CTNF 18/195,626 CTNF 90432 DETAILED ACTION Notice of Pre-AIA or AIA Status 07-03-aia AIA 15-10-aia The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA. Claim Rejections - 35 USC § 103 07-06 AIA 15-10-15 In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status. 07-20-aia AIA 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. 07-21-aia AIA Claim (s) 1-6, 9-15, and 18-20 is/are rejected under 35 U.S.C. 103 as being unpatentable over Shah, Syed Yousaf, et al. "Dependency analysis of cloud applications for performance monitoring using recurrent neural networks." in view of Hoenig et al. US 2022/0269579 . Regarding claims 1, 10, and 19 , Shah teaches “a method for facilitating automated application performance monitoring threshold management through deep learning model, [the method being implemented by at least one processor], the method comprising: retrieving, by the at least one processor via an application programming interface, raw data that correspond to at least one application, the raw data including application performance data” (abstract “we use performance monitoring data collected from two sources: a controlled experiment involving a sample cloud application that we deployed in a public cloud infrastructure and cloud monitoring data collected from the monitoring service of an operational, public cloud service provider” raw data, or data collected directly without processing and pg. 4 §A ¶1 “We collected cloud performance data using AWS Cloud Watch and implemented a python web service client, which automatically downloads the cloud monitoring data (total of 11 metrics shown in Fig. 4 for application and database) using the AWS CloudWatch APIs. The monitoring data is collected every minute, so the dataset has a data point for every minute”) ; “generating, by the at least one processor, at least one data frame based on the raw data, the at least one data frame relating to a multi-dimensional data structure” (pg. 3 right col. second to last ¶ “When new monitoring data from any layer is available, the system prepares it for model (re)training. It generates the feature matrix ‘d’ using the user defined Learning Function, as described in section III-A1 and initializes the probe vector ‘V’. It then feeds the feature matrix ‘d’ along with the monitoring metric for which the forecasting is desired to the LSTM neural network” converting new (raw) data to a feature matrix is analogous to a data frame as the feature matrix is multi-dimensional) ; “converting, by the at least one processor, the at least one data frame into at least one model” (previous citation, “It then feeds the feature matrix ‘d’ along with the monitoring metric for which the forecasting is desired to the LSTM neural network” feeding the data to the neural network is analagous to converting the data into a model) ; “developing, by the at least one processor, at least one error function that optimizes a regression coefficient” (pg. 4 right col. “For training we use 500 epochs, the activation function as tanh, loss function as MSE and for optimizer we use RMSProp” MSE (mean squared error) is an error function that optimizes a regression coefficient, as is known in the art) ; “training, by the at least one processor, the at least one model by using the at least one error function” (previous citation) ; and “determining, by the at least one processor using the trained at least one model, at least one forecasted threshold value that relates to at least one application performance metric for the at least one application” (pg. 7 left col. “The improved forecasting achieved using our LSTM-based approach can be used for various cloud performance monitoring tasks such as anomaly detection.”) Shah teaches forecasting using the data gathered but does not explicitly teach thresholds. Hoenig however teaches “at least one processor” ([0051] “Computer system 400 may include one or more processors (also called central processing units, or CPUs), such as a processor 404. Processor 404 may be connected to a communication infrastructure or bus 406.”) “[…] threshold value that relates to at least one application performance metric […]” (Hoenig [0011] “FIG. 1 is a block diagram 100 illustrating example functionality for providing a performance metric system (PMS) 102, according to some embodiments. PMS 102 may monitor the performance of applications 104A, 104B based on one or more performance metrics provided by users 110A-110C, aggregate that information from distributed application components, and provide it on a single user interface 112. The user interface 112 may provide for real-time monitoring of one or more user-defined performance metrics 106 against the applications 104A, 104B and may provide alerts 130 when any metric 106 falls below a threshold 116 .”) It would have been obvious to one having ordinary skill in the art at the time that the invention was effectively filed to combine the teachings of Shah with that of Hoenig since a combination of known methods would yield predictable results. As shown in Hoenig, it is known to monitor thresholds in the context of application performance monitoring in order to determine if an anomaly occurs. As Shah pertains to forecasting anomalies, it would be a natural extension to forecast threshold values since they are functionally the same thing. Note that independent claims 10 and 19 recite the same substantial subject matter as independent claim 1, only differing in embodiment. The difference in embodiments, a device and non-transitory medium are taught by Hoenig in the alternative embodiments. Regarding claims 2 and 11, the Shah and Hoenig references have been addressed above. Shah further teaches “wherein the at least one model relates to a relationship between at least one independent feature and at least one dependent feature of the application performance data” (previous citation, the model, a neural network uses a independent feature as input to predict dependent features) Regarding claims 3 and 12, the Shah and Hoenig references have been addressed above. Shah further teaches “wherein the training includes a recurrent training process that minimizes the at least one error function, the recurrent training process including a plurality of computing layers of neural networks” (abstract “In this paper, we propose a novel use of the modeling capability of Long-Short Term Memory (LSTM) recurrent neural networks”) Regarding claims 4 and 13, the Shah and Hoenig references have been addressed above. Shah further teaches “wherein the plurality of computing layers include at least one memory cell that persists learning acquired from a relationship between at least one independent feature and at least one dependent feature of the application performance data” (pg. 2 right col. PNG media_image1.png 328 466 media_image1.png Greyscale ) Regarding claims 5 and 14, the Shah and Hoenig references have been addressed above. Shah further teaches “wherein the plurality of computing layers include a second layer that uses a first output from a first layer to compute at least one partial derivative and update a model parameter, a third layer that trains a second output from the second layer by recomputing the model parameter with a new set of parameters, and a fourth layer that trains a third output from the third layer until the at least one error function converges to a minimum value” (pg. 2 right col. §II “Deep neural networks such as Recurrent Neural Networks (RNNs) have been shown to be very effective in modeling time series and sequential data. The Long Short-Term Memory (LSTM) networks [10] are a type of RNNs that are suitable for capturing long-term dependencies in sequential data, a property that makes them particular appealing for our appli cation domain. A brief description of their design follows, while the interested reader is referred to [10] for more details. Neurons (also called “cells” in neural networks) are organized in layers that are connected with one another. The first (input) layer receives the input data, computes weighted sums on it and then applies an activation function (e.g. tanh, sigmoid, etc.) that produces an output, which is then consumed by the neurons or cells in the next layer and/or itself, as in the case of RNNs . During the training phase, the network tries to learn the weights that minimize the error between the final output of the network and the real value of the data.” it is known in the art that backpropagation is the standard algorithm to train neural networks, which uses partial derivatives) Regarding claims 6 and 15, the Shah and Hoenig references have been addressed above. Shah further teaches “wherein the at least one forecasted threshold value is automatically determined for the at least one application according to a time interval, the time interval including a period of time that is dynamically adjusted based on time series data” (pg. 2 ¶1 “We apply LSTM modeling using the time series data that is collected as part of the performance monitoring of various key performance indicators, which spans over any of the infrastructure/platform/application layers of the cloud applications software stack”) Regarding claims 9 and 18, the Shah and Hoenig references have been addressed above. Shah further teaches “wherein the at least one model includes at least one from among a machine learning model, a statistical model, a mathematical model, a process model, and a data model” (the neural network model from above is a machine learning model) 07-21-aia AIA Claim (s) 7-8 and 16-17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Shah in view of Hoenig further in view of Paul et al. US 2023/0136230 . Regarding claims 7 and 16 , the Shah and Hoenig references have been addressed above. While they discuss API accesses, they do not explicitly teach the claim limitations. Paul however teaches “wherein retrieving the raw data further comprises: generating, by the at least one processor, at least one access token for each of a plurality of access calls that corresponds to the application programming interface” (Paul [0049] “In step 510, the system receives a call to an API backend at an authentication service […] In step 511, an authentication service determines whether a valid access token is associated with the call” wherein determining a valid access token inherently includes the generation of said access token at some point) ; “passing, by the at least one processor, the at least one access token together with the plurality of access calls to the application programming interface, the plurality of access calls including a predetermined expiration time and a set of parameters” (previous citation “If an access token is not present, is expired, or is not associated with any of the authentication services in the overall system, in step 520, the user is redirected to a log-in page to create a new account or sign in to an existing account by providing user login credential” expired i.e. predetermined time) It would have been obvious to one having ordinary skill in the art at the time that the invention was effectively filed to combine the teachings of Shah and Hoenig with that of Paul since a combination of known methods would yield predictable results. As shown in Paul, API accesses often involve access tokens and validation. This is standard practice in the computing world and these techniques would act out as predicted with the system above in order to ensure proper access to data. Shah further teaches “retrieving, by the at least one processor via the application programming interface, the raw data from at least one application performance monitoring toolset” (pg. 4 §A ¶1 “We collected cloud performance data using AWS Cloud Watch and implemented a python web service client, which automatically downloads the cloud monitoring data (total of 11 metrics shown in Fig. 4 for application and database) using the AWS CloudWatch APIs. The monitoring data is collected every minute, so the dataset has a data point for every minute”) Regarding claims 8 and 17 , the Shah, Hoenig, and Paul references have been addressed above. Shah further teaches “wherein the raw data includes at least one application performance metric and an associated hardware performance metric, the at least one application performance metric including an application latency metric” (pg. 6 table I shows latency) Conclusion Any inquiry concerning this communication or earlier communications from the examiner should be directed to KEVIN W FIGUEROA whose telephone number is (571)272-4623. The examiner can normally be reached Monday-Friday, 10AM-6PM EST. 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, MIRANDA HUANG can be reached at (571)270-7092. 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. KEVIN W FIGUEROA Primary Examiner Art Unit 2124 /Kevin W Figueroa/Primary Examiner, Art Unit 2124 Application/Control Number: 18/195,626 Page 2 Art Unit: 2124 Application/Control Number: 18/195,626 Page 3 Art Unit: 2124 Application/Control Number: 18/195,626 Page 4 Art Unit: 2124 Application/Control Number: 18/195,626 Page 5 Art Unit: 2124 Application/Control Number: 18/195,626 Page 6 Art Unit: 2124 Application/Control Number: 18/195,626 Page 7 Art Unit: 2124 Application/Control Number: 18/195,626 Page 8 Art Unit: 2124
Prosecution Insights