DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Responsive to the communication dated 3/6/2023.
Claims 1 – 20 are presented for examination.
Priority
ADS dated 2/16/2023 does not claim any domestic or international priority.
The effective filing date is 2/16/2023.
Information Disclosure Statement
IDS dated 2/16/2023 and 3/6/2023 have been reviewed. See attached.
Drawings
The drawings dated 2/16/2023 have been reviewed. They are accepted.
Specification
The abstract dated 2/16/2023 has 174 words, 14 lines, and no legal phraseology. The abstract is objected to because it contains more than 150 words.
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 2 and 3 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.
Claim 2 recites the limitation "the alert" . There is insufficient antecedent basis for this limitation in the claim.
Claim 3 is rejected due to its dependency from claim 1.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 16, 2, 3, 6, 8, 10, 9, 19, 20 are rejected under 35 U.S.C. 103 as being unpatentable over Hofmann_2020 (Hidden Markov Model and Their Application for Predicting Failure Events, Springer Natural Switzerland AG 2020) in view of Polze_2011 (Timely Virtual Machine Migration for Pro-Active Fault Tolerance 2011 14th IEEE International Symposium on Object/Component/Service-Oriented Real-Time Distributed Computing Workshops) in view of Browning_2006 (Multilocus Association Mapping Using Variable-Length Markov Chains, The American Journal of Human Genetics Volume 78 June 2006).
Claim 1, Hofmann_2020 makes obvious “A method comprising: accessing data representing a directed graph of model states for a [asset], wherein the directed graph comprises a plurality of transitions among the model states, and the directed graph comprises, for each transition of the transitions, a transition identifier associated with the transition and a probability associated with the transition;
Characterizing a history of the computer system to reach a current state of the computer system, wherein the current state corresponds to a given model state of the model states, the history corresponds to first transitions of the plurality of transitions, and characterizing the history comprises:
Based on the data, identifying first transition identifiers of the transition identifiers associated with the first transitions; and
Applying a first arithmetic operator to the first transition identifiers
Based on the likelihood, initiating responsive action in anticipation of the [asset] transitioning to the given future state” (page 464 abstract: “… Markov… models… can be used to predict the degradation of assets. We model the degradation path of individual assets, to predict overall failure… Markov states… Markov decision process (POMDP) to dynamically optimize the policy for when and how to maintain the asset.” [page 464 section 1: “predictive maintenance is an important topic in asset management… in this paper we introduce… Markov mixed membership models (MMMM). We show how we use these models to learn complex degradation patterns from data by introducing a terminal state that represents the failure state of the asset, whereas other states represent health-states of the asset as it progresses towards failure. The probability distribution of these non-terminal states and the transition probabilities between states are learned from non-stationary time-series data gathered as historic data, as well as real time streaming data… dynamic failure prediction… to find optimal replacement and repair policies…”; page 466 section 2 A Brief Introduction to the Hidden Markov Model: “a hidden Markov model is a generative graph model that represents probability distributions over sequence of observations [7]. It involves two interconnected models. The state model consists of a discrete-time, discrete-state first-order Markov chain zt ∈ {1,...,N} that transitions according to p(zt|zt−1), while the observation model is governed by p(xt|zt), where xt are the observations. The corresponding joint distribution of a sequence of states and observations can be factored as:
PNG
media_image1.png
102
665
media_image1.png
Greyscale
Therefore, to fully define this probability distribution, we need to specify a probability distribution over the initial state p(z1), a N × N state transition matrix defining all transition probabilities p(zt|zt−1), and the emission probability model p(xt|zt). To summarize, the HMM generative model has the following
assumptions:
Each observation xt is generated by a hidden state zt.
Transition probabilities between states p(zt|zt−1) represented by a transition matrix are constant.
At time t, an observation xt has a certain probability distribution corresponding to possible hidden states.
States are finite and satisfy first-order Markov property.”
Page 470 Fig. 3 Value Function: replace, repair, do-nothing; Fig. 5, Fig. 8, , Fig. 9).
Hofmann_2020 at page 468 states: “… additionally, one can be interested in computing the most probable state sequence path, Z*, given the entire data sequence. This is the maximum a posteriori (MAP) estimate and can be computed through the Viterbi algorithm” implicitly teaches to utilize historic state information. In an HMM, the underlying system is a Markove chain, but the states are not directly visible. Here, the identifier tags track the most likely sequence of hidden states based on the observed outputs and is often implemented via the Viterbi Algorithm. A Viterbi decoding is effectively an identifier (ID) that represents the exact mathematical state (or condition) of an encoding system at a specific point in time. Nevertheless, Hofmann_2020 doesn’t explicitly recite: “to provide a history identifier corresponding to the history; and Based on the history identifier, determining a likelihood”
While Hofmann_2020 teaches to use Markov chains and the directed state changes (i.e., graph) to predict failures and to identify the optimal action to take with regard to the asset, Hofmann_2020 does not recite that the asset is specifically a “computer system” asset.
Polze_2011, however, makes obvious that an asset may be a “computer system” (abstract: “… we propose an architectural blueprint for managing server system dependability in a pro-active fashion, in order to keep service-level promises for response times and availability even with increasing hardware failure rates. We introduce the concept of anticipatory virtual machine migration that proactively moves computation away from faulty or suspicious machines. The migration decision is based on health indicators… live migration techniques can be triggered in order to move computation to healthy machines even before a failure…”; page 234 section 1: “… live migration capabilities for virtual machines running the actual service. In such frameworks, a virtual machine (VM) can be migrated during runtime from one physical machine to another physical machine… See Figure 1…”; page 235: “… we propose as alternative a recovery scheme called proactive migration, were the virtual machine is migrated to another physical host with respect to timing constraints – before a failure occurs…” page 235 section II Approach: “… health-indicator… migration controller finally performs the migration…”; Figure 2; Page 238: “… in early research we have successfully predicted performance failures of a commercial telecommunication platform based on error log analysis in hidden semi-Markov models [19]. We showed how… provide very good prediction accuracy…” Figure 3; page 239: “… Bronevetsky et al. [26] facilitate semi-Markov models of normal timing behavior of MPI applications… our analysis shows that the existing body of research results for time failure prediction is suitable…”; page 242 Conclusion: “… prediction of failures at various machine levels and the combination of these indicators into a global health indicator… prediction approach can provide enough lead time for the failure avoidance mechanism (such as virtual machine migration), the overall pro-active fault tolerance approach has a good chance to act as supplement to reactive fault tolerance schemes in use today…”).
Hofmann_2020 and Polze_2011 are analogous art because they are from the same field of endeavor called fault prediction. Before the effective filing date, it would have been obvious to a person of ordinary skill in the art to combine Hofmann_2020 and Polze_2011. The rationale for doing so would have been that Hofmann_2020 teaches to predict asset faults using Markov chain models follow the degradation path between asset states and then to preemptively find the optimal replacement and repair policies for the failing asset. Polze_2011 teaches to build a health-indicator that incorporates Markov models and to sue the health-indicator to preemptively perform a VM migration on an asset known as a computer server system. Therefore, it would have been obvious to combine Hofmann_2020 and Polze_2011 for the benefit of predicting failure of an asset called a computer server system to improve fault tolerance of server farms that to obtain the invention as specified in the claims.
Browning_2006 makes obvious “to provide a history identifier corresponding to the history; and Based on the history identifier, determining a likelihood” (page 904:
PNG
media_image2.png
533
575
media_image2.png
Greyscale
PNG
media_image3.png
426
571
media_image3.png
Greyscale
EXAMINER NOTE: Instead of the standard Markov assumption where the next state depends solely on the current state (memoryless), a history identifier (or history state) is assigned to the transition edge. The transition graph’s edges are conditioned on the sequence history of preceding states (e.g., Xt-1, Xt-2) rather than just Xt). This identifier allows the graph to "remember" previous steps and dynamically change transition probabilities depending on the trajectory the chain took to reach the current node
Hofmann_2020 and Browning_2006 are analogous art because they are from the same field of endeavor called Markov Chains. Before the effective filing date, it would have been obvious to a person of ordinary skill in the art to combine Hofmann_2020 and Browning_2006. The rationale for doing so would have been that Hofmann_2020 teaches to use Hidden Markov Chains. Browning_2006 teaches that Hidden Markov Chains do not “allow memory (i.e., choice of variables on which a conditional probability is based) to depend on the state of some of the variables” (page 904) and that VLMC have some important advantages of HMMs and that VLMC’s take advantage of linkage disequilibrium to create a parsimonious/efficient model. Therefore, it would have been obvious to combine Hofmann_2020 and Browning_2006 for the benefit of having a more efficient model that takes advantage of linkage disequilibrium to obtain the invention as specified in the claims.
Claim 16. The limitations of claim 16 are substantially the same as those of claim 1 and are rejected due to the same reasons as outlined above for claim 1. The further limitations of claim 1 are also made obvious by the combination of Hofmann_2020 and Polze_2011.
Claim 2. Polze_2011 makes obvious “wherein the computer system comprises a first computer system, the given future state corresponds to a failure of the first computer system, and the method further comprises, responsive to the alert, failing over the first computer system to a second computer system” (abstract: “… we propose an architectural blueprint for managing server system dependability in a pro-active fashion, in order to keep service-level promises for response times and availability even with increasing hardware failure rates. We introduce the concept of anticipatory virtual machine migration that proactively moves computation away from faulty or suspicious machines. The migration decision is based on health indicators… live migration techniques can be triggered in order to move computation to healthy machines even before a failure…”; page 234 section 1: “… live migration capabilities for virtual machines running the actual service. In such frameworks, a virtual machine (VM) can be migrated during runtime from one physical machine to another physical machine… See Figure 1…”; page 235: “… we propose as alternative a recovery scheme called proactive migration, were the virtual machine is migrated to another physical host with respect to timing constraints – before a failure occurs…” page 235 section II Approach: “… health-indicator… migration controller finally performs the migration…”; Figure 2; page 242 Conclusion: “… prediction of failures at various machine levels and the combination of these indicators into a global health indicator… prediction approach can provide enough lead time for the failure avoidance mechanism (such as virtual machine migration), the overall pro-active fault tolerance approach has a good chance to act as supplement to reactive fault tolerance schemes in use today…”).
Claim 3. Polze_2011 makes obvious “wherein failing over the first computer system to the second computer system comprises at least one of migrating a virtual machine of the first computer system to the second computer system, or transferring a workload assigned to the first computer system to the second computer system” (abstract: “… we propose an architectural blueprint for managing server system dependability in a pro-active fashion, in order to keep service-level promises for response times and availability even with increasing hardware failure rates. We introduce the concept of anticipatory virtual machine migration that proactively moves computation away from faulty or suspicious machines. The migration decision is based on health indicators… live migration techniques can be triggered in order to move computation to healthy machines even before a failure…”; page 234 section 1: “… live migration capabilities for virtual machines running the actual service. In such frameworks, a virtual machine (VM) can be migrated during runtime from one physical machine to another physical machine… See Figure 1…”; page 235: “… we propose as alternative a recovery scheme called proactive migration, were the virtual machine is migrated to another physical host with respect to timing constraints – before a failure occurs…” page 235 section II Approach: “… health-indicator… migration controller finally performs the migration…”; Figure 2; page 242 Conclusion: “… prediction of failures at various machine levels and the combination of these indicators into a global health indicator… prediction approach can provide enough lead time for the failure avoidance mechanism (such as virtual machine migration), the overall pro-active fault tolerance approach has a good chance to act as supplement to reactive fault tolerance schemes in use today…”).
Claim 6. Hofmann_2020 makes obvious “wherein: Determining the likelihood that the computer system will transition to the next state comprises identifying a path of the directed graph based on the history page 456: “… the probability distribution of these non-terminal states and the transition probabilities between states are learned from non-stationary time-series data gathered as historic data…”; page 466 section 2 A Brief Introduction to the Hidden Markov Model: “a hidden Markov model is a generative graph model that represents probability distributions over sequence of observations [7]. It involves two interconnected models. The state model consists of a discrete-time, discrete-state first-order Markov chain zt ∈ {1,...,N} that transitions according to p(zt|zt−1), while the observation model is governed by p(xt|zt), where xt are the observations. The corresponding joint distribution of a sequence of states and observations can be factored as:
PNG
media_image1.png
102
665
media_image1.png
Greyscale
Therefore, to fully define this probability distribution, we need to specify a probability distribution over the initial state p(z1), a N × N state transition matrix defining all transition probabilities p(zt|zt−1), and the emission probability model p(xt|zt). EXAMINER NOTE: transition probabilities/probability distributions are the claimed likelihood of a transition to the next state. These probabilities, for each edge, of the graph model are contained in the transition matrix. These probabilities are obtained by observing/identifying the paths taken in the graph historically. ) and the path comprises the given model state, the first transitions and a second model state of the model states corresponding to the given future state” (abstract: “… Markov states…”; page 464 introduction: “… learn complex stochastic degradation patterns from data by introducing a terminal state that represents the failure state of the asset, whereas other states represent health-states of the asset as it progresses towards failure…”; Fig. 5 is an illustration of state transition graph where transitions are made from a given model state to a second model state where the second model state is a future model state.
Browning_2006 makes obvious “history identifiers” (Figure 1 illustrates blocks and each block has multiple paths each path is defined by different numerical identifiers.).
Claim 8. Browning_2006 makes obvious: “Determining the likelihood that the computer system will transition to the given future state comprises identifying a path of the directed graph based on the history identifier and a probability associated with the path; The path comprises the given model state, the first transition and a second model state of the model states corresponding to the next state; The path comprises at least one transition of the plurality of transitions from the given model state to the second model state; and the probability associated with the path corresponds to the probability or probabilities associated with the at least one transition”
( Page 905:
PNG
media_image4.png
332
571
media_image4.png
Greyscale
Claim 9. Hofmann_2020 makes obvious “wherein the given future state corresponds to a failure of the computer system” (page 464 Introduction: “… introducing a terminal state that represents the failure state of the asset…”).
Claim 10. Hofmann_2020 makes obvious “further comprising: Detecting a new state of the computer system that does not correspond to a model state of the model states; Responsive to the detection, updating the data to add another model state to the directed graph correspond to the new state; and
Responsive to the detection, further updating the data to add a transition identifier associated with a transition to the other model state” (page 465: “… adopting a Bayesian approach allows for starting with an estimate of the probability that can be subsequently refined by observation, as more sensor data is revealed in real time. In particular, our approach allows task specific knowledge to be included into the model. For example, the number of health-states, the number of mixtures of (topics, or archetypes) and the structure of the transition matrix may be modeled explicitly using engineering knowledge of practitiioners. Typically, the data structure for LSTM is fixed. E.g., matrix, or time-series, while MMMM is more flexible… MMM can also worth with missing data out of the box, using expert knowledge as priors…” EXAMINER NOTE: the above indicates that the Markov states as represented in the matrix may be incomplete due to, for example, using expert knowledge as priors but through the acquisition of observations by, for example, real time sensor data refinement is performed that completes the Markov states.)
Claim 19. Polze_2011 makes obvious “wherein the management controller comprises a chassis management controller, and the plurality of computer systems comprise computer systems of a rack-based computing system” (Figure 1: “server blade”; page 234: “… CPUs in commodity servers such as blade centers…” EXAMINER NOTE: a blade server inherently implies a chassis because a “blade” is not a standalone computer. It is an thin modular circuit board containing processors, memory, and network adaptors but it lacks its own power supply, cooling fans, and external ports. Accordingly, every blade server must be installed into a specialized blade chassis. Additionally, the chassis, which holds the blades, is typically mounted into a server rack.)
Claim 20. Polze_2011 makes obvious “wherein: The directed graph corresponds to a Markov chain; and the directed graph is associated with one or a memory subsystem of the first computer system, a processor subsystem of the first computer system, or an expansion card subsystem of the first computer system” (Figure 2: “Hardware: MCA, CPU Profiling, IPMI” is illustrated as input into the health indicator. Figure 3 illustrates “Hardware Failure Predictors”).
Claims 4, 5 are rejected under 35 U.S.C. 103 as being unpatentable over Hofmann_2020 in view of Polze_2011 in view of Browning_2006 view of Keifer_2020 (Comparing Labelled Markov Decision Processes, 40th IARCS Annual Conference on Foundations of Software Technology and Theoretical Computer Science (FSTTCS 2020)).
Claim 4. Keifer_2020 makes obvious “wherein the first transition identifiers comprise prime numbers, and the directed graph corresponds to a Markov chain” (page 49:9 Lemma 8 “… the witness strategy might not be MD, we compute a set of prime numbers wthat can be used to form the weights of the actions. The prime numbers are used to rule out certain “accidental” bisimulations…”; Lemma 9, 10, 11 EXAMINER NOTE: The paper outlines that, In the domain of probabilistic bisimulation and equivalence checking for Markov chains, algorithms map actions and transitions using numerical weights. When generating transition weights or assigning transition identifiers, prime numbers are used to rule out certain "accidental" bisimulations. Because prime numbers are only divisible by \(1\) and themselves, utilizing them for transition identifiers prevents overlapping divisors in the algebraic calculations of transition probabilities. This coprimality guarantees that linear combinations of transitions remain mathematically unique, allowing the algorithm to correctly identify and distinguish between different paths in the Markov chain without introducing false equivalences).
Browning_2006 and Keifer_2020 are analogous art because they are from the same field of endeavor called Markov chains. Before the effective filing date, it would have been obvious to a person of ordinary skill in the art to combine Browning_2006 and Keifer_2020. The rationale for doing so would have been that In the context of variable length Markov chains bisimulation be used to identify sections of the Markov chain which are conditionally independent from other sections of the chain because probabilistic bisimulation can be used to identify sections of a Variable Length Markov Chain (VLMC) that are conditionally independent of other sections. In a VLMC, conditional independence manifests as variable memory depth: the future depends only on a specific historical context, making it conditionally independent of older history. Probabilistic bisimulation maps history-based states into equivalence classes based on identical future transition distributions, fundamentally finding and grouping these conditionally independent sections. Therefore, it would have been obvious to combine Hofmann_2020 and Keifer_2020 for the benefit of having a more robust Markov decision process by making use of multilocus data and identifying linkage disequilibrium between failure markers to create a parsimonious model to obtain the invention as specified in the claims.
Claim 5. Keifer_2020 makes obvious “wherein applying the first arithmetic operator comprises determining a product of the prime numbers” (page 49:9 Lemma 8, 9, 10, 11 EXAMINER NOTE: Using prime numbers as unique transition identifiers in Markov chain operations—such as in bisimulation and probabilistic equivalence checking—often involves applying arithmetic operators like multiplication to uniquely combine those transitions. Applying arithmetic operators by computing the product of prime numbers creates a mathematical property known as unique factorization. Because every number has a unique prime factorization, multiplying the prime identifiers of parallel or branching transitions allows systems to perfectly compress branching information without losing track of which specific branches or paths were taken.
This approach serves a few core purposes in formal verification:
Unique Identification: Combining transitions via multiplication (e.g., P_1 X P_2 X P_3 ) guarantees a single, distinct composite identifier. You can always uniquely decompose the product to determine exactly which transitions went into making it.
Bisimulation Partitioning: In probabilistic bisimulation, algorithms must check whether states have equivalent probabilities of transitioning into equivalent equivalence classes. Using prime products provides a deterministic way to aggregate transition probabilities and structural identifiers.
Avoiding Collisions: Using addition or hash functions can result in collisions (where two different combinations sum to the same number). Multiplication by primes avoids this)
Claims 7, 17, 18 are rejected under 35 U.S.C. 103 as being unpatentable over Hofmann_2020 in view of Polze_2011 in view of Browning_2006 in view of Giscard_2015 (Walk-Sums, Continued Fractions and Unique Factorisation on Digraphs, Jan 2015).
Claim 7. Hofmann_2020 makes obvious wherein:
The path is associated with a second number; and identifying the path comprises determining whether the path is consistent with the history of the computer system based on and the second number” (page 456: “… the probability distribution of these non-terminal states and the transition probabilities between states are learned from non-stationary time-series data gathered as historic data…”; page 466 section 2 A Brief Introduction to the Hidden Markov Model: “a hidden Markov model is a generative graph model that represents probability distributions over sequence of observations [7]. It involves two interconnected models. The state model consists of a discrete-time, discrete-state first-order Markov chain zt ∈ {1,...,N} that transitions according to p(zt|zt−1), while the observation model is governed by p(xt|zt), where xt are the observations. The corresponding joint distribution of a sequence of states and observations can be factored as:
PNG
media_image1.png
102
665
media_image1.png
Greyscale
Therefore, to fully define this probability distribution, we need to specify a probability distribution over the initial state p(z1), a N × N state transition matrix defining all transition probabilities p(zt|zt−1), and the emission probability model p(xt|zt). EXAMINER NOTE: the state transition probabilities are the weights on the edges of the directed graph for the Markov chain. These are the claimed “second number” on each edge.)
Giscard_2015, however, makes obvious “the history identifier comprises a first number” and “the first number” (abstract: “… our results are based on an equivalent to the fundamental theorem of arithmetic: we demonstrate that arbitrary walks on G factorize uniquely into nesting products… where nesting is a product operation… paths… are the prime elements of the set of all walks on G equipped with the nesting product…” page 2 section 1.1: “walks on graphs are pervasive mathematical objects that appear in a wide range of fields from mathematics and physics to engineering, biology and social scoence…”; page 2 section 1.2: “in this work we consider walks on (possibly weighted) directed graphs as mathematical entities in their own right. We demonstrate that… any walk can be uniquely factorized into a product of prime walks… this continued fraction is the prime representation of the walk series…”; page 3: “… the edges of G. Consequently, the factorization of a walk W on G into concatenations of prime walks…”; page 3 section 2.1: a directed graph or digraph is a set of vertices connected by directed edges, also known as arrows…”; Fig. 1) EXAMINER NOTE: The above teaches that the edges of a weighted directed graph can be characterized by prime numbers and the walk along the graph can be uniquely characterized by the product of the prime number from each edge traversed during the walk. Because this is done on a weighted directed graph each edge has two numbers: the weight (i.e., claimed second number and the prime number (I.e., claimed first number: edge history ID)).
Hofmann_2020 and Giscard_2015 are analogous art because they are from the same field of endeavor called directed graphs. Before the effective filing date, it would have been obvious to a person of ordinary skill in the art to combine Hofmann_2020 and Giscard_2015. The rationale for doing so would have been that Hofmann_2020 teaches to have a directed graph called a Markov chain where the edges are identified by weights (e.g., probabilities) and to use the Markov chains to dynamically optimize maintenance policy for failing assets. Giscard_2015 teaches that according to the fundamental theorem of arithmetic, that weighted directed graphs can further be characterized according to prime numbers which, when multiplied, provide a unique indication of the historic path taken through the directed graph. Giscard_2015 also teaches that the use of prime factorization allows the identification of walks into, for example, families of prime factors. Therefore, it would have been obvious to combine Hofmann_2020 and Giscard_2015 for the benefit of, for example, identifying unique failure paths that may require unique or different dynamic optimization policies to obtain the invention as specified in the claims.
Claim 17. Giscard_2015 makes obvious “wherein the first transition identifiers comprises first prime numbers; and the management controller to further multiply the prime numbers together to determine a second prime number representing the history identifier” (abstract: “… our results are based on an equivalent to the fundamental theorem of arithmetic: we demonstrate that arbitrary walks on G factorize uniquely into nesting products… where nesting is a product operation… paths… are the prime elements of the set of all walks on G equipped with the nesting product…” page 2 section 1.1: “walks on graphs are pervasive mathematical objects that appear in a wide range of fields from mathematics and physics to engineering, biology and social scoence…”; page 2 section 1.2: “in this work we consider walks on (possibly weighted) directed graphs as mathematical entities in their own right. We demonstrate that… any walk can be uniquely factorized into a product of prime walks… this continued fraction is the prime representation of the walk series…”; page 3: “… the edges of G. Consequently, the factorization of a walk W on G into concatenations of prime walks…”; page 3 section 2.1: a directed graph or digraph is a set of vertices connected by directed edges, also known as arrows…”; Fig. 1)
Hofmann_2020 and Giscard_2015 are analogous art because they are from the same field of endeavor called directed graphs. Before the effective filing date, it would have been obvious to a person of ordinary skill in the art to combine Hofmann_2020 and Giscard_2015. The rationale for doing so would have been that Hofmann_2020 teaches to have a directed graph called a Markov chain where the edges are identified by weights (e.g., probabilities) and to use the Markov chains to dynamically optimize maintenance policy for failing assets. Giscard_2015 teaches that according to the fundamental theorem of arithmetic, that weighted directed graphs can further be characterized according to prime numbers which, when multiplied, provide a unique indication of the historic path taken through the directed graph. Giscard_2015 also teaches that the use of prime factorization allows the identification of walks into, for example, families of prime factors. Therefore, it would have been obvious to combine Hofmann_2020 and Giscard_2015 for the benefit of, for example, identifying unique failure paths that may require unique or different dynamic optimization policies to obtain the invention as specified in the claims.
Claim 18. Giscard_2015 makes obvious “wherein the management controller to further: identify a path of the directed graph based on the second prime number, wherein the path comprises the given model state, the first transitions and a second model state of the model state corresponding to the second model state, and the path is associated with a third number, wherein identifying the path comprises: determining whether the second prime number is a factor of the third prime number; and selecting the path from a plurality of candidate paths responsive to a result of the determination of whether the second prime number is a factor of the third prime number” (abstract: “… our results are based on an equivalent to the fundamental theorem of arithmetic: we demonstrate that arbitrary walks on G factorize uniquely into nesting products… where nesting is a product operation… paths… are the prime elements of the set of all walks on G equipped with the nesting product…” page 2 section 1.1: “walks on graphs are pervasive mathematical objects that appear in a wide range of fields from mathematics and physics to engineering, biology and social scoence…”; page 2 section 1.2: “in this work we consider walks on (possibly weighted) directed graphs as mathematical entities in their own right. We demonstrate that… any walk can be uniquely factorized into a product of prime walks… this continued fraction is the prime representation of the walk series…”; page 3: “… the edges of G. Consequently, the factorization of a walk W on G into concatenations of prime walks…”; page 3 section 2.1: a directed graph or digraph is a set of vertices connected by directed edges, also known as arrows…”; Fig. 1 EXAMINER NOTE: using prime numbers to identify paths through a directed graph exploits the Fundamental Theorem of Arithmetic (the rule that every integer has a unique prime factorization). By assigning unique prime numbers to the edges of a graph, any path’s sequence of edges can be collapsed into a single, unique composite number. Because the product is derived exclusively from primes, prime factorization guarantees that only one combination of edges could ever produce that number).
Claims 11, 14, 15 are rejected under 35 U.S.C. 103 as being unpatentable over Hofmann_2020 in view of Polze_2011 in view of Browning_2006 in view of NGO_2013 (WO 2013/109702 A1).
Claim 11. The limitations of claim 11 are substantially the same as those of claim 1 and are therefore rejected due to the same reasons as outlined above for claim 1. Additionally, NGO_2013 makes obvious the further limitations of “a non-transitory storage medium storing machine-readable instructions that, when executed by a machine, cause the machine to: “ (par 21: “… a non-transitory computer-readable storage medium tangibly embodying a program of instructions executable by a server within a server cluster…”)
Hofmann_2020 and NGO_2013 are analogous art because they are from the same field of endeavor called asset monitoring. Before the effective filing date, it would have been obvious to a person of ordinary skill in the art to combine Hofmann_2020 and NGO_2013. The rationale for doing so would have been that Hofmann_2020 teaches to monitor assets for failure. NGO_2013 teaches to monitor assets using computer executable programs in memory. Therefore, it would have been obvious to combine Hofmann_2020 and NGO_2013 for the benefit of having a computer to execute asset monitoring to obtain the invention as specified in the claims.
Claim 14. Hofmann_2020 makes obvious “wherein the given future computer state comprises a failure state (page 464 Introduction: “… introducing a terminal state that represents the failure state of the asset…”) and Polze_2011 makes obvious “and the remedial action comprises at least one of migrating a virtual machine from the computer system to another computer system, or migrating a workload to the computer system to another computer system” (abstract: “… we propose an architectural blueprint for managing server system dependability in a pro-active fashion, in order to keep service-level promises for response times and availability even with increasing hardware failure rates. We introduce the concept of anticipatory virtual machine migration that proactively moves computation away from faulty or suspicious machines. The migration decision is based on health indicators… live migration techniques can be triggered in order to move computation to healthy machines even before a failure…”; page 234 section 1: “… live migration capabilities for virtual machines running the actual service. In such frameworks, a virtual machine (VM) can be migrated during runtime from one physical machine to another physical machine… See Figure 1…”; page 235: “… we propose as alternative a recovery scheme called proactive migration, were the virtual machine is migrated to another physical host with respect to timing constraints – before a failure occurs…” page 235 section II Approach: “… health-indicator… migration controller finally performs the migration…”; Figure 2; page 242 Conclusion: “… prediction of failures at various machine levels and the combination of these indicators into a global health indicator… prediction approach can provide enough lead time for the failure avoidance mechanism (such as virtual machine migration), the overall pro-active fault tolerance approach has a good chance to act as supplement to reactive fault tolerance schemes in use today…”).
Claim 15. Hofmann_2020 makes obvious “wherein the directed graph comprises a Markov Chain” (page 466: “… a hidden Markov model is a generative graph model that represents probability distributions over sequences of observations [7]. It involves two interconnected models. The state model consists of… Markov chain…”).
Claims 12, 13 are rejected under 35 U.S.C. 103 as being unpatentable over Hofmann_2020 in view of Polze_2011 in view of Browning_2006 in view of NGO_2013 in view of Giscard_2015
Claim 12. Giscard_2015 makes obvious “wherein the instructions, when executed by the machine, further cause the machine to: identify first transitions of the plurality of transitions corresponding to the transition history (page 456: “… the probability distribution of these non-terminal states and the transition probabilities between states are learned from non-stationary time-series data gathered as historic data…”; page 466 section 2 A Brief Introduction to the Hidden Markov Model: “a hidden Markov model is a generative graph model that represents probability distributions over sequence of observations [7]. It involves two interconnected models. The state model consists of a discrete-time, discrete-state first-order Markov chain zt ∈ {1,...,N} that transitions according to p(zt|zt−1), while the observation model is governed by p(xt|zt), where xt are the observations. The corresponding joint distribution of a sequence of states and observations can be factored as:
PNG
media_image1.png
102
665
media_image1.png
Greyscale
Therefore, to fully define this probability distribution, we need to specify a probability distribution over the initial state p(z1), a N × N state transition matrix defining all transition probabilities p(zt|zt−1), and the emission probability model p(xt|zt).)
Giscard_2015, however, makes obvious and multiply the first prime numbers associated with the first transitions together to determine the second number (abstract: “… our results are based on an equivalent to the fundamental theorem of arithmetic: we demonstrate that arbitrary walks on G factorize uniquely into nesting products… where nesting is a product operation… paths… are the prime elements of the set of all walks on G equipped with the nesting product…” page 2 section 1.1: “walks on graphs are pervasive mathematical objects that appear in a wide range of fields from mathematics and physics to engineering, biology and social scoence…”; page 2 section 1.2: “in this work we consider walks on (possibly weighted) directed graphs as mathematical entities in their own right. We demonstrate that… any walk can be uniquely factorized into a product of prime walks… this continued fraction is the prime representation of the walk series…”; page 3: “… the edges of G. Consequently, the factorization of a walk W on G into concatenations of prime walks…”; page 3 section 2.1: a directed graph or digraph is a set of vertices connected by directed edges, also known as arrows…”; Fig. 1) EXAMINER NOTE: The above teaches that the edges of a weighted directed graph can be characterized by prime numbers and the walk along the graph can be uniquely characterized by the product of the prime number from each edge traversed during the walk. )
Hofmann_2020 and Giscard_2015 are analogous art because they are from the same field of endeavor called directed graphs. Before the effective filing date, it would have been obvious to a person of ordinary skill in the art to combine Hofmann_2020 and Giscard_2015. The rationale for doing so would have been that Hofmann_2020 teaches to have a directed graph called a Markov chain where the edges are identified by weights (e.g., probabilities) and to use the Markov chains to dynamically optimize maintenance policy for failing assets. Giscard_2015 teaches that according to the fundamental theorem of arithmetic, that weighted directed graphs can further be characterized according to prime numbers which, when multiplied, provide a unique indication of the historic path taken through the directed graph. Giscard_2015 also teaches that the use of prime factorization allows the identification of walks into, for example, families of prime factors. Therefore, it would have been obvious to combine Hofmann_2020 and Giscard_2015 for the benefit of, for example, identifying unique failure paths that may require unique or different dynamic optimization policies to obtain the invention as specified in the claims.
Claim 13. Giscard_2015 makes obvious “wherein the instructions, when executed by the machine, further cause the machine to: Access third data representing candidate paths of the directed graph, wherein each candidate path of the candidate paths comprises a second model state corresponding to the given future computer system state, and each candidate path of the candidate paths has an associated third prime number; and Select a given candidate path of the candidate paths based on the third prim numbers and the second prime number, wherein the given candidate path comprises at least one transition of the plurality of transitions from the given model state to the second model state (Fig. 1 illustrates nesting products. There are more than three edges.) a Hofmann_2020 makes obvious “and determine a likelihood of the computer system transitioning to the given future computer system state based on the probability or probabilities associated with the at least one transition” ( page 466 section 2 a brief introduction to the hidden Markov model).
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BRIAN S COOK whose telephone number is (571)272-4276. The examiner can normally be reached 8:00 AM - 5:00 PM.
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, Emerson Puente can be reached at 571-272-3652. 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.
/BRIAN S COOK/Primary Examiner, Art Unit 2187
Prosecution Insights