DETAILED ACTION
This action is responsive to the amendment filed on 10/07/2025. Claims 1-15 and 20-25 are pending in the case. Claims 1, 8 and 15 are independent claims. Claims 1, 8 and 15 are amended claims. Claims 21-25 are new. Claims 16-20 are cancelled.
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 .
Response to Arguments
Applicant's arguments filed 10/07/2025 have been fully considered but they are not persuasive.
With respect to the rejection under 35 U.S.C 101:
Applicant does not appear to provide specific arguments for the claims eligibility but for noting that the amendments clarify patentability.
Examiner disagrees. Routing as claimed amounts a claim which “recites only the idea of a solution or outcome i.e., the claim fails to recite details of how a solution to a problem is accomplished. The recitation of claim limitations that attempt to cover any solution to an identified problem with no restriction on how the result is accomplished and no description of the mechanism for accomplishing the result”.
The claims does not recite how the routing is performed beyond merely being in connection or based on the resource cost and workload balancing. Further, this claim feature merely makes use of the recited exception for an application at a high degree of generality, again because the mechanism for how the routing is performed is not reflected in the claims.
With respect to the rejection under 35 U.S.C 103:
Applicant first appears to highlight the amendment “routing said workload to a node based on said resource cost and workload balancing” clarifies advantages over the prior art, and as such the rejection should be withdrawn.
Examiner disagrees. Routing based on a predicted resources cost and workload balancing data is taught by Park as highlighted in the updated rejection. Park describes using workload balance information to predict latency (i.e resource cost) in order to inform the configuration and deployment of said workload (see Figure 8 instance deployment, and Section 3.8).
Further, Applicant argues that the claim explicitly requires “a graph neural network model” for “improving stability of said model”. Further noting that Park does not use a graph neural network (GNN) itself for the improvement but rather a particular loss function, the huber loss function.
Examiner disagrees the claim limitation reads:
“improving stability of said model with a graph neural network model”
The claim does not specify that the GNN itself performs the improvement, rather that the improving is performed in an environment of context “with” a GNN. The claim is not read so narrowly to require the GNN is operatively performing the improvement. The improvement to training stability of said model described by Park’s choice of loss function is in the context or “with” training of a graph neural network model (called latency prediction model in the art).
Further, paragraph 0024 of the Specification describes the stability reinforced as a product of how the GNN model is trained. Park describes similarly how particular training accomplishes improved stability in the model.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-15, 21-25 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more. The analysis of the claims will follow the 2016 Revised Patent Subject Matter Eligibility Guidance, 84 Fed. Reg. 50 (“2019 PEG”).
Claim 1
Step 1: The claim recites a system; therefore, it is directed to the statutory category of a machine.
Step 2A Prong 1: The claim recites, inter alia:
training a model: This limitation encompasses the mathematical concept of training a machine learning model. Note that no explicit definition or process for training is disclosed in the specification; therefore, the term is given its ordinary meaning in the art encompassing a gradient descent algorithm, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
enhancing said model with reinforcement learning: This limitation encompasses the mathematical concept of updating a machine learning model’s parameters using the outputs of a reward function, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
predicting, with said model, a resource cost of a workload; and: This limitation encompasses the mental process of predicting the resource cost associated with a given computing task, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: The abstract ideas listed above are not integrated into a practical application. Specifically, the additional elements, a memory; and a processor in communication with said memory, said processor being configured to perform operations, said operations comprising, amount to invoking computers or other machinery merely as tools to perform an existing process. Thus, these additional elements are recited in a manner that represents no more than mere instructions to apply the abstract ideas on a computer (see MPEP § 2106.05(f)).
The additional element, improving stability of said model with a graph neural network model, amounts to invoking computers or other machinery merely as tools to perform an existing process. Specifically, improving stability is interpreted as improving the robustness of a machine learning model which can be achieved by applying noisy inputs as part of the training algorithm. This mathematical concept, executed by a graph neural network model, amounts to mere instructions to apply the abstract ideas on a computer (see MPEP § 2106.05(f)).
The additional element, routing said workload to a node based on said resource cost and workload balancing, describes an application at a high degree of generality which makes use of the recited exception, see MPEP 2106.05(f)
Nothing in the claim integrates the abstract ideas into a practical application, and the claim is thus directed to the abstract ideas.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. The claim recites [the] additional element[s] [of/such as] [what elements are/do] [2106.05(f-h) reasoning]. The additional element[s] listed above do not amount to significantly more than the abstract ideas. Therefore, the claim is subject-matter ineligible.
Claim 2
Step 1: A machine, as above.
Step 2A Prong 1: The claim recites:
generating a callback with an actual consumption and a predicted consumption; and: This limitation encompasses the mental process of calculating a difference between a predicted value with an actual value, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
updating said model with said callback: This limitation encompasses the mathematical concept of updating the parameters of a machine learning model using an error value, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: There are no additional elements in the claim that integrate the abstract idea into a practical application, and the claim is thus directed to the abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. Therefore, the claim is subject-matter ineligible.
Claim 3
Step 1: A machine, as above.
Step 2A Prong 1: The claim recites:
estimating a minimum consumption of said workload and a maximum consumption of said workload; and: This limitation encompasses the mental process of estimating a minimum and maximum consumption associated with a computing task, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
using said minimum consumption and said maximum consumption to predict said resource cost: This limitation encompasses the mental process of predicting a resource cost from consumption estimations, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: There are no additional elements in the claim that integrate the abstract idea into a practical application, and the claim is thus directed to the abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. Therefore, the claim is subject-matter ineligible.
Claim 4
Step 1: A machine, as above.
Step 2A Prong 1: The claim recites:
training said graph neural network model for adversarial attacks: This limitation encompasses the mental process of introducing noise into the training process to make the resultant model robust to adversarial attacks, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: There are no additional elements in the claim that integrate the abstract idea into a practical application, and the claim is thus directed to the abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. Therefore, the claim is subject-matter ineligible.
Claim 5
Step 1: A machine, as above.
Step 2A Prong 1: The claim recites:
confirming a model result using a maximum cut method: This limitation encompasses the mathematical concept of applying a maximum cut algorithm to a graph network, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: There are no additional elements in the claim that integrate the abstract idea into a practical application, and the claim is thus directed to the abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. Therefore, the claim is subject-matter ineligible.
Claim 6
Step 1: A machine, as above.
Step 2A Prong 1: The claim recites:
adopting said model within a first specified time period: This limitation encompasses the mental process of adopting a model within a certain time period, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
wherein parameters are updated within a second specified time period to enable adopting said model within said first specified time period: This limitation encompasses the mental process of updating model parameters within a certain time period, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: There are no additional elements in the claim that integrate the abstract idea into a practical application, and the claim is thus directed to the abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. Therefore, the claim is subject-matter ineligible.
Claim 7
Step 1: A machine, as above.
Step 2A Prong 1: The claim recites, inter alia:
providing a real-time response to an incoming request: This limitation encompasses the mental process of responding to a system request, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: The abstract idea presented above is not integrated into a practical application. Specifically, the additional element, with said model, amounts to invoking computers or other machinery merely as tools to perform an existing process. Thus, this additional element is recited in a manner that represents no more than mere instructions to apply the abstract ideas on a computer (see MPEP § 2106.05(f)).
Nothing in the claim integrates the abstract ideas into a practical application, and the claim is thus directed to the abstract ideas.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. The claim recites an additional element directed to using a model to respond to a resource request thereby applying the judicial exceptions on a computing device. As such, the additional element listed above does not amount to significantly more than the abstract ideas. Therefore, the claim is subject-matter ineligible.
Claim 8
Step 1: The claim recites a computer-implemented method; therefore, it is directed to the statutory category of a process.
Step 2A Prong 1: The claim recites, inter alia:
training a model: This limitation encompasses the mathematical concept of training a machine learning model. Note that no explicit definition or process for training is disclosed in the specification; therefore, the term is given its ordinary meaning in the art encompassing a gradient descent algorithm, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
enhancing said model with reinforcement learning: This limitation encompasses the mathematical concept of updating a machine learning model’s parameters using the outputs of a reward function, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
predicting, with said model, a resource cost of a workload; and: This limitation encompasses the mental process of predicting the resource cost associated with a given computing task, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: The abstract ideas listed above are not integrated into a practical application. Specifically, the additional element, improving stability of said model with a graph neural network model, amounts to invoking computers or other machinery merely as tools to perform an existing process. Specifically, improving stability is interpreted as improving the robustness of a machine learning model which can be achieved by applying noisy inputs as part of the training algorithm. This mathematical concept, executed by a graph neural network model, amounts to mere instructions to apply the abstract ideas on a computer (see MPEP § 2106.05(f)).
The additional element, routing said workload to a node based on said resource cost and workload balancing, describes an application at a high degree of generality which makes use of the recited exception, see MPEP 2106.05(f)
Nothing in the claim integrates the abstract ideas into a practical application, and the claim is thus directed to the abstract ideas.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. The claim recites [the] additional element[s] [of/such as] [what elements are/do] [2106.05(f-h) reasoning]. The additional element[s] listed above do not amount to significantly more than the abstract ideas. Therefore, the claim is subject-matter ineligible.
Claim 9
Step 1: A process, as above.
Step 2A Prong 1: The claim recites:
generating a callback with an actual consumption and a predicted consumption; and: This limitation encompasses the mental process of calculating a difference between a predicted value with an actual value, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
updating said model with said callback: This limitation encompasses the mathematical concept of updating the parameters of a machine learning model using an error value, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: There are no additional elements in the claim that integrate the abstract idea into a practical application, and the claim is thus directed to the abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. Therefore, the claim is subject-matter ineligible.
Claim 10
Step 1: A process, as above.
Step 2A Prong 1: The claim recites:
estimating a minimum consumption of said workload and a maximum consumption of said workload; and: This limitation encompasses the mental process of estimating a minimum and maximum consumption associated with a computing task, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
using said minimum consumption and said maximum consumption to predict said resource cost: This limitation encompasses the mental process of predicting a resource cost from consumption estimations, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: There are no additional elements in the claim that integrate the abstract idea into a practical application, and the claim is thus directed to the abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. Therefore, the claim is subject-matter ineligible.
Claim 11
Step 1: A process, as above.
Step 2A Prong 1: The claim recites:
training said graph neural network model for adversarial attacks: This limitation encompasses the mental process of introducing noise into the training process to make the resultant model robust to adversarial attacks, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: There are no additional elements in the claim that integrate the abstract idea into a practical application, and the claim is thus directed to the abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. Therefore, the claim is subject-matter ineligible.
Claim 12
Step 1: A process, as above.
Step 2A Prong 1: The claim recites:
confirming a model result using a maximum cut method: This limitation encompasses the mathematical concept of applying a maximum cut algorithm to a graph network, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: There are no additional elements in the claim that integrate the abstract idea into a practical application, and the claim is thus directed to the abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. Therefore, the claim is subject-matter ineligible.
Claim 13
Step 1: A process, as above.
Step 2A Prong 1: The claim recites:
adopting said model within a first specified time period: This limitation encompasses the mental process of adopting a model within a certain time period, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
wherein parameters are updated within a second specified time period to enable adopting said model within said first specified time period: This limitation encompasses the mental process of updating model parameters within a certain time period, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: There are no additional elements in the claim that integrate the abstract idea into a practical application, and the claim is thus directed to the abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. Therefore, the claim is subject-matter ineligible.
Claim 14
Step 1: A process, as above.
Step 2A Prong 1: The claim recites, inter alia:
providing a real-time response to an incoming request: This limitation encompasses the mental process of responding to a system request, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: The abstract idea presented above is not integrated into a practical application. Specifically, the additional element, with said model, amounts to invoking computers or other machinery merely as tools to perform an existing process. Thus, this additional element is recited in a manner that represents no more than mere instructions to apply the abstract ideas on a computer (see MPEP § 2106.05(f)).
Nothing in the claim integrates the abstract ideas into a practical application, and the claim is thus directed to the abstract ideas.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. The claim recites an additional element directed to using a model to respond to a resource request thereby applying the judicial exceptions on a computing device. As such, the additional element listed above does not amount to significantly more than the abstract ideas. Therefore, the claim is subject-matter ineligible.
Claim 15
Step 1: The claim recites a computer program product; therefore, it is directed to the statutory category of an article of manufacture. Note that paragraph [0121] indicates that “[a] computer readable storage medium, as used herein, is not to be construed as being
transitory signals per se” thereby qualifying it as statutory subject matter.
Step 2A Prong 1: The claim recites, inter alia:
training a model: This limitation encompasses the mathematical concept of training a machine learning model. Note that no explicit definition or process for training is disclosed in the specification; therefore, the term is given its ordinary meaning in the art encompassing a gradient descent algorithm, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
enhancing said model with reinforcement learning: This limitation encompasses the mathematical concept of updating a machine learning model’s parameters using the outputs of a reward function, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
predicting, with said model, a resource cost of a workload; and: This limitation encompasses the mental process of predicting the resource cost associated with a given computing task, which is an evaluation practically capable of being performed in the human mind with the assistance of pen and paper.
Step 2A Prong 2: The abstract ideas listed above are not integrated into a practical application. Specifically, the additional elements, computer readable storage medium having program instructions embodied therewith, said program instructions executable by a processor to cause said processor to perform a function, said function comprising, amount to invoking computers or other machinery merely as tools to perform an existing process. Thus, these additional elements are recited in a manner that represents no more than mere instructions to apply the abstract ideas on a computer (see MPEP § 2106.05(f)).
The additional element, improving stability of said model with a graph neural network model, amounts to invoking computers or other machinery merely as tools to perform an existing process. Specifically, improving stability is interpreted as improving the robustness of a machine learning model which can be achieved by applying noisy inputs as part of the training algorithm. This mathematical concept, executed by a graph neural network model, amounts to mere instructions to apply the abstract ideas on a computer (see MPEP § 2106.05(f)).
The additional element, routing said workload to a node based on said resource cost and workload balancing, describes an application at a high degree of generality which makes use of the recited exception, see MPEP 2106.05(f)
Nothing in the claim integrates the abstract ideas into a practical application, and the claim is thus directed to the abstract ideas.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. As such, the additional element listed above does not amount to significantly more than the abstract ideas. Therefore, the claim is subject-matter ineligible.
Claim 21
Step 1: A machine, as above.
Step 2A Prong 1: The claim recites no further abstract ideas beyond those already considered by the parent claim.
Step 2A Prong 2: The abstract idea presented above is not integrated into a practical application. Specifically, the additional element, balancing a plurality of workloads over a cluster, describes an application at a high degree of generality which makes use of the recited exception, see MPEP 2106.05(f)
Nothing in the claim integrates the abstract ideas into a practical application, and the claim is thus directed to the abstract ideas.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. As such, the additional element listed above does not amount to significantly more than the abstract ideas. Therefore, the claim is subject-matter ineligible.
Claim 22
Step 1: A machine, as above.
Step 2A Prong 1: The claim recites no further abstract ideas beyond those already considered by the parent claim.
Step 2A Prong 2: The abstract idea presented above is not integrated into a practical application. Specifically, the additional element, workload balancing improves use of impacted resources, recites only the idea of a solution or outcome i.e., the claim fails to recite details of how a solution to a problem is accomplished (2106.05(f))
Nothing in the claim integrates the abstract ideas into a practical application, and the claim is thus directed to the abstract ideas.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. As such, the additional element listed above does not amount to significantly more than the abstract ideas. Therefore, the claim is subject-matter ineligible.
Claim 23
Step 1: A process, as above.
Step 2A Prong 1: The claim recites:
selecting said node for said workload because said node is an optimal node for said workload based on a goal: This limitation encompasses the mental process of selecting a node based on data which is an evaluation practically capable of being performed in the human mind.
Step 2A Prong 2: There are no additional elements in the claim that integrate the abstract idea into a practical application, and the claim is thus directed to the abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. Therefore, the claim is subject-matter ineligible.
Claim 24
Step 1: A machine, as above.
Step 2A Prong 1: The claim recites no further abstract ideas beyond those already considered by the parent claim.
Step 2A Prong 2: The abstract idea presented above is not integrated into a practical application. Specifically, the additional element, improving a robustness of said model by minimizing noise, recites only the idea of a solution or outcome i.e., the claim fails to recite details of how a solution to a problem is accomplished (2106.05(f))
Nothing in the claim integrates the abstract ideas into a practical application, and the claim is thus directed to the abstract ideas.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. As such, the additional element listed above does not amount to significantly more than the abstract ideas. Therefore, the claim is subject-matter ineligible.
Claim 25
Step 1: A process, as above.
Step 2A Prong 1: The claim recites:
correcting said model for data noise This limitation encompasses the mental process of selecting a node based on data which is an evaluation practically capable of being performed in the human mind.
Step 2A Prong 2: There are no additional elements in the claim that integrate the abstract idea into a practical application, and the claim is thus directed to the abstract idea.
Step 2B: The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception because when considered separately or in combination, they do not constitute an inventive concept. Therefore, the claim is subject-matter ineligible.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 23 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
The term “because said node is an optimal node” in claim 23 is a relative term which renders the claim indefinite. The term “optimal” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention.
Claim Rejections - 35 USC § 103
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.
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, 6-11, 13-18 and 21-25 are rejected under 35 U.S.C. 103 as being unpatentable over Park et al. (“GRAF: A Graph Neural Network based Proactive Resource Allocation Framework for SLO-Oriented Microservices,” hereinafter Park) in view of Yang et al. (“MIRAS: Model-based Reinforcement Learning for Microservice Resource Allocation over Scientific Workflows,” hereinafter Yang).
Regarding claim 1, Park teaches [a] system, said system comprising: a memory; and a processor in communication with said memory, said processor being configured to perform operations, said operations comprising: (Park, 5 Evaluation, pp. 161, cols. 1 and 2; “We deploy Kubernetes clusters on 7 machines equipped with 2 Intel E5-2650 CPUs and 128GB of memory. We use a machine for the Kubernetes master node and the rest for the worker nodes.”)
training a model; (Park, 3.4 Latency Prediction Model, pp. 159, col. 1, paragraph 2; “To successfully train the end-to-end tail latency prediction model, we carefully design input state features, neural network structure, and loss function,” wherein to “train the end-to-end latency prediction model” encompasses training a model.)
Park does not explicitly teach enhancing said model with reinforcement learning. However, Yang, in the area of microservice resource allocation, teaches this limitation (Yang, I. Introduction, pp. 123, col. 1, paragraph 2; “We take the microservice workflow system (to be introduced in Section II) as the target environment where we perform resource adaptation and leverage recent advances in model-based reinforcement learning to design a control policy for microservice workflow system using past experience of interacting with the microservice infrastructure”).
Yang is analogous to the claimed invention as both are from the same field of endeavor, that is, microservice resource allocation in distributed computing systems. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the training method of Park to incorporate reinforcement learning, as taught by Yang. The motivation to do so is to harness the adaptability of reinforcement learning’s reward scheme to guarantee optimization in a dynamic system (Yang, I. Introduction, pp. 123, col. 1, paragraph 2; “We choose reinforcement learning because it directly optimizes long-term reward in dynamic environment.”).
Park further teaches improving stability of said model with a graph neural network model; (Park, 3.1 GRAF Overview, pp. 158, col. 1, paragraph 7; “Note that Latency Prediction Model asynchronously trains the end-to-end tail latency prediction model with collected samples utilizing a GNN.” Park, 3.4 Latency Prediction Model, pp. 159, col. 2, paragraph 4; “Second, we choose the hüber loss function instead of the mean-square loss function to increase stability during training.”)
predicting, with said model, a resource cost of a workload; and (Park, Figure 9;
PNG
media_image1.png
210
366
media_image1.png
Greyscale
In the figure above, the “graph neural network” corresponds to said model, and “to predict latency” is equivalent to predicting…a resource cost of a workload.)
routing said workload to a node based on said resource cost and workload balancing (Park, 3.6 Resource Controller, pp 160 “The resource control module scale observed workload moderately to fit into the latency prediction model. Scaled workloads are fed into the configuration solver, which gives out optimal resource configuration as the output. … With the processed resource configuration, the resource controller calculates the number of instances to scale in/out for every microservices,” Park Section 3.1 pp 158 “Then, the resource control module in Resource Controller makes scaling decisions on microservices according to calculated resource configuration” Figure 8 pp 158
PNG
media_image2.png
485
1035
media_image2.png
Greyscale
as shown in the instance deployment amounts to routing workloads to a node, based on the predicted latency (i.e resource cost) and workload balancing managed by the configuration solver.)
Regarding claim 2, the combination of Park and Yang teaches the system of claim 1 (and thus the rejection of claim 1 is incorporated).
Park further teaches said operations further comprising: generating a callback with an actual consumption and a predicted consumption; and (Park, 3.4 Latency Prediction Model, pp. 159, col. 2, paragraph 4; “Third, we introduce asymmetry in the loss function, it gives more penalty if the latency prediction of the model is lower than the actual value and gives less penalty, otherwise,” wherein the “penalty” of a “loss function” is equivalent to a callback with an actual consumption and a predicted consumption.)
updating said model with said callback (Park, 3.4 Latency Prediction Model, pp. 159, col. 2, paragraph 4; “Therefore, we avoid underestimation of our latency prediction model by penalizing more when it guess latency to be shorter than actual,” wherein “penalizing” a model necessarily involves updating said model with said callback generated by the “loss function.”).
Regarding claim 3, Park teaches the system of claim 1 (and thus the rejection of claim 1 is incorporated).
Park further teaches estimating a minimum consumption of said workload and a maximum consumption of said workload; and (Park, Algorithm 1, lines 11 and 16;
PNG
media_image3.png
320
353
media_image3.png
Greyscale
Park, 3.7 Sample Collector, pp. 160, col. 2, paragraph 2; “An upper bound for each microservice
is found by collocating large CPU resources for every microservices then reducing the target microservice’s CPU resource step by step. The latency for each microservice has a lower bound due to the required minimal CPU cycles to handle a request,” wherein finding “the required minimal CPU cycles to handle a request” or “a lower bound” is equivalent to estimating a minimum consumption of said workload. Thus, finding the corresponding “upper bound for each microservice” is equivalent to estimating…a maximum consumption of said workload)
using said minimum consumption and said maximum consumption to predict said resource cost (Park, 3.7 Sample Collector, pp. 160, col. 1, paragraph 5; “Building a big enough training dataset is necessary to train GNN at predicting end-to-end percentile latency with high accuracy.” Therefore, finding “a lower bound” and “a higher bound” corresponding to said minimum consumption and said maximum consumption is done in part to predict said resource cost, or “latency.”).
Regarding claim 4, the combination of Park and Yang teaches the system of claim 1 (and thus the rejection of claim 1 is incorporated).
Park further teaches said operations further comprising: training said graph neural network for adversarial attacks (Park, 3.4 Latency Prediction Model, pp. 159, col. 2, paragraph 4; “Second, we choose the hüber loss function instead of the mean-square loss function to increase stability during training. The Hüber loss function is designed to give out mean-square-error towards small error within bounds and linear error when a large error that’s out of bounds occurs. Thus, irregular samples that show extreme values in some of the collected samples have less effect during the training process,” wherein to “increase stability during training” by adjusting the error given for “small error within bounds” amounts to defending against data poisoning, which is a form of adversarial attack.).
Regarding claim 6, the combination of Park and Yang teaches the system of claim 1 (and thus the rejection of claim 1 is incorporated).
Park further teaches said operations further comprising: adopting said model within a first specified time period, (Park, 5.3 End-to-End Performance Evaluation, pp. 164, col. 1, paragraph 3; “GRAF creates the required instances concurrently at around 50 seconds,” wherein the time taken for “creat[ing] the required instances,” or responding to a request, is equivalent to adopting said model within a first specified time period) wherein parameters are updated within a second specified time period to enable adopting said model within said first specified time period (Park, 5.2 Resource Optimization Analysis, pp. 162, col. 2, paragraph 1; “Also, the gradient descent algorithm’s 90%-tile latency to reach the target tolerance threshold takes about 6.7 seconds, fast enough to make resource allocation decisions synchronously in microservices.” The “6.7 seconds” wherein parameters are updated via gradient descent encompasses a second specified time period that falls within the “50 seconds” required to respond to a request, and, thus, adopt said model.).
Regarding claim 7, the combination of Park and Yang teaches the system of claim 1 (and thus the rejection of claim 1 is incorporated).
Park further teaches said operations further comprising: providing a real-time response to an incoming request with said model (Park, 6 Discussion and Future Work, pp. 164, col. 1, paragraph 4; “GRAF has complete scalability toward handling multiple chains of different request types at once. This is possible due to our workload analyzer and node embedding procedure in the system design,” wherein “handling multiple chains of different request types at once” encompasses providing a real-time response to an incoming request with said model.).
Claims 8-11, 13 and 14 are method claims corresponding to the steps of claims 1-4, 6 and 7, and are therefore rejected for the same reasons as claims 1-4, 6 and 7.
Claim 15 is a product claims corresponding to the steps of claims 1, and are therefore rejected for the same reasons as claims 1
Regarding claim 21, the combination of Park and Yang teaches the system of claim 1 (and thus the rejection of claim 1 is incorporated).
Park further teaches balancing a plurality of workloads over a cluster. (Section 3.1 pp 158 “GRAFoperates as an end-to-end resource allocator with six components geared up together within a microservice application deployed onto real Kubernetes cluster… Once the configuration is found, the resource controller (§ 3.6) calculates corresponding instances for each microservice and make scaling decision to the cluster.”)
Regarding claim 22, the combination of Park and Yang teaches the system of claim 1 (and thus the rejection of claim 1 is incorporated).
Park further teaches workload balancing improves use of impacted resources. (Section 3.1 pp 158 “GRAFoperates as an end-to-end resource allocator with six components geared up together within a microservice application deployed onto real Kubernetes cluster… The optimal configuration is found by iterating through possible resource combinations…Once the configuration is found, the resource controller (§ 3.6) calculates corresponding instances for each microservice and make scaling decision to the cluster.” This claim described the intended result and is not given patentable weight, nevertheless the clustering improves impacted resources by providing optimal configuration.)
Regarding claim 23, the combination of Park and Yang teaches the system of claim 1 (and thus the rejection of claim 1 is incorporated).
Park further teaches selecting said node for said workload because said node is an optimal node for said workload based on a goal.
(Section 3.1 pp 158 “GRAFoperates as an end-to-end resource allocator with six components geared up together within a microservice application deployed onto real Kubernetes cluster… The optimal configuration is found by iterating through possible resource combinations…Once the configuration is found, the resource controller (§ 3.6) calculates corresponding instances for each microservice and make scaling decision to the cluster.” The optimal configuration of resources on a set of clusters for a workload is a selection of the optimal workload based on resource constrains such as latency (i.e based on a goal). In the context of the art, scaling is the placement of resources and workloads on different clusters or nodes)
Regarding claim 24, the combination of Park and Yang teaches the system of claim 1 (and thus the rejection of claim 1 is incorporated).
Park further teaches improving a robustness of said model by minimizing noise. (Park, 3.4 Latency Prediction Model, pp. 159, col. 2, paragraph 4; “Second, we choose the hüber loss function instead of the mean-square loss function to increase stability during training. The Hüber loss function is designed to give out mean-square-error towards small error within bounds and linear error when a large error that’s out of bounds occurs. Thus, irregular samples that show extreme values in some of the collected samples have less effect during the training process,” Examiner notes that “improving a robustness” is an intended result and is not given patentable weight. The Huber loss minimizes the affect of outlier noise (i.e irregular extreme values ) via the use of the Huber loss function. Thus the model noise is minimized via training with such a loss function, which is interpreted as improving robustness)
Regarding claim 25, the combination of Park and Yang teaches the system of claim 1 (and thus the rejection of claim 1 is incorporated).
Park further teaches correcting said model for data noise. (Park, 3.4 Latency Prediction Model, pp. 159, col. 2, paragraph 4; “Second, we choose the hüber loss function instead of the mean-square loss function to increase stability during training. The Hüber loss function is designed to give out mean-square-error towards small error within bounds and linear error when a large error that’s out of bounds occurs. Thus, irregular samples that show extreme values in some of the collected samples have less effect during the training process,” The Huber loss minimizes the affect of outlier noise (i.e irregular extreme values ) via the use of the Huber loss function. Thus the training with such a loss corrects the model for outlier data noise. Further, Examiner highlights the specification paragraph 0071 notes minimizing the impact of unreasonable data on the model is an example of correcting for noise. This is precisely the goal of the Huber loss in the art)
Claims 5, 12 are rejected under 35 U.S.C. 103 as being unpatentable over Park in view of Yang in further view of Baraki et al. (“Optimizing Applications for Mobile Cloud Computing Through MOCCAA,” hereinafter Baraki).
Regarding claim 5, the combination of Park and Yang teaches the system of claim 1 (and thus the rejection of claim 1 is incorporated).
Park does not explicitly teach said operations further comprising: confirming a model result using a maximum cut method. However, Baraki, in the area of optimizing resource allocations in cloud computing systems, teaches this limitation (Baraki, 4.2 Max-Flow-Max0Cut Heuristic, pp. 658, col. 1, paragraph 1; “We apply our Max-Flow-Max-Cut heuristic in order to decide which methods shall be executed in the Cloud,” wherein “decid[ing] which methods shall be executed” encompasses confirming a model result in accordance with the explanation provided at paragraph [0048] of the specification of the claimed invention, “The results from the model may be confirmed using the maximum cut method 152. The maximum cut method 152 may be used, for example, to identify, compare, address, and/or correct workload balances across the cluster 170” (emphasis added).).
Baraki is analogous to the claimed invention as both are from the same field of endeavor, that is, workload optimization in distributed computing systems. Therefore, it would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify the GNN-based latency prediction model of Park to incorporate the Max-Flow-Max-Cut algorithm of Baraki. The motivation to do so is to model nested method calls that arise when microservices invoke other microservices as is the case in practical applications ranging from operating system optimization to resource management in cloud computing (Baraki, 4.2 Max-Flow-Max-Cut, pp. 658, col. 1, paragraph 3; “Assuming that each edge has the weight 1, a zigzag course of method invocations will be obtained since as many edges as possible are cut…The remote methods are calling local methods. Nonetheless, the weights of the edges in our graph include the gained benefit of all succeeding nodes as we assume that they are offloaded too. This represents the character of applications where methods invoke other methods in turn.”).
Claim 12 is a method claim corresponding to the steps of claims 5, and is therefore rejected for the same reasons as claim 5.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Joseph et al. (“Fuzzy Reinforcement Learning based Microservice Allocation in Cloud Computing Environments”) discloses a reinforcement-learning based microservice allocation algorithm for cloud computing systems.
Ni et al. (“Generalizable Resource Allocation in Stream Processing via Deep Reinforcement Learning”) discloses a framework for resource allocation generally applicable to any distributed computing system utilizing reinforcement learning and graph partitioning.
Xie et al. (“Virtualized Network Function Forwarding Graph Placing in SDN and NFV-Enabled IoT Networks: A Graph Neural Network Assisted Deep Reinforcement Learning Method”) discloses a framework for optimized resource allocation in Internet of Things (IoT) networks utilizing graph neural networks and reinforcement learning.
THIS ACTION IS MADE FINAL. 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 mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JOHNATHAN R GERMICK whose telephone number is (571)272-8363. The examiner can normally be reached M-F 9:30-4:30.
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, Kakali Chaki can be reached on 571-272-3719. 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.
/J.R.G./
Examiner, Art Unit 2122
/KAKALI CHAKI/Supervisory Patent Examiner, Art Unit 2122
Prosecution Insights