DETAILED ACTION
Claims 1, 4-8, 10-14, 17-22 are presented for examination.
This office action is in response to submission of application on 16-MARCH-2026.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 17-OCTOBER-2022 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Response to Amendment
The amendment filed 16-MARCH-2026 in response to the previous office action mailed 15-DECEMBER-2025 has been entered. Claims 1, 4-8, 10-14, 17-22 remain pending in the application.
With regards to the non-final office action’s rejections under 103, the amendments to the claims necessitated a new consideration of the art. After this consideration, the examiner respectfully disagrees with the applicant’s arguments that the art referenced in the previous office action does not teach the amendment claim limitations. A new 103 rejection over the prior art has been provided.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 4-8, 10-12, 14, and 17-19, 21-22 are rejected under 35 U.S.C. 103 as being unpatentable over Thibaux et al. (Pub. No. US 20080133288 A1, filed November 30th 2006, hereinafter Thibaux) in view of Childress et al. (Pub. No. US 20180174069 A1, filed December 15th 2016, hereinafter Childress) further in view of Liu et al. (Pub. No. US 20210365339 A1, filed May 19th 2020, hereinafter Liu) further in view of Tsou et al. (Pub. No. US 20180189667 A1, filed December 29th 2016, hereinafter Tsou).
Regarding claim 1:
Claim 1 recites:
A method of building an artificial intelligence (AI) model using big data associated with a plurality of servers, the method comprising: forming a matrix of data time series and statistic types, wherein each row of the matrix corresponds to a time series of a different server parameter of two or more server parameters and each column of the matrix corresponds to a different statistic type of two or more statistic types, wherein the two or more server parameters includes at least two of a field programmable gate array (FPGA) parameter, an air flow parameter, a CPU parameter, a memory parameter, and an interrupt parameter, wherein the two or more statistic types includes at least two of a first moving average of a first server parameter of the two or more server parameters, a first entire average of the first server parameter, a z-score of the first server parameter, a second moving average of standard deviation of the first server parameter, a second entire average of standard deviation of the first server parameter, and a spectral residual of the first server parameter; determining a first content of the matrix at a first time; determining a second content of the matrix at a second time; determining at least one leading indicator by processing at least the first content and the second content; building a plurality of decision trees based on the at least one leading indicator, wherein the building the plurality of decision trees comprises: starting with a single decision tree configured to predict hardware failure using a decision referring to one server parameter, and adding a second decision tree out of a plurality of decision trees by using an objective function to identify the second decision tree, wherein the addition of the second decision tree increases a reliability of the prediction and controls complexity of the AI model; outputting the plurality of decision trees as the AI model, wherein the AI model is used commonly for the plurality of servers; predicting, based on the AI model, failure of at least one of the plurality of servers; and reducing a load on a server for which failure is predicted.
Thibaux discloses determining a first content of the matrix at a first time; determining a second content of the matrix at a second time:
Thibaux teaches an intermediate matrix that can be time sliced to optimize finding correlation between failures and causes (Paragraph 42). The ability to gather information about the matrix at different times would include determine a first content of the matrix at a first time and the second content at a second time.
Thibaux discloses determining at least one leading indicator by processing at least the first content and the second content:
Thibaux teaches that implied causes of failure may be determined by the hypothesized grouping matrix from the intermediate matrix that comprises that first and second contents (Paragraph 52). The implied causes of failure would be the leading indicator.
Thibaux discloses forming a matrix of data time series and statistic types, wherein each row of the matrix corresponds to a time series of a different server parameter of two or more server parameters and [each column of the matrix corresponds to a different statistic type of two or more statistic types]
Thibaux teaches a component failure matrix that is a time series of component failures across the internet along with a hypothesized grouping matrix that for each of the component failures connects them with causes of failure or events (Paragraph 42). Therefore, the hypothesized grouping matrix would be a matrix of time series of different server parameters, wherein the server parameters are components. Therefore, Thibaux forming a matrix of time series and component failures, wherein functional status would be a type of server parameter.
However, Thibaux does not teach columns that correspond to a statistical type consisting of two or more of the recited list.
Thibaux does not disclose building an artificial intelligence (Al) model using big data. Rather, this limitation is disclosed by Childress:
Childress in the same field of endeavor of detection of anomalies in hardware components teaches the use of a random forest, which is a type of machine learning model (Paragraph 15).
Thibaux, Childress, and the present application are all in the same field of endeavor, detection of anomalies in hardware components.
Thibaux further does not disclose building a plurality of decision trees based on the at least one leading indicator; wherein the building the plurality of decision trees comprises: starting with a single decision tree configured to predict hardware failure using a decision referring to one server parameter, and adding a second decision tree out of a plurality of decision trees [by using an objective function to identify the second decision tree], wherein the addition of the second decision tree increases a reliability of the prediction and controls complexity of the AI model; outputting the plurality of decision trees as the Al model. This limitation is disclosed by Childress.
Childress teaches building a random forest model, which is a plurality of decision trees, based on a plurality of failures of individual hardware pieces (Paragraph 33). The plurality of failures of individual hardware pieces would be the leading indicators and as the decision trees are a random forest model, they are the AI model. These may be output, as the methodology of Childress may be included in a stand-alone software package including the random forest model (Paragraph 52). This would also demonstrate starting with a single decision tree configured to predict hardware failure using a decision referring to one server parameter, as each of the failures would represent a classification of that hardware as failing i.e. a decision, and the random forest by necessity starts its process of creating decision trees with a single decision tree configured to predict hardware failure.
Furthermore, Childress teaches using more than one decision tree (i.e., adding a second decision tree) as seen in the random forest improves classification, or reliability of prediction (Childress, Paragraph 15). Furthermore, adding a second decision tree would by nature control the complexity of an AI model, as further expanding the model would increase the complexity of the random forest.
Childress does not disclose using an objective function to identify the second decision tree. This limitation is taught further below by Tsou.
However, Thibaux does not disclose wherein the two or more server parameters includes at least two of a field programmable gate array (FPGA) parameter, an air flow parameter, a CPU parameter, a memory parameter, and an interrupt parameter. This limitation is disclosed below by Childress:
Additionally, Childress teaches that its training data includes CPU information and memory information (Paragraph 34), which would be two or more server parameters includes at least two of a field programmable gate array (FPGA) parameter, an air flow parameter, a CPU parameter, a memory parameter, and an interrupt parameter
Furthermore, Thibaux nor Childress disclose wherein the two or more statistic types includes at least two of a first moving average of a first server parameter of the two or more server parameters, a first entire average of the first server parameter, a z-score of the first server parameter, a second moving average of standard deviation of the first server parameter, a second entire average of standard deviation of the first server parameter, and a spectral residual of the first server parameter. This limitation is disclosed below by Liu:
Liu in same field of endeavor of machine learning models teaches that statistics for multiple server parameters (in Liu, virtual machines) may be computed (Paragraph 107) and that this statistic may consist of a moving average by way of using ARIMA (Autoregressive Integrated Moving Average, wherein the moving average is used) (Paragraph 108). As within Liu the moving average of both the first server parameter and second server parameter would be calculated, it teaches both the first moving average and the second moving average from the recited list, which is at least two of a first moving average of a first server parameter of the two or more server parameters, a first entire average of the first server parameter, a z-score of the first server parameter, a second moving average of standard deviation of the first server parameter, a second entire average of standard deviation of the first server parameter, and a spectral residual of the first server parameter.
Therefore using the statistics of Liu and the statistics of Childress with the matrix of Thibaux would produce a matrix in accordance with the above limitation.
Thibaux, Childress, Liu, and the present application are analogous art because they are all in the same field of endeavor of machine learning.
Neither Thibaux nor Childress disclose wherein the AI model is used commonly for the plurality of servers; predicting, based on the AI model, failure of at least one of the plurality of servers; and reducing a load on a server for which failure is predicted. Instead, this limitation is disclosed by Liu:
Liu teaches a prediction engine that is configured to predict failure at a replica site, or server (Paragraph 102) which would be an AI model used commonly for the plurality of servers predicting failure of at least one of the plurality of servers. Furthermore, Liu teaches the use of load balancing, which may involve a virtual machine being off loaded from one server to a second server such that there is no down time for users of the virtual machine (Paragraph 94).
None of Thibaux, Childress, or Liu disclose using an objective function to identify the second decision tree. This limitation is disclosed below by Tsou:
Tsou in the same field of endeavor of machine learning teaches an optimization function which identifies the new decision tree by identifying its weights (Tsou, Paragraph 125). This would be using an objective function to identify the second decision tree, since an optimization function is a type of objective function.
Tsou and the present application are analogous art because they are in the same field of endeavor.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Thibaux, the teachings of Childress, the teachings of Liu, and the teachings of Tsou. This would have granted the advantage of increasing reliability of hardware components (Childress, Paragraph 4) as well as data protection in the case of system failure (Liu, Paragraph 2) as well as improved accuracy and robustness (Tsou, Paragraph 51).
Regarding claim 4, which depends upon claim 1:
Claim 4 recites:
The method of claim 3, wherein the FPGA parameter is message queue, the CPU parameter is load and/or processes, the memory parameter is IRQ or DISKIO, and the interrupt parameter is IPMI and/or IOWAIT.
Thibaux in view of Childress, further in view of Liu, further in view of Tsou discloses the method of claim 1 upon which claim 4 depends. However, Thibaux does not teach the limitation of claim 4:
Childress teaches that its CPU parameters is regarding CPU usage, which is the loads and/or processes of the CPU (Paragraph 22). Therefore the CPU parameter of claim 3 would load and/or processes.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Thibaux, the teachings of Childress, the teachings of Liu, and the teachings of Tsou. This would have granted the advantage of increasing reliability of hardware components (Childress, Paragraph 4) as well as data protection in the case of system failure (Liu, Paragraph 2) as well as improved accuracy and robustness (Tsou, Paragraph 51).
Regarding claim 5, which depends upon claim 1:
Claim 5 recites:
The method of claim 1, wherein each decision tree of the plurality of decision trees includes a plurality of decision nodes, a corresponding plurality of decision thresholds are associated with the plurality of decision nodes, and the building the plurality of decision trees comprises choosing the plurality of decision thresholds to detect anomaly patterns of the at least one leading indicator over a first time interval.
Thibaux in view of Childress, further in view of Liu, further in view of Tsou discloses the method of claim 1 upon which claim 5 depends. However, Thibaux does not teach the limitation of claim 5 that wherein each decision tree of the plurality of decision trees includes a plurality of decision nodes, a corresponding plurality of decision thresholds are associated with the plurality of decision nodes:
Childress teaches its decision trees have leaves, wherein item observations are mapped to conclusions about the item’s target value (Paragraph 15). Leaves, in the context of decision trees, refer to nodes. Furthermore, the mapping of observations to conclusions to continue down the tree would be analogous to decision thresholds associated with the nodes.
Furthermore, Thibaux does not teach the building the plurality of decision trees comprises choosing the plurality of decision thresholds to detect anomaly patterns of the at least one leading indicator over a first time interval:
Childress teaches detecting outliers through low proximity between the output of a hardware device and a random forest model which would indicate an anomalous pattern (Paragraph 24). The output of a hardware device would then be the leading indicator.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Thibaux, the teachings of Childress, the teachings of Liu, and the teachings of Tsou. This would have granted the advantage of increasing reliability of hardware components (Childress, Paragraph 4) as well as data protection in the case of system failure (Liu, Paragraph 2) as well as improved accuracy and robustness (Tsou, Paragraph 51).
Regarding claim 6, which depends upon claim 5:
Claim 6 recites:
The method of claim 5, wherein the big data comprises a plurality of server diagnostic files associated with a first server of a plurality of servers, a dimension of the plurality of server diagnostic files indicating that there is a first number of files in the plurality of server diagnostic files, and the first number is more than 1,000.
Thibaux in view of Childress, further in view of Liu, further in view of Tsou discloses the method of claim 5 upon which claim 6 depends. Furthermore, regarding the limitation of claim 6:
Thibaux teaches that thousands of health indications are received from systems over the Internet (Paragraph 99). Health indications would be server diagnostic files associated with the first servers.
Regarding claim 7, which depends upon claim 6:
Claim 7 recites:
The method of claim 6, wherein the first time interval is about one month.
Thibaux in view of Childress, further in view of Liu, further in view of Tsou discloses the method of claim 6 upon which claim 7 depends. However, Thibaux in view of Childress does not teach the limitation of claim 7:
Liu teaches that seasonality may be the model’s approximated best seasonality (Paragraph 108). While Liu does not explicitly mention monthly as a time interval, Liu does list ‘weekly’ as an example of an appropriate time interval, demonstrating the time intervals of Liu are able to be significantly longer than e.g., minutes. Therefore it would be obvious that the seasonality of Liu may include monthly in order to allow for Liu to adapt to an anticipated monthly seasonality (Paragraph 108).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Thibaux, the teachings of Childress, the teachings of Liu, and the teachings of Tsou. This would have granted the advantage of increasing reliability of hardware components (Childress, Paragraph 4) as well as data protection in the case of system failure (Liu, Paragraph 2) as well as improved accuracy and robustness (Tsou, Paragraph 51).
Regarding claim 8, which depends upon claim 7:
Claim 8 recites:
The method of claim 7, wherein a most recent version of a first file of the plurality of server diagnostic files associated with the first server is obtained about every 10 minutes.
Thibaux in view of Childress, further in view of Liu, further in view of Tsou discloses the method of claim 7 upon which claim 8 depends. Furthermore, regarding the limitation of claim 8:
Thibaux teaches that a data point is received regarding a single component’s state during a one-minute time interval (Paragraph 49). This data point would be the first file of server diagnostic files and gathering it every minute would also mean it is gathered at every ten minute interlude.
Regarding claim 10, which depends upon claim 8:
Claim 10 recites:
The method of claim 9, wherein the plurality of decision trees are configured to process the second number of copies of the first file to make a prediction of hardware failure related to the first server.
Thibaux in view of Childress, further in view of Liu, further in view of Tsou discloses the method of claim 8 upon which claim 10 depends. However, Thibaux does not teach the limitation of claim 10:
Childress teaches using outlier information to predict possible failure of a device (Paragraph 23). The outlier information would be the second number of copies of the first file which represent the status of the server parameters over time.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Thibaux, the teachings of Childress, the teachings of Liu, and the teachings of Tsou. This would have granted the advantage of increasing reliability of hardware components (Childress, Paragraph 4) as well as data protection in the case of system failure (Liu, Paragraph 2) as well as improved accuracy and robustness (Tsou, Paragraph 51).
Regarding claim 11, which depends upon claim 10:
Claim 11 recites:
The method of claim 10, wherein a second dimension of the plurality of servers indicating that there is a second number of servers in the plurality of servers, and the second number is greater than 1,000 servers.
Thibaux in view of Childress, further in view of Liu, further in view of Tsou discloses the method of claim 10 upon which claim 11 depends. Furthermore, regarding the limitation of claim 11:
Thibaux teaches that the number of its autonomous systems, which may be servers, is 10,000-20,000, which is greater than 1000 (Paragraph 39). This may also be the second number of servers as it includes all the autonomous systems of Thibaux.
Regarding claim 12, which depends upon claim 11:
Claim 12 recites:
The method of claim 11, wherein the plurality of decision trees are configured to implement a light-weight process, and the plurality of decision trees are configured to output a health score for each server of the plurality of servers, the plurality of decision trees being scalable with respect to a second number of servers, wherein scalable includes an exponential increase in the second number of servers causing at most a linear increase in a complexity of the plurality of decision trees.
Thibaux in view of Childress, further in view of Liu, further in view of Tsou discloses the method of claim 11 upon which claim 12 depends.
Neither Thibaux nor Childress disclose wherein the plurality of decision trees are configured to implement a light-weight process. Instead, this limitation is disclosed by Liu:
Liu teaches the use of threads of execution, which are synonymous with light-weight processes (Paragraph 162).
Thibaux does not disclose and the plurality of decision trees are configured to output a health score for each server of the plurality of servers. Instead, this limitation is disclosed by Childress:
Childress teaches the goal of its processes, including the creation of the random forest model, is to predict which device or devices are prone to failure (Paragraph 20). Predicting devices that are prone to failure is a prediction of the health of a system, and hence a health score.
Thibaux does not disclose the plurality of decision trees being scalable with respect to a second number of servers, wherein scalable includes an exponential increase in the second number of servers causing at most a linear increase in a complexity of the plurality of decision trees. Instead, this limitation is disclosed by Childress:
Childress teaches uses decision trees to map information about at item to conclusion about the item (Paragraph 15). Therefore, the number of decision trees will scale with the number of items, which may be servers. Furthermore, this would be in a 1:1 ratio of decision trees to server, which would be a linear increase in complexity of the decision trees.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Thibaux, the teachings of Childress, the teachings of Liu, and the teachings of Tsou. This would have granted the advantage of increasing reliability of hardware components (Childress, Paragraph 4) as well as data protection in the case of system failure (Liu, Paragraph 2) as well as improved accuracy and robustness (Tsou, Paragraph 51).
Claims 14, 17-19 recite an apparatus that parallels the method of claims 1, 4-6 respectively. Furthermore, Childress teaches the processors, memories, and storage for the computer program (Paragraph 37). Therefore, the analysis discussed above with respect to claims 1, 4-6 also applies to claims 14, 17-19 respectively. Accordingly, claims 14, 17-19 are rejected based on substantially the same rationale as set forth above with respect to claims 1, 4-6 respectively.
Regarding claim 21, which depends upon claim 1:
Claim 21 recites:
The method of claim 1, wherein the predicting comprises predicting, based on the AI model, failure of a first server; and off-loading a first virtual machine from the first server to a second server, wherein the plurality of servers comprises the first server and the second server, a first user is supported by the first virtual machine, a user down time for the first user is avoided as the first virtual machine has been moved from the first server before a failure of the first server, thereby improving a system availability to the first user
Thibaux in view of Childress, further in view of Liu, further in view of Tsou disclose the method of claim 1 upon which claim 21 depends. However, Thibaux in view of Childress do not teach the limitations of claim 21.
Liu teaches a prediction engine that is configured to predict failure at a replica site, or server (Paragraph 102) which would be an AI model used commonly for the plurality of servers predicting failure of at least one of the plurality of servers. Furthermore, Liu teaches the use of load balancing, which may involve a virtual machine being off loaded from one server to a second server such that there is no down time for users of the virtual machine (Paragraph 94).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Thibaux, the teachings of Childress, the teachings of Liu, and the teachings of Tsou. This would have granted the advantage of increasing reliability of hardware components (Childress, Paragraph 4) as well as data protection in the case of system failure (Liu, Paragraph 2) as well as improved accuracy and robustness (Tsou, Paragraph 51).
Regarding claim 22, which depends upon claim 1:
Claim 22 recites:
The method of claim 1, wherein the building a plurality of decision trees comprises: starting with a single decision tree configured to predict hardware failure using a decision referring to one server parameter; adding a second decision tree out of a plurality of decision trees by using an objective function to identify the second decision tree, wherein the addition of the second decision tree increases a reliability of the prediction while an increase in complexity is controlled by a regularization term in the objective function
Thibaux in view of Childress, further in view of Liu, further in view of Tsou discloses claim 1 upon which claim 22 depends. Furthermore, Childress discloses starting with a single decision tree configured to predict hardware failure using a decision referring to one server parameter […] wherein the addition of the second decision tree increases a reliability of the prediction:
Childress teaches that any individual decision tree (i.e., a single decision tree) to identify observations about an item, for example a server parameter, to conclusion about the items (Paragraph 15), such as hardware failure prediction (Paragraph 4). Furthermore, use of a random forest adds at least one additional free, the addition of a second decision tree, in order to improve the classification rate, which would be the reliability of the prediction (Paragraph 15).
Furthermore, Thibaux teaches wherein […] an increase in complexity is controlled by a regularization term in the objective function:
Thibaux teaches a complexity evaluator that uses a cost function in order to evaluate (Paragraph 70) and limit (Paragraph 73) complexity. The cost function would be another term for the objective function.
Furthermore, Tsou teaches adding a second decision tree out of a plurality of decision trees by using an objective function to identify the second decision tree:
Tsou teaches that decision tree weighting is used to determine whether a new tree should be added (Paragraph 60), wherein the decision tree weights is determined by an objective function (Paragraph 125). Therefore, the objective function is used to identify and add a second decision tree out of a plurality of decision trees.
Tsou and the present application are analogous art because they are in the same field of endeavor.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Thibaux, the teachings of Childress, the teachings of Liu, and the teachings of Tsou. This would have granted the advantage of increasing reliability of hardware components (Childress, Paragraph 4) as well as data protection in the case of system failure (Liu, Paragraph 2) as well as improved accuracy and robustness (Tsou, Paragraph 51).
Claim 13 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Thibaux in view of Childress, further in view of Liu, further in view of Tsou, further in view of Biem et al. (Pub. No. US 20160125315 A1, filed June 18th 2015, hereinafter Biem).
Regarding claim 13:
Claim 13 recites:
A model builder computer comprising: one or more processors; and one or more memories, the one or more memories storing a computer program, the computer program including: interface code configured to obtain server log data associated with a plurality of servers, and calculation code configured to: form a matrix of data time series and statistic types, wherein each row of the matrix corresponds to a time series of a different server parameter of a plurality of server parameters and each column of the matrix corresponds to a different statistic type of a plurality of statistic types, wherein a number of the plurality of server parameters is on the order of one hundred, wherein a number of the different statistic types is on an order of five, and a rate of the plurality of server parameters is on the order of one million per minute; determine a first content of the matrix at a first time; determine a second content of the matrix at a second time; and determine at least one leading indicator by processing at least the first content and the second content, and build a plurality of decision trees based on the at least one leading indicator, wherein the interface code is further configured to send the plurality of decision trees, as a trained Artificial Intelligence (AI) model, to a computer, and the trained AI model becoming a component of an AI inference engine, wherein the AI model is used commonly for the plurality of servers by predicting, based on the AI model, failure of at least one of the plurality of servers, thus reducing a load on a server for which failure is predicted, wherein the building of the decision trees by the calculation code comprises: starting with a single decision tree configured to predict hardware failure using a decision referring to one server parameter, and adding a second decision tree out of a plurality of decision trees by using an objective function to identify the second decision tree, wherein the addition of the second decision tree increases a reliability of the prediction and controls complexity of the AI model.
Thibaux discloses interface code configured to obtain server log data associated with a plurality of servers, and calculation code configured to: form a matrix of data time series and statistic types, wherein each row of the matrix corresponds to a time series of a different server parameter of a plurality of server parameters and [each column of the matrix corresponds to a different statistic type of a plurality of statistic types, wherein a number of the plurality of server parameters is on the order of one hundred, wherein a number of the different statistic types is on an order of five, and a rate of the plurality of server parameters is on the order of one million per minute] and determine at least one leading indicator by processing at least the first content and the second content:
Thibaux teaches receiving health indications from autonomous systems across the Internet (Paragraph 99) which is obtaining server log data associated with a plurality of servers.
Thibaux teaches a component failure matrix that is a time series of component failures across the internet along with a hypothesized grouping matrix that for each of the component failures connects them with causes of failure or events (Paragraph 42). Therefore, the hypothesized grouping matrix would be a matrix of time series of different server parameters, wherein the server parameters are components. Therefore, Thibaux forming a matrix of time series and component failures, wherein functional status would be a type of server parameter.
Furthermore, Thibaux teaches that implied causes of failure may be determined by the hypothesized grouping matrix from the intermediate matrix that comprises that first and second contents (Paragraph 52). The implied causes of failure would be the leading indicator which are determined from the intermediate matrix, i.e. the first and second contents.
Thibaux does not teach each column of the matrix corresponds to a different statistic type of a plurality of statistic types or wherein a number of the plurality of server parameters is on the order of one hundred, wherein a number of the different statistic types is on an order of five, and a rate of the plurality of server parameters is on the order of one million per minute. These limitations are taught by Childress and Biem respectively below.
Thibaux discloses determining a first content of the matrix at a first time; determining a second content of the matrix at a second time:
Thibaux teaches an intermediate matrix that can be time sliced to optimize finding correlation between failures and causes (Paragraph 42). The ability to gather information about the matrix at different times would include determine a first content of the matrix at a first time and the second content at a second time.
However, Thibaux does not disclose build a plurality of decision trees based on the at least one leading indicator. Rather, Childress discloses this limitation:
Childress teaches detecting outliers through low proximity between the output of a hardware device and a random forest model which would indicate an anomalous pattern (Paragraph 24). The output of a hardware device would then be the leading indicator. Furthermore, Childress teaches the building of decision trees in the form of a random forest based on training data including a plurality of failures of hardware devices (Paragraph 4), which would be an example of this leading indicator.
However, Thibaux does not disclose one or more processors; and one or more memories, the one or more memories storing a computer program. Rather, this limitation is disclosed by Childress:
Childress teaches a processor, memory, and storage to run the computer program (Paragraph 37).
Thibaux does not disclose wherein the interface code is further configured to send the plurality of decision trees, as a trained Artificial Intelligence (AI) model, to a computer, and the trained Al model becoming a component of an Al inference engine. Instead, this limitation is disclosed by Childress:
Childress teaches that for its computer system including the trained random forest which is a plurality of decision trees as a trained AI model may execute entirely on a user’s computer or partly on a remote computer, which indicated the trained model being sent to the user’s computer. Furthermore, it may be included as a stand-alone software package, wherein it becomes a component of an AI inference engine. (Paragraph 52).
Thibaux does not disclose wherein the building of the decision trees by the calculation code comprises: starting with a single decision tree configured to predict hardware failure using a decision referring to one server parameter, and adding a second decision tree out of a plurality of decision trees [by using an objective function to identify the second decision tree], wherein the addition of the second decision tree increases a reliability of the prediction and controls complexity of the AI model. Instead, Childress discloses:
Childress teaches building a random forest model, which is a plurality of decision trees, based on a plurality of failures of individual hardware pieces (Paragraph 33). The plurality of failures of individual hardware pieces would be the leading indicators and as the decision trees are a random forest model, they are the AI model. These may be output, as the methodology of Childress may be included in a stand-alone software package including the random forest model (Paragraph 52). This would also demonstrate starting with a single decision tree configured to predict hardware failure using a decision referring to one server parameter, as each of the failures would represent a classification of that hardware as failing i.e. a decision, and the random forest by necessity starts its process of creating decision trees with a single decision tree configured to predict hardware failure.
Furthermore, Childress teaches using more than one decision tree (i.e., adding a second decision tree) as seen in the random forest improves classification, or reliability of prediction (Childress, Paragraph 15). Furthermore, adding a second decision tree would by nature control the complexity of an AI model, as further expanding the model would increase the complexity of the random forest.
Childress does not disclose using an objective function to identify the second decision tree. This limitation is taught further below by Tsou.
However, Thibaux does not disclose each column of the matrix corresponds to a different statistic type of a plurality of statistic types. Instead, this is disclosed by Childress:
Additionally, Childress teaches that its training data includes CPU information and memory information (Paragraph 34), which would be different statistic types of a plurality of statistic types.
Therefore the statistics of Childress with the previously described matrix of Thibaux would produce a matrix in accordance with the above limitation.
Neither Thibaux nor Childress disclose wherein the AI model is used commonly for the plurality of servers; predicting, based on the AI model, failure of at least one of the plurality of servers; and reducing a load on a server for which failure is predicted. Instead, this limitation is disclosed by Liu:
Liu teaches a prediction engine that is configured to predict failure at a replica site, or server (Paragraph 102) which would be an AI model used commonly for the plurality of servers predicting failure of at least one of the plurality of servers. Furthermore, Liu teaches the use of load balancing, which may involve a virtual machine being off loaded from one server to a second server such that there is no down time for users of the virtual machine (Paragraph 94).
None of Thibaux, Childress, or Liu disclose wherein a number of the plurality of server parameters is on the order of one hundred, wherein a number of the different statistic types is on an order of five, and a rate of the plurality of server parameters is on the order of one million per minute. Instead, this limitation is disclosed by Biem:
Biem in the same field of inference models teaches a data stream with millions of events or message per second, wherein the events would be, for example, server parameters (Paragraph 17), wherein this would also be millions per minute. Furthermore, Biem teaches treating data streaming from thousands of real-time sources (Paragraph 17) which would be include an order of one hundred as the sources would be analogous to the plurality of server parameters as they are the source of server parameters data. Finally, Biem shows that three coefficients may be calculated for the sources, in addition to a key and a status (Figure 7A). This would consist of five statistic types as each is a different point of data about the source.
Biem and the present application are analogous art because they are in the same field of endeavor of inference models.
None of Thibaux, Childress, Liu, or Biem disclose using an objective function to identify the second decision tree. This limitation is disclosed below by Tsou:
Tsou in the same field of endeavor of machine learning teaches an optimization function which identifies the new decision tree by identifying its weights (Tsou, Paragraph 125). This would be using an objective function to identify the second decision tree, since an optimization function is a type of objective function.
Tsou and the present application are analogous art because they are in the same field of endeavor.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the present application to implement a method that utilized the teachings of Thibaux, the teachings of Childress, the teachings of Liu, and the teachings of Tsou, and the teachings of Biem. This would have granted the advantage of increasing reliability of hardware components (Childress, Paragraph 4) as well as data protection in the case of system failure (Liu, Paragraph 2) as well as improved accuracy and robustness (Tsou, Paragraph 51) as well as processing big data effectively (Biem, Paragraph 2).
Claim 20 recites a non-transitory computer readable storage medium that parallels the method of claim 13. Therefore, the analysis discussed above with respect to claim 13 also applies to claim 20. Accordingly, claim 20 is rejected based on substantially the same rationale as set forth above with respect to claim 13.
Response to Arguments
Applicant’s arguments filed 16-MARCH-2026 have been fully considered, but the examiner believes that not all are fully persuasive.
Regarding the applicant’s remarks on the non-final office action’s 103 rejection of the claims, the applicant argues that the previously presented art does not teach the amended limitations of these claims. As such, the applicant argues that all claims dependent on the above would additionally not be obvious under 103. However, the examiner believes that the previously presented art does teach the amended limitations and respectfully requests applicant’s consideration of the following:
The applicant argues that none of Thibaux, Childress, or Liu disclose “the building the plurality of decision trees comprises: starting with a single decision tree configured to predict hardware failure using a decision referring to one server parameter, and adding a second decision tree out of a plurality of decision trees by using an objective function to identify the second decision tree” and the “addition of the second decision tree increases a reliability of the prediction and controls complexity of the AI model”. The examiner believes that Childress does address some of these limitations, but has further incorporated previous art Tsou to address its deficiencies.
With regards to Childress, the applicant states that nothing within the art discloses “that the addition of a second decision tree ‘increases a reliability of the prediction’ and ‘controls complexity of the AI model.’” The examiner believes that the random forest of Childress demonstrates these improvements, as using more than one decision tree as seen in the random forest is stated to improve classification (Childress, Paragraph 15). Furthermore, adding a second decision tree would by nature control the complexity of an AI model, as further expanding the model would increase the complexity of the random forest.
Tsou is incorporated in order to address the using an objective function to identify the second decision tree, wherein the optimization function identifies the new decision tree by identifying its weights (Tsou, Paragraph 125). While this alone does not fully teach the amended limitations, when combined with the teachings of Childress the examiner believes that the prior art discloses these limitations.
Any limitation of the amendment not explicitly addressed within the arguments is disclosed above in the prior art rejection. For these reasons, the examiner believes that the amendment to claim 1 and the analogous limitations are disclosed.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ALEXANDRIA JOSEPHINE MILLER whose telephone number is (703)756-5684. The examiner can normally be reached Monday-Thursday: 7:30 - 5:00 pm, every other Friday 7:30 - 4:00.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Mariela Reyes can be reached on (571) 270-1006. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/A.J.M./Examiner, Art Unit 2142
/Mariela Reyes/ Supervisory Patent Examiner, Art Unit 2142
Prosecution Insights