# US 11,907,191 B2 · Content based log retrieval by using embedding feature extraction > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,907,191 B2 - Document ID: US11907191B2 - Status: Granted - Kind: B2 - Application number: 17/722,902 - Filed: 2022-04-18 - Published: 2023-10-19 - Granted: 2024-02-20 - Earliest claimed: 2022-04-18 - Assignees: Dell Products L.P. - Applicants: Dell Products L.P. - Inventors: Dale Wang; Min Gong; Ashok Narayanan Potti - Web page: https://patentgrep.com/patent/US11907191B2 ## CPC - G06F 16/2237: {Vectors, bitmaps or matrices} - G06F 18/23: Clustering techniques - G06F 17/40: Data acquisition and logging (for input to computer G06F3/00) - G06N 20/00: Machine learning - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract Methods and systems for managing data processing systems are disclosed. A data processing system may include hardware and/or software components. The operation of the data processing system may depend on the operation of these components. To manage the operation of the data processing system, a system may include a data processing system manager. The data processing system manager may obtain logs for components of the data processing system reflecting the historical operation of these components and use the log to predict the future operation of the data processing system, identify similar operation of other data processing systems, and/or for other purposes. Based on the predictions, the data processing system manager may take action to reduce the likelihood of the data processing system becoming impaired. ## Claims ### Claim 1 A computer-implemented method for managing data processing systems, the method comprising: obtaining logs for the data processing systems, each of the logs comprising log entries associated with an event; obtaining a trained machine learning model using the logs for the data processing systems, the trained machine learning model comprising a plurality of embedding vectors; selecting, after obtaining the trained machine learning model, one of the embedding vectors as an output for an embedding generator, wherein the embedding generator is obtained by pruning the trained machine learning model; obtaining, using the logs and the embedding generator, vector representations for the logs; storing the vector representations in a vector indexing based database; obtaining a data processing system support request for a data processing system of the data processing systems, the support request including a log pattern, wherein the method is executed by a data processing system manager that is distinct and separate from the data processing system from which the data processing system support request is obtained; indexing, using the log pattern, the database to identify a related log of the logs; identifying the event associated with the related log; and performing an action set, based on the event, to service the data processing system support request, wherein performing the action set based on the event comprises: disabling, by the data processing system manager, one or more hardware components of the data processing system from which the data processing system support request is obtained. ### Claim 2 (depends on 1) The computer-implemented method of claim 1, wherein the data processing system support request is a request to obtain a log of the logs that is similar to a new log from which the log pattern was obtained, wherein the data processing system support request is obtained from the data processing system of the data processing systems, and wherein the data processing system is separate from the trained machine learning model. ### Claim 3 (depends on 2) The computer-implemented method of claim 2, wherein indexing the database comprises: obtaining a vector representation for the new log using the embedding generator; performing a nearest neighbor lookup in the database using the vector representation of the new log and the database to identify a vector of the vector representations in the database; identifying the related log of the logs based on the identity of the vector of the vector representations in the database; and returning a copy of the related log to a requestor to service the data processing system support request. ### Claim 4 (depends on 2) The computer-implemented method of claim 2, wherein indexing the database comprises: identifying clusters of the vector representations of the logs; obtaining a vector representation for the new log using the embedding generator; obtaining a graphical representation of the clusters and the vector representation for the new log; obtaining user input based on the graphical representation, the user input indicating one cluster of the clusters; and identifying the related log based on the one cluster. ### Claim 5 (depends on 1) The computer-implemented method of claim 1, wherein the logs are labeled, the trained machine learning model is trained using a supervised training method, the trained machine learning model is adapted to output an event associated with the log entries of the logs based on an input. ### Claim 6 (depends on 1) The computer-implemented method of claim 1, wherein the logs are unlabeled, the trained machine learning model is trained using an unsupervised training method, the trained machine learning model is adapted to output a log entry of the logs based on an input. ### Claim 7 (depends on 1) The computer-implemented method of claim 1, wherein the action set further comprises: transferring a workload from a first data processing system of the data processing systems to a second data processing system of the data processing systems. ### Claim 8 (depends on 1) The computer-implemented method of claim 1, wherein the action set further comprises: disabling a hardware component of a data processing system of the data processing systems. ### Claim 9 (depends on 1) The computer-implemented method of claim 1, wherein the action set comprises: sending a notification to an administrator of a data processing system of the data processing systems of a predicted impairment of the data processing system. ### Claim 10 (depends on 1) The computer-implemented method of claim 1, wherein pruning the trained machine learning model to obtain the embedding generator comprises disabling, within an embedding vector architecture of the trained machine learning model comprising the plurality of embedding vectors, all embedding vectors that are subsequent to the one embedding vector selected as the output for the embedding generator. ### Claim 11 A non-transitory machine-readable medium having instructions stored therein, which when executed by a processor, cause the processor to perform operations for managing data processing systems, the operations comprising: obtaining logs for the data processing systems, each of the logs comprising log entries associated with an event; obtaining a trained machine learning model using the logs for the data processing systems, the trained machine learning model comprising a plurality of embedding vectors; selecting, after obtaining the trained machine learning model, one of the embedding vectors as an output for an embedding generator, wherein the embedding generator is obtained by pruning the trained machine learning model; obtaining, using the logs and the embedding generator, vector representations for the logs; storing the vector representations in a vector indexing based database; obtaining a data processing system support request for a data processing system of the data processing systems, the support request including a log pattern, wherein the method is executed by a data processing system manager that is distinct and separate from the data processing system from which the data processing system support request is obtained; indexing, using the log pattern, the database to identify a related log of the logs; identifying the event associated with the related log; and performing an action set, based on the event, to service the data processing system support request, wherein performing the action set based on the event comprises: disabling, by a data processing system manager, one or more hardware components of the data processing system from which the data processing system support request is obtained. ### Claim 12 (depends on 11) The non-transitory machine-readable medium of claim 11, wherein the data processing system support request is a request to obtain a log of the logs that is similar to a new log from which the log pattern was obtained, wherein the data processing system support request is obtained from the data processing system of the data processing systems, and wherein the data processing system is separate from the trained machine learning model. ### Claim 13 (depends on 12) The non-transitory machine-readable medium of claim 12, wherein indexing the database comprises: obtaining a vector representation for the new log using the embedding generator; performing a nearest neighbor lookup in the database using the vector representation of the new log and the database to identify a vector of the vector representations in the database; identifying the related log of the logs based on the identity of the vector of the vector representations in the database; and returning a copy of the related log to a requestor to service the data processing system support request. ### Claim 14 (depends on 12) The non-transitory machine-readable medium of claim 12, wherein indexing the database comprises: identifying clusters of the vector representations of the logs; obtaining a vector representation for the new log using the embedding generator; obtaining a graphical representation of the clusters and the vector representation for the new log; obtaining user input based on the graphical representation, the user input indicating one cluster of the clusters; and identifying the related log based on the one cluster. ### Claim 15 (depends on 11) The non-transitory machine-readable medium of claim 11, wherein the logs are labeled, the trained machine learning model is trained using a supervised training method, the trained machine learning model is adapted to output an event associated with the log entries of the logs based on an input. ### Claim 16 (depends on 11) The non-transitory machine-readable medium of claim 11, wherein the logs are unlabeled, the trained machine learning model is trained using an unsupervised training method, the trained machine learning model is adapted to output a log entry of the logs based on an input. ### Claim 17 A data processing system manager, comprising: a processor; and a memory coupled to the processor to store instructions, which when executed by the processor, cause the processor to perform operations for managing data processing systems, the operations comprising: obtaining logs for the data processing systems, each of the logs comprising log entries associated with an event; obtaining a trained machine learning model using the logs for the data processing systems, the trained machine learning model comprising a plurality of embedding vectors; selecting, after obtaining the trained machine learning model, one of the embedding vectors as an output for an embedding generator, wherein the embedding generator is obtained by pruning the trained machine learning model; obtaining, using the logs and the embedding generator, vector representations for the logs; storing the vector representations in a vector indexing based database; obtaining a data processing system support request for a data processing system of the data processing systems, the support request including a log pattern, wherein the method is executed by a data processing system manager that is distinct and separate from the data processing system from which the data processing system support request is obtained; indexing, using the log pattern, the database to identify a related log of the logs; identifying the event associated with the related log; and performing an action set, based on the event, to service the data processing system support request, wherein performing the action set based on the event comprises: disabling, by a data processing system manager, one or more hardware components of the data processing system from which the data processing system support request is obtained. ### Claim 18 (depends on 17) The data processing system manager of claim 17, wherein the data processing system support request is a request to obtain a log of the logs that is similar to a new log from which the log pattern was obtained, wherein the data processing system support request is obtained from the data processing system of the data processing systems, and wherein the data processing system is separate from the trained machine learning model of the data processing system manager. ### Claim 19 (depends on 18) The data processing system manager of claim 18, wherein indexing the database comprises: obtaining a vector representation for the new log using the embedding generator; performing a nearest neighbor lookup in the database using the vector representation of the new log and the database to identify a vector of the vector representations in the database; identifying the related log of the logs based on the identity of the vector of the vector representations in the database; and returning a copy of the related log to a requestor to service the data processing system support request. ### Claim 20 (depends on 17) The data processing system manager of claim 17, wherein the logs are labeled, the trained machine learning model is trained using a supervised training method, the trained machine learning model is adapted to output an event associated with the log entries of the logs based on an input.