# US 11,113,629 B2 · Adaptive configuration of a heterogeneous cluster environment > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,113,629 B2 - Document ID: US11113629B2 - Status: Granted - Kind: B2 - Application number: 15/859,542 - Filed: 2017-12-31 - Published: 2019-02-14 - Granted: 2021-09-07 - Earliest claimed: 2017-08-14 - Assignees: International Business Machines Corporation - Applicants: International Business Machines Corporation - Inventors: Hitham Ahmed Assem Aly Salama; Teodora Sandra Buda; Faisal Ghaffar; Lei Xu - Web page: https://patentgrep.com/patent/US11113629B2 ## CPC - G06N 20/00: Machine learning - G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound - H04L 41/0886: {Fully automatic configuration} - H04L 41/145: {involving simulating, designing, planning or modelling of a network} - H04L 41/16: using machine learning or artificial intelligence - H04L 41/24: {using dedicated network management hardware} - H04L 41/0803: Configuration setting - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M) ## Abstract An embodiment includes a method for use in managing a system comprising one or more computers, each computer comprising at least one hardware processor coupled to at least one memory, the method comprising a computer-implemented manager: generating a potential configuration for hardware resources of the system; determining whether the potential configuration satisfies accuracy and time constraints for a selected machine learning model; if the potential configuration satisfies the constraints, indicating the potential configuration to be the optimal configuration for the system; and if the potential configuration does not satisfy the constraints, adapting the potential configuration to satisfy the constraints. The adapting may comprise repeating the generating and determining steps. The adapting may be based at least in part on the hardware resources and the selected machine learning model. ## Claims ### Claim 1 A method for use in managing a system comprising one or more computers, each computer comprising at least one hardware processor coupled to at least one memory, the method comprising a computer-implemented manager: generating a plurality of potential configurations for hardware resources of the system for each of a plurality of machine learning models; determining, for each of the plurality of machine learning models, the potential configuration of the plurality of potential configurations that exhibits a minimal execution time while satisfying accuracy and time constraints for the corresponding machine learning model, wherein the determining step comprises configuring the system according to a selected potential configuration, training the corresponding machine learning model on the configured system, and recording time required for training and accuracy reached; sorting, based on execution time to train the corresponding machine learning model of the plurality of machine learning models, the plurality of machine learning models and corresponding plurality of potential configurations that exhibit the minimal execution time; and configuring at least a subset of the hardware resources based on an optimal machine learning model of the plurality of sorted machine learning models. ### Claim 2 (depends on 1) The method of claim 1, wherein the determining is based at least in part on the hardware resources and a selected machine learning model. ### Claim 3 (depends on 1) The method of claim 1, further comprising the step of, once a selected potential configuration is indicated to be the optimal configuration for the system, configuring the system according to the optimal configuration. ### Claim 4 (depends on 1) The method of claim 1, further comprising the step of, once a selected potential configuration is indicated to be the optimal configuration for the system, saving the optimal configuration for the system to a configuration database. ### Claim 5 (depends on 4) The method of claim 4, further comprising the step of, if the configuration database includes a saved optimal configuration for the system, configuring the system according to the optimal configuration without generating additional potential configurations. ### Claim 6 (depends on 5) The method of claim 5, wherein configuring the system according to the optimal configuration comprises training the machine learning model on the configured system. ### Claim 7 (depends on 1) The method of claim 1, wherein the determining step comprises utilizing a selected potential configuration to estimate time required for training the machine learning model. ### Claim 8 (depends on 1) The method of claim 1, wherein a selected potential configuration for the system is generated based at least in part on a fixed set of available hardware resources. ### Claim 9 (depends on 1) The method of claim 1, further comprising selecting the machine learning model from a set of available machine learning models. ### Claim 10 (depends on 9) The method of claim 9, further comprising configuring the system based at least in part on the selected model and the optimal configuration for the selected model. ### Claim 11 (depends on 9) The method of claim 9, wherein the machine learning model is selected based at least in part on the accuracy and time constraints for the machine learning model. ### Claim 12 (depends on 9) The method of claim 9, wherein selecting the machine learning model from the set of available machine learning models comprises: determining a subset of the set of models which satisfies accuracy and time constraints; and selecting at least one of the subset which satisfies the constraints. ### Claim 13 (depends on 12) The method of claim 12, wherein the optimal configuration for the given model minimizes a training time of the given model relative to one or more other configurations. ### Claim 14 (depends on 13) The method of claim 13, wherein determining whether at least the given one of the set of available models satisfies accuracy and time constraints comprises estimating the training time of the given model based on the optimal configuration for the given model. ### Claim 15 (depends on 12) The method of claim 12, wherein selecting at least one of the subset which satisfies the constraints comprises the step of selecting the at least one model based at least in part on its position within the plurality of sorted machine learning models. ### Claim 16 (depends on 15) The method of claim 15, wherein the plurality of sorted machine learning models is sorted based at least in part on computational complexity. ### Claim 17 (depends on 15) The method of claim 15, wherein the plurality of sorted machine learning models is sorted based at least in part on training time. ### Claim 18 (depends on 15) The method of claim 15, wherein the plurality of sorted machine learning models is sorted based on both computational complexity and training time. ### Claim 19 (depends on 1) The method of claim 1, wherein at least one of the potential configurations for hardware resources is defined in a virtualized environment and comprises one or more virtual machines having a specified number of virtual central processing units, RAM and storage capacity of a specified size, a specified operating system, a specified hardware architecture, and specified network settings.