Notice of Pre-AIA or AIA Status
1. The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
DETAILED ACTION
2. This Office Action is sent in response to Applicant’s Communication received on 7/22/2026 for application number 18/387,230.
Response to Amendments
3. The Amendment filed 07/22/2026 has been entered. Claims 1-3, 5, 10, 11 have been amended. Claims 6 and 8 have been canceled. Claims 12-22 have been added. Claims 1-5, 7, and 9-22 remain pending in the application.
4. Applicant’s amendments to claims 1, 8, 10, and 11 have been fully considered and are persuasive. The objections to these claims are respectfully withdrawn.
5. Applicant’s amendment to claim 10 has been fully considered and is persuasive. The amendment provided to overcome the §112(f) set forth in the previous office action is sufficient. Accordingly, the 35 U.S.C § 112(f) of claim 10 is respectfully withdrawn.
6. Applicant’s amendments to claims 1, 10, and 11 have been fully considered and are persuasive. The amendments provided to overcome the §101 rejection set forth in the previous office action is sufficient. Accordingly, the 35 U.S.C § 101 rejection of claims 1-11 is respectfully withdrawn.
Response to Arguments
Applicant argues that cited references fail to disclose or suggest all the features of the amended independent claims 1, 10, and 11. However, the argument is moot since this is a newly presented limitation, thus changes the scope of the claims. However, newly found references are applied.
Claim Rejections – 35 USC § 103
7. 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 of this title, 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.
8. Claims 1, 5, 7, 9-12, and 15-17 are rejected under 35 U.S.C. 103 as being unpatentable over Wright et al. (U.S. Patent Application Pub. No. US 20210320936 A1) in view of Terra et al. (U.S. Patent Application Pub. No. US 20250005450 A1), and further in view of Ping (U.S. Patent Application Pub. No. US 20180260007 A1).
Claim 1: Wright teaches a fault prediction method implemented by one or more processors (i.e. Processing element 410 may be, one or multiple central processing unit (CPU); para. [0061-0063]) of a baseboard management controller (BMC) (i.e. A management controller such as a baseboard management controller (BMC) can be used to perform the analysis on incoming health information according to the deep learning model; para. [0020]), the method comprising:
collecting, by the one or more processors, monitoring information regarding computing modules installed on a main board (i.e. A management controller 116 in a production or test environment can communicate with components 110 of the computing device 102 to generate health information 214. The management controller 116 can use one or more bus to communicate with the components to receive health information. Examples of health information 214 can include temperature associated with a chip, such as one or more a central processing unit, memory, etc., log information, session counters, user action information, network packet counters, power consumption, error information, a record of firmware or software installed on the computing device, processing workload information, peripheral device configuration information, peripheral device sensor information, etc; para. [0024, 0064]);
calculating, by the one or more processors, a fault outbreak feedback level (FOFL) from the collected monitoring information (i.e. a Deep Neural Network (DNN) makes immediate predictions as it consumes input vectors. A fully connected architecture can be employed with the final stage being the output that predicts whether an anomaly has occurred … the scores can be used to rank the probability that each of the event or a group of events represents an anomaly of normal operation of the computing device 102; para. [0030, 0031, 0033]);
selecting, by an on-device Al prediction processor of the one or more processors, one artificial intelligence (AI) model from among a plurality of Al models (i.e. Various deep learning models can be used. Examples of deep learning models include long short-term memory (LSTM), a convolution neural networks, recurrent neural networks, neural history compressor, recursive neural networks, gated recurrent unit (GRU), etc; para. [0027]);
constructing the one Al model (i.e. The health data can be provided to an analysis platform that can generate the model. In some examples, the training using benchmark workloads can be considered pre-training the deep learning model for the specific workload. That deep learning model can further be updated based on additional feedback from computing devices; para. [0026]) using Al model data received from an external Al model platform based on the selecting Al model (i.e. the management controller 400 can receive updated deep learning models from the analysis platform. As noted above, this can be updated based on health information parameter feedback received from multiple computing devices. Other computing devices can act upon updated models based on health information from the management controller (e.g., by using the updated model on new health information to detect another anomaly); para. [0038, 0045, 0074]), and predicting a FOFL from the monitoring information using the constructed Al model (i.e. When the deep learning model 112 is used in production mode, the management controller 116 can process its respective health information 214 according to the deep learning model 112 to determine whether an anomaly has occurred based on the deep learning model 112; para. [0032]); and
controlling the computing modules of the main board based on the predicted FOFL (i.e. the management controller that detected the anomaly or the associated management controllers may perform an action on their computing devices, such as isolating network traffic, quarantining, updating of a software or firmware component, communicating with other management controllers of other computing devices, preemptive actions, and the like; para. [0040, 0053, 0072]).
Wright does not explicitly teach selecting, one artificial intelligence (AI) model from among a plurality of Al models corresponding to respective FOFL levels based on the calculated FOFL; wherein the on-device Al prediction processor is implemented in a secondary service processor (SSP) that is distinguished from a primary service processor (PSP) of the BMC, and the on-device Al prediction processor runs the selected Al model using a lightweight inference engine.
However, Terra teaches selecting, by an on-device Al prediction processor of the one or more processors, one artificial intelligence (AI) model from among a plurality of Al models corresponding to respective FOFL levels based on the calculated FOFL (i.e. The risk value determined for the state of the environment, based on the safety information, is then used to dynamically select one of a plurality of ML models available to the ML agent … FIG. 4 shows three different alternative options which may be selected depending upon which of model_1, model_2 and model_n is selected. In the example shown in FIG. 4 , model_1 is selected in a high risk scenario, model_2 is selected in a medium risk scenario and model_n is selected in a low risk scenario; para. [0025-0029, 0032, 0041]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the invention of Wright to include the feature of Terra. One would have been motivated to make this modification because it reduces processing resource and power consumption when a lower risk condition permits use of a less complex model while retaining the ability to employ a more complex model when a higher risk condition is detected.
However, Ping teaches wherein the on-device Al prediction processor is implemented in a secondary service processor (SSP) that is distinguished from a primary service processor (PSP) of the BMC (i.e. Referring to FIG. 4, in some embodiments, a neural network (or “artificial neural network”) is used to estimate the heat load in the SSD. The neural network may be implemented as a suitable set of program instructions executed by the SSD controller 210 or by the baseboard management controller 140, or it may be implemented as an analog neural network configured, with suitable digital to analog and analog to digital converters, to operate as an (on-chip or off-chip) analog co-processor of the SSD controller 210 or of the baseboard management controller 140; para. [0048, 0057]), and the on-device Al prediction processor runs the selected Al model using a lightweight inference engine (i.e. the output of the first node of the output layer is an estimated heat load, and an output of the second node of the output layer is an estimated effective temperature; para. [0048, 0049]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Wright and Terra to include the feature of Ping. One would have been motivated to make this modification because it provides a dedicated hardware resource for performing the neural network computations rather than burdening the principal BMC processing circuitry.
Claim 2: Wright, Terra, and Ping teach the fault monitoring method of claim 1. Wright further teaches wherein the collecting and the calculating are periodically performed (i.e. some or all of the health information 214 can be provided to the analysis platform 250 to update the deep learning model. This can be performed periodically, for example, health information 214 collected for a period (e.g., at a low workflow part of each day). The analysis platform 250 can process the health information 214 as a matter of course and send updated deep learning models 112 to the computing devices to use; para. [0045, 0047]).
Claim 5: Wright, Terra, and Ping teach the fault monitoring method of claim 1. Wright further teaches storing the predicted FOFL in a database (DB) as log data (i.e. the health information 214 can be stored as logs or other data structures in a persistent memory or volatile memory; para. [0025]).
Claim 7: Wright, Terra, and Ping teach the fault monitoring method of claim 1. Wright further teaches wherein the AI model is trained by an external platform (i.e. The health data can be provided to an analysis platform that can generate the model. In some examples, the training using benchmark workloads can be considered pre-training the deep learning model for the specific workload. That deep learning model can further be updated based on additional feedback from computing devices; para. [0026]).
Claim 9: Wright, Terra, and Ping teach the fault monitoring method of claim 1. Wright further teaches (i.e. The action can be a reporting action such as notifying an administrator of the issue; para. [0018]).
Claims 10-12 and 15-17 are similar in scope to Claims 1, 2, 5, 7, 9 and are rejected under a similar rationale.
9. Claims 3 and 13 are rejected under 35 U.S.C. 103 as being unpatentable over Wright in view of Terra, Ping, and further in view of Shigehara (U.S. Patent Application Pub. No. US 20130061098 A1).
Claim 3: Wright, Terra, and Ping teach the fault monitoring method of claim 2. Wright does not explicitly teach wherein the predicting is performed only when the FOFL is a defined level.
However, Shigehara teaches wherein the predicting is performed only when the FOFL is a defined level (i.e. if the predicted process load is higher than a first threshold (the predetermined level), the failure check is not performed, if the predicted process load is lower than the first threshold and higher than a second threshold (second predetermined level), a part of the failure check is performed, and if the predicted process load is lower than the second threshold, the entire failure check (with respect to the single CPU core) is performed; para. [0115]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Wright, Terra, and Ping to include the feature of Shigehara. One would have been motivated to make this modification because it provides a way to make the fault-prediction workflow more efficient.
Claim 13 is similar in scope to Claim 3 and is rejected under a similar rationale.
10. Claims 4 and 14 rejected under 35 U.S.C. 103 as being unpatentable over Wright in view of Terra, Ping, and further in view of Kudva et al. (U.S. Patent Pub. No. US 8122158 B1).
Claim 4: Wright, Terra, and Ping teach the fault monitoring method of claim 1. Wright does not explicitly teach wherein the FOFL is set based on a policy set by a user.
However, Kudva teaches wherein the FOFL is set based on a policy set by a user (i.e. policy engine 474 may also predict faults based on user specified rules; col. 11, lines 55-60).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Wright, Terra, and Ping to include the feature of Kudva. One would have been motivated to make this modification because it provides a way for a system to tailor when and how faults are predicted, reported, or acted on based on the operating environment and the user’s preferred strategy.
Claim 14 is similar in scope to Claim 4 and is rejected under a similar rationale.
11. Claim 18 is rejected under 35 U.S.C. 103 as being unpatentable over Wright in view of Terra, Ping, and further in view of Srivastava et al. (U.S. Patent Pub. No. US 20080215924 A1).
Claim 18: Wright, Terra, and Ping teach the fault prediction method of claim 1. Wright further teaches a second level corresponding to a fault prediction report (i.e. The action can be a reporting action such as notifying an administrator of the issue; para. [0018]).
Wright does not explicitly teach a first level corresponding to a warning report; a third level corresponding to a critical report; and a fourth level corresponding to a self-repair action performed on the computing modules of the main board.
However, Srivastava teaches wherein the FOFL comprises a plurality of levels (i.e. the health information may indicate the sub-system health is at one of multiple pre-defined levels; para. [0022]) including: a first level corresponding to a warning report; a third level corresponding to a critical report (i.e. One level may correspond to the sub-system's health condition as being good. Another health level may correspond to the sub-system's condition as being failed. Other health levels may indicate that a sub-system's condition is somewhere between good and failed, such as a critical level or a warning level. Many variations of levels may be used to characterize the health of a subsystem, all of which are considered within the scope of the present invention; para. [0022]); and a fourth level corresponding to a self-repair action performed on the computing modules of the main board (i.e. the sub-system may perform recovery or corrective action if the sub-system's health status is not acceptable; para. [0030]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Wright, Terra, and Ping to include the feature of Srivastava. One would have been motivated to make this modification because it provides fault management response appropriate to the severity of the detected condition.
12. Claim 19 is rejected under 35 U.S.C. 103 as being unpatentable over Wright in view of Terra, Ping, and further in view of Bhatia et al. (U.S. Patent Pub. No. US 20140195669 A1).
Claim 19: Wright, Terra, and Ping teach the fault prediction method of claim 1. Wright does not explicitly teach wherein the collecting comprises communicating with sensors associated with the computing modules via a system data bus using at least one of an Intelligent Platform Management Interface (IPMI) protocol and an Intelligent Platform Management Bus (IPMB) protocol.
However, Bhatia teaches wherein the collecting comprises communicating with sensors associated with the computing modules (i.e. The assisting management instance 132 can control and collect information of sensors, sensor data record (SDR) devices, and field replaceable units (FRUs); para. [0078]) via a system data bus (i.e. Both master components and slave components are operable to transmit and receive communications over the communication bus 137; para. [0037]) using at least one of an Intelligent Platform Management Interface (IPMI) protocol and an Intelligent Platform Management Bus (IPMB) protocol (i.e. IPMI and IPMB; para. [0009, 0050]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Wright, Terra, and Ping to include the feature of Bhatia. One would have been motivated to make this modification because it provides a standardized and interoperable mechanism for BMC to obtain health information from managed computing components.
13. Claim 20 is rejected under 35 U.S.C. 103 as being unpatentable over Wright in view of Terra, Ping, and further in view of Yellin et al. (U.S. Patent Pub. No. US 20200285891 A1).
Claim 20: Wright, Terra, and Ping teach the fault prediction method of claim 11. Wright does not explicitly teach prior to the receiving, requesting the Al model data from the external Al model platform based on the calculated FOFL, wherein the requested Al model data corresponds to the one Al model selected from among the plurality of Al models respectively corresponding to a plurality of FOFL levels.
Terra further teaches requesting the Al model data based on the calculated FOFL, wherein the requested Al model data corresponds to the one Al model selected from among the plurality of Al models respectively corresponding to a plurality of FOFL levels (i.e. The risk value determined for the state of the environment, based on the safety information, is then used to dynamically select one of a plurality of ML models available to the ML agent … FIG. 4 shows three different alternative options which may be selected depending upon which of model_1, model_2 and model_n is selected. In the example shown in FIG. 4 , model_1 is selected in a high risk scenario, model_2 is selected in a medium risk scenario and model_n is selected in a low risk scenario; para. [0025-0029, 0032, 0041]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Wright and Ping to include the feature of Terra. One would have been motivated to make this modification because it reduces processing resource and power consumption when a lower risk condition permits use of a less complex model while retaining the ability to employ a more complex model when a higher risk condition is detected.
However, Yellin teaches prior to the receiving, requesting the Al model data from the external Al model platform (i.e. building, by an artificial intelligence system, a number of machine learning models and storing the models on a server, wherein each model is customized to different device and user requirements. The server receives a request for a machine learning model from an electronic device, wherein the request includes a device profile comprising device attributes and user characteristics. The server determines which of the machine learning models is best suited for the device profile according to a set of model policies. The machine learning model best suited for the device profile is selected from among the number of models and downloaded to the electronic device; para. [0005]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Wright, Terra, and Ping to include the feature of Yellin. One would have been motivated to make this modification because it would avoid unnecessarily storing or transferring every available model to the resource constrained device.
14. Claim 21 is rejected under 35 U.S.C. 103 as being unpatentable over Wright in view of Terra, Ping, and further in view of Maheshwari et al. (U.S. Patent Pub. No. US 20160104076 A1).
Claim 21: Wright, Terra, and Ping teach the fault prediction method of claim 11. Wright does not explicitly teach receiving, a fault feedback policy set by a user; and setting the FOFL based on the received fault feedback policy, wherein the fault feedback policy defines FOFL levels and corresponding detailed actions for each FOFL level.
However, Maheshwari teaches receiving, a fault feedback policy set by a user (i.e. a user may define a Normal state, a Warning state, and a Critical state for a KPI, and the value produced by the search query of the KPI can indicate the current state of the KPI. In one implementation, one or more thresholds are created for each KPI. Each threshold defines an end of a range of values that represent a particular state of the KPI. A graphical interface can be provided to facilitate user input for creating one or more thresholds for each KPI, naming the states for the KPI, and associating a visual indicator (e.g., color, pattern) to represent a respective state; para. [0242]); and setting the FOFL based on the received fault feedback policy, wherein the fault feedback policy defines FOFL levels (i.e. Each threshold defines an end of a range of values representing a certain state for the KPI. Multiple states can be defined for the KPI (e.g., unknown state, trivial state, informational state, normal state, warning state, error state, and critical state); para. [0279]) and corresponding detailed actions for each FOFL level (i.e. The actions to be triggered can be pre-defined or specified by the user via a user interface (e.g., a GUI or a command line interface) while the user is defining thresholds or after the thresholds have been defined. The action to be triggered in view of thresholds can be included in the service definition identifying the respective KPI(s) or can be stored in a data structure dedicated to store various KPI settings of a relevant KPI; para. [0680]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Wright, Terra, and Ping to include the feature of Maheshwari. One would have been motivated to make this modification because it permits a system administrator to tailor fault severity classifications and corresponding responses to the operating requirements of the system.
15. Claim 22 is rejected under 35 U.S.C. 103 as being unpatentable over Wright in view of Terra, Ping, and further in view of Claussen et al. (U.S. Patent Pub. No. US 20210263493 A1).
Claim 22: Wright, Terra, and Ping teach the fault prediction method of claim 11. Wright does not explicitly teach comprising: transmitting, by the one or more processors, the calculated FOFL together with the monitoring information to the on- device Al prediction processor to request the predicting; and receiving, by the one or more processors, a prediction result from the on-device Al prediction processor, wherein the controlling is performed based on the received prediction result.
Terra further teaches comprising: transmitting, by the one or more processors, the calculated FOFL together with the monitoring information to the on- device Al prediction processor to request the predicting (i.e. The analyser performs a safety analysis to determine a risk value for the state of the environment (see step 3), and then passes the determined risk value to a selector for use in the selection of a ML model from among the plurality of ML models (see step 4). In this example, the information about the state of the environment, including the sensor data, is also passed to a selector; para. [0040, 0041]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the invention of Wright to include the feature of Terra. One would have been motivated to make this modification because it reduces processing resource and power consumption when a lower risk condition permits use of a less complex model while retaining the ability to employ a more complex model when a higher risk condition is detected.
Ping further teaches the on- device Al prediction processor (i.e. Referring to FIG. 4, in some embodiments, a neural network (or “artificial neural network”) is used to estimate the heat load in the SSD. The neural network may be implemented as a suitable set of program instructions executed by the SSD controller 210 or by the baseboard management controller 140, or it may be implemented as an analog neural network configured, with suitable digital to analog and analog to digital converters, to operate as an (on-chip or off-chip) analog co-processor of the SSD controller 210 or of the baseboard management controller 140; para. [0048, 0057]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Wright and Terra to include the feature of Ping. One would have been motivated to make this modification because it provides a dedicated hardware resource for performing the neural network computations rather than burdening the principal BMC processing circuitry.
However, Claussen teaches comprising: transmitting, by the one or more processors, together with the monitoring information to the on- device Al prediction processor (i.e. an artificial intelligence (AI) accelerator processor … to generate the input data values for the machine learning models based on data received from the devices 135, 140, 145. In some instances, the raw input from the device may be directly used as the input data values for the machine learning models; para. [0021, 0022, 0023]) to request the predicting (i.e. Once these input data values are received, each AI accelerator processor 115A, 120A, 125A executes a machine learning model; para. [0022]); and receiving, by the one or more processors, a prediction result from the on-device Al prediction processor (i.e. Each AI accelerator processor 115A, 120A, and 125A applies one or more machine learning models to the input data values to generate one or more output data values. Once generated, the technology modules 115, 120, and 125 transfer the output data values to the CPU module 110 over the backplane bus 130 and the processor in the CPU module 110 uses output data values during execution of the control program; para. [0025]), wherein the controlling is performed based on the received prediction result (i.e. the CPU module executes an application program that utilizes the output data values. The application program may then use the output data values to control the devices. In the example of FIG. 2, at step 235, the CPU module generates one or more control instructions for the devices based on the output data values and at step 240 those control instructions are transferred to the devices; para. [0033]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention to modify the combination of Wright, Terra, and Ping to include the feature of Claussen. One would have been motivated to make this modification because it permits computationally intensive AI processing to be offloaded from the primary controller while making the AI results synchronously available for real-time control.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant’s disclosure.
Wu et al. (Pub. No. US 20210377102 A1), a method for detecting a server fault includes: collecting sample monitoring data of a plurality of servers, the sample monitoring data signifying operating states of the plurality of servers; performing training, based on the sample monitoring data, to obtain a fault detection model for the plurality of servers; and collecting current monitoring data of a target server, and inputting the current monitoring data into the fault detection model to determine an operating fault corresponding to the current monitoring data.
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 extension fee 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 date of this final action.
It is noted that any citation to specific pages, columns, lines, or figures in the prior art references and any interpretation of the references should not be considered to be limiting in any way. A reference is relevant for all it contains and may be relied upon for all that it would have reasonably suggested to one having ordinary skill in the art. In re Heck, 699 F.2d 1331, 1332-33, 216 U.S.P.Q. 1038, 1039 (Fed. Cir. 1983) (quoting In re Lemelson, 397 F.2d 1006, 1009, 158 U.S.P.Q. 275, 277 (C.C.P.A. 1968)).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to TAN TRAN whose telephone number is (303)297-4266. The examiner can normally be reached on Monday - Thursday - 8:00 am - 5:00 pm MT.
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, Matt Ell can be reached on 571-270-3264. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/TAN H TRAN/Primary Examiner, Art Unit 2141
Prosecution Insights