources to be used (step S31). The generation unit 136 stores the information of the generated logical server in the resource table storage unit 121, the logical server table storage unit 122, and the bus usage table storage unit 126. This allows the information processing device 100 to search for the optimal resource from the resource pool 200.

[0076] In this way, the information processing device 100 receives logical server order information from users utilizing the computer system composed of CDI. The information processing device 100 also obtains information on unused resources among the resources included in the tenant. Based on the order information, the information processing device 100 selects one or more combinations of resource candidates from the obtained unused resource information. The information processing device 100 also calculates the estimated number of virtual switch crossovers in the path connecting the resource candidates in the selected combination, based on the number of crossovers and weighting for each type of virtual switch between resources. The information processing device 100 then determines the resource candidate combination with the minimum estimated number of switch crossovers as the resource to be used corresponding to the order information. As a result, the information processing device 100 can search for the optimal resource from the resource pool 200.

[0077] Furthermore, the information processing device 100 calculates the estimated number of switches crossed by multiplying the number of switches crossed by a weight for each type of virtual switch and adding this value for each type of virtual switch. As a result, the information processing device 100 can search for the optimal resource from the resource pool 200, taking into account the physical distance and performance of the virtual switches.

[0078] Furthermore, if there are multiple combinations that result in the minimum number of switches crossed, the information processing device 100 determines, for each combination, the number of buses used by other logical servers included in the route, based on the bus usage status along the route. The information processing device 100 also determines the resource candidates for combinations with a small number of buses included in the route as resources to be used. As a result, the information processing device 100 can search for the optimal resources from the resource pool 200, taking into account the resource usage status of other logical servers.

[0079] Furthermore, the resources are connected via buses in a hierarchical structure through virtual switches. As a result, the information processing device 100 can search for the optimal resource from the resource pool 200, which contains resources at different physical distances.

[0080] Furthermore, the information processing device 100 monitors resources and updates resource information if there are any failed resources, added resources, or removed resources. Here, adding and removing resources refers to, for example, attaching or detaching hardware resources such as SSD cards (NVNeSSD212,213) or GPU cards (PCIe device 207) to PCIe boxes 204,205. As a result, the information processing device 100 can reflect the failed, added, or removed resources in the resource table storage unit 121, the management table storage unit 123, the crossover count table storage unit 124, and the usage bus table storage unit 126.

[0081] Furthermore, the resources are one or more types of resources from among CPU, GPU, and SSD. As a result, the information processing device 100 can generate a logical server corresponding to the order information.

[0082] Furthermore, the types of virtual switches include PCIe fabric switches, PCIe switches, and NVMe switches. As a result, the information processing device 100 can search for the optimal resource from the resource pool 200, which contains resources located at different physical distances.

[0083] In the above embodiment, a logical server was generated by performing a resource search based on order information for generating a new logical server, but this is not limited to that. For example, it may be applied to resource searches when a user adds resources to an existing logical server. Also, for example, if it is determined in the monitoring process that resources have been added, a search process may be performed on the existing logical server to search for more optimal resources.

[0084] Furthermore, although the information processing device 100 generated the logical server in the above embodiment, the invention is not limited to this. For example, the information processing device 100 may transmit information about the resources to be used, which it has determined, to another information processing device such as a switching device, and the other information processing device such as the switching device may generate the logical server (switch the configuration).

[0085] Furthermore, the components of each part shown in the diagram do not necessarily have to be physically configured as depicted. In other words, the specific forms of distribution and integration of each part are not limited to those shown in the diagram, and all or part of them can be functionally or physically distributed and integrated in any unit according to various loads, usage conditions, etc.

[0086] Furthermore, the various processing functions performed by each device may be executed in whole or in part on a CPU (or a microcomputer such as an MPU or MCU (Micro Controller Unit)). It goes without saying that the various processing functions may also be executed in whole or in part on a program analyzed and executed by the CPU (or a microcomputer such as an MPU or MCU), or on wired logic hardware.

[0087] Incidentally, the various processes described in the above embodiment can be achieved by executing a pre-prepared program on a computer. Therefore, below we will describe an example of a computer that executes a program having the same functions as in the above embodiment. Figure 17 is a diagram showing an example of a computer that executes a resource discovery program.

[0088] As shown in Figure 17, the computer 300 includes a CPU 301 that performs various calculations, an input device 302 that accepts data input, and a monitor 303. The computer 300 also includes a media reader 304 that reads programs and the like from a storage medium, an interface device 305 for connecting to various devices, and a communication device 306 for connecting to other information processing devices via wired or wireless connections. The computer 300 also includes a RAM 307 for temporarily storing various information and a hard disk drive 308. Each of the devices 301 to 308 is connected to a bus 309.

[0089] The hard disk drive 308 stores a resource search program having the same functions as the receiving unit 131, acquisition unit 132, selection unit 133, calculation unit 134, determination unit 135, generation unit 136, and monitoring unit 137 shown in Figure 4. The hard disk drive 308 also stores a resource table storage unit 121, a logical server table storage unit 122, a management table storage unit 123, a crossover count table storage unit 124, a weighting table storage unit 125, and a bus usage table storage unit 126. The input device 302 receives various information inputs, such as operation information, from, for example, the administrator of the computer 300. The monitor 303 displays various screens, such as the display screen, to, for example, the administrator of the computer 300. The interface device 305 is connected to, for example, a printing device. The communication device 306 has the same functions as, for example, the communication unit 110 shown in Figure 4, is connected to the network N, and exchanges various information with the terminal device 10, the resource pool 200, and other information processing devices.

[0090] The CPU 301 reads each program stored in the hard disk drive 308, loads it into the RAM 307, and executes it to perform various processes. These programs can also cause the computer 300 to function as the receiving unit 131, acquisition unit 132, selection unit 133, calculation unit 134, determination unit 135, generation unit 136, and monitoring unit 137 shown in Figure 4.

[0091] The resource discovery program described above does not necessarily need to be stored on the hard disk drive 308. For example, the computer 300 may read and execute a program stored on a storage medium that the computer 300 can read. Examples of storage media that the computer 300 can read include portable recording media such as CD-ROMs, DVDs (Digital Versatile Discs), and USB (Universal Serial Bus) memory, semiconductor memory such as flash memory, and hard disk drives. Alternatively, the resource discovery program may be stored on a device connected to a public network, the internet, or a LAN, and the computer 300 may read and execute the resource discovery program from there. [Explanation of symbols]

[0092] 1. Information Processing System 10 Terminal devices 100 Information Processing Devices 110 Communications Department 120 Storage section 121 Resource Table Storage Unit 122 Logical Server Table Storage Unit 123 Management Table Storage Unit 124 Cross-table storage unit 125 Weighted Table Storage Unit 126 Usage Bus Table Storage Unit 130 Control Unit 131 Reception Department 132 Acquisition Department 133 Selection Section 134 Calculation Section 135 Decision Section 136 Generation part 137 Monitoring Department 200 resource pools N Network

Claims

1. The system receives order information for logical servers from users utilizing a computer system composed of CDI (Composable Disaggregated Infrastructure). Among the resources included in the aforementioned tenant, information on unused resources is obtained. Based on the aforementioned order information, one or more combinations of resource candidates are selected from the previously acquired unused resource information. Based on the number of virtual switch crossovers and weighting for each type of virtual switch between the resources, the estimated number of virtual switch crossovers in the path connecting the candidate resources in the selected combination is calculated. The resource candidate for the combination that minimizes the number of switches crossed is determined as the resource to be used corresponding to the order information. A resource discovery program characterized by having a computer perform the processing.

2. The calculation process involves multiplying the number of crossovers by the weighting for each type of virtual switch, and adding this value to the number of virtual switch types to calculate the estimated number of crossovers. The resource discovery program according to feature 1.

3. The process for making the determination, if there are multiple combinations that result in the minimum number of switches crossed, determines, based on the bus usage status on the path, the number of buses used by other logical servers included in the path for each combination, and determines the resource candidate for the combination with the fewest buses included in the path as the resource to be used. A resource discovery program according to claim 1 or 2, characterized by the above.

4. The aforementioned resources are connected by buses in a hierarchical structure via the virtual switch. A resource discovery program according to claim 1 or 2, characterized by the above.

5. The system monitors the aforementioned resources and updates the resource information if any resources have failed, been added, or been removed. A resource discovery program according to claim 1 or 2, characterized by the above.

6. The aforementioned resources are one or more types of resources from among CPU (Central Processing Unit), GPU (Graphics Processing Unit), and SSD (Solid State Drive). A resource discovery program according to claim 1 or 2, characterized by the above.

7. The types of virtual switches are PCIe fabric switches, PCIe switches, and NVMe switches. A resource discovery program according to claim 1 or 2, characterized by the above.

8. We accept order information for logical servers from users using computer systems composed of CDI (Composable Disaggregated Infrastructure). Among the resources included in the aforementioned tenant, information on unused resources is obtained. Based on the aforementioned order information, one or more combinations of resource candidates are selected from the previously acquired unused resource information. Based on the number of virtual switch crossovers and weighting for each type of virtual switch between the resources, the estimated number of virtual switch crossovers in the path connecting the candidate resources in the selected combination is calculated. The resource candidate for the combination that minimizes the number of switches crossed is determined as the resource to be used corresponding to the order information. A resource search method characterized by having a computer perform the processing.

9. A reception desk that receives order information for logical servers from users of a computer system composed of CDI (Composable Disaggregated Infrastructure), An acquisition unit that acquires information on unused resources among the resources included in the tenant, Based on the order information, a selection unit selects one or more combinations of resource candidates from the acquired unused resource information, A calculation unit calculates the estimated number of virtual switch crossovers in the path connecting the candidate resources in the selected combination, based on the number of crossovers and weighting of virtual switches for each type between the resources, A determination unit that determines the resource candidate for the combination that minimizes the number of switches crossed, as the resource to be used corresponding to the order information, An information processing device characterized by having the following features.

Citation Information

Patent Citations

  • VM assignment management system and VM assignment management method

    WO2017056238A1