{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp09_archetype_breadth150_20260804","research_id":"eoa_inverse_innovation_exp09_light_prior_art_20260804","cell_id":"invariant_mode_decomposition_design__computer_science","search_lanes":{"direct_problem_and_intervention":{"queries":["microservices retry storm coupled dynamics eigenvalue modal analysis telemetry","dynamic mode decomposition microservice performance anomaly detection","state space model microservices retry queue latency control eigenvalues","\"modal analysis\" \"retry\" distributed software"],"source_ids":["SRC1","SRC4"],"no_result_note":null},"synonyms_and_historical_terms":{"queries":["metastable failures distributed systems retry storm paper queue saturation","site:research.google/pubs cascading failure retries overload distributed systems","eigenvalue stability analysis distributed system retry control"],"source_ids":["SRC1","SRC2"],"no_result_note":null},"products_practices_and_standards":{"queries":["site:sre.google cascading failures retries overload","Retry Storm antipattern Azure Architecture Center","site:aws.amazon.com/builders-library retries backoff overload"],"source_ids":["SRC1"],"no_result_note":null},"component_combination":{"queries":["Dynamic mode decomposition with control Proctor Brunton Kutz 2016 paper","DMDc sensitivity control eigenmodes system identification intervention","dynamic mode decomposition for interconnected control systems","\"Predicting Instability Waves in Distributed Software\""],"source_ids":["SRC3","SRC4"],"no_result_note":null}},"sources":[{"source_id":"SRC1","title":"Addressing Cascading Failures","publisher":"Google Site Reliability Engineering","url":"https://sre.google/sre-book/addressing-cascading-failures/","source_type":"OFFICIAL_GUIDANCE","claims_supported":["Retries can destabilize an overloaded service, grow recursively, consume resources, and sustain overload after the initiating load subsides.","Retry graphs may be mistaken for symptoms rather than compounding causes.","Operational mitigations include randomized backoff, bounded retries, retry budgets, holistic multi-layer analysis, testing, load reduction, and cautious canaries."]},{"source_id":"SRC2","title":"Metastable Failures in the Wild","publisher":"USENIX Association","url":"https://www.usenix.org/system/files/osdi22-huang-lexiang.pdf","source_type":"PRIMARY_RESEARCH","claims_supported":["The study catalogued public cloud and distributed-system incidents involving retries, cascading overload, queue growth, timeouts, and contention.","Retry policy was the most common sustaining effect, affecting more than half of the studied incidents.","Recovery commonly required coordinated load shedding, throttling, policy changes, rollback, or capacity intervention."]},{"source_id":"SRC3","title":"Dynamic Mode Decomposition for Interconnected Control Systems","publisher":"arXiv","url":"https://arxiv.org/abs/1709.02883","source_type":"PRIMARY_RESEARCH","claims_supported":["Network DMDc estimates data-driven linear dynamics for interconnected controlled subsystems while exploiting known network structure.","Eigenvectors or modes expose prominent system behavior, while eigenvalues encode growth, decay, and oscillation.","The method incorporates measured control inputs and dynamical relationships between connected components."]},{"source_id":"SRC4","title":"Predicting Instability Waves in Distributed Software: A Finite-Element Analogy and the Spectral Resiliency Index","publisher":"ResearchGate (author-uploaded preprint)","url":"https://www.researchgate.net/publication/404397422_Predicting_Instability_Waves_in_Distributed_Software_A_Finite-Element_Analogy_and_the_Spectral_Resiliency_Index","source_type":"PRIMARY_RESEARCH","claims_supported":["The preprint proposes telemetry-compatible eigenvalue and modal analysis for self-amplifying instabilities propagating across distributed-software services.","It links modal growth to resource exhaustion and discusses damping, isolation, capacity headroom, and observability hardening as defensive implications.","It is early-stage preprint work and does not establish the proposal's retry-specific control-selection and replay workflow."]}],"problem_evidence":{"status":"PARTLY_SUPPORTED","finding":"Retry-driven, self-sustaining overload is clearly visible: Google describes recursively growing retries, resource exhaustion, latency, and diagnostic ambiguity, while the OSDI study found retries sustaining more than half of its collected metastable incidents and documented queue growth and cross-service outages. The narrower assertion that coordinate-level dashboards remain tolerable while a reproducible latent modal direction is already growing was not directly demonstrated by the retained evidence.","source_ids":["SRC1","SRC2","SRC4"]},"closest_prior_art":[{"name":"Predicting Instability Waves in Distributed Software / Spectral Resiliency Index","source_ids":["SRC4"],"overlap":"Applies eigenvalue spectra and modal analysis to telemetry-compatible distributed-software models to reveal self-amplifying cross-service instability and motivate damping-oriented mitigation.","remaining_difference":"It uses a finite-element-inspired resilience model and broad defensive implications; the retained description does not fit a one-step transition from retry, queue, saturation, rejection, and latency telemetry, estimate control sensitivity for retry budgets and timeout relationships, or require replay, reconstruction-residual, spectral-separation, coupling, and drift gates."},{"name":"Network Dynamic Mode Decomposition with Control","source_ids":["SRC3"],"overlap":"Provides the central mathematical pattern: learn interconnected state-and-control dynamics from data, extract modes and modal gains, and represent how connected subsystems and inputs influence evolution.","remaining_difference":"It is a general identification method rather than a microservice retry-storm operating procedure; it does not supply the proposed state definition, incident relevance rule, reversible SRE control sweep, replay comparison, or production authority contract."},{"name":"Metastable-failure analysis and established retry-storm response","source_ids":["SRC1","SRC2"],"overlap":"Recognizes retry amplification as a coupled sustaining mechanism and recommends coordinated controls such as retry budgets, backoff, throttling, load shedding, deadline changes, testing, rollback, and cautious canaries.","remaining_difference":"These sources diagnose mechanisms and prescribe practices without estimating a cross-service transition operator, selecting controls by their projected leverage on a separated mode, or retaining mode-specific residual and drift invalidation criteria."}],"prior_art_disposition":"ADJACENT_PRIOR_ART","contrastive_claim_remaining":"For a declared microservice subgraph and stable regime, a held-out-predictive local transition model can expose a separated, reproducible retry-saturation mode and use the measured projection of several reversible service controls onto that mode to select one coordinated configuration that reduces its replayed gain without worsening completion, latency, residuals, or other safety-relevant modes. The retained sources establish the failure, the modal ingredients, and nearby distributed-software spectral analysis, but not this retry-specific, guarded selection-and-replay workflow.","contrastive_claim_falsifier":"The contrast disappears if an existing system or publication is found that already fits cross-service queue, retry, saturation, rejection, and latency transitions; identifies persistent or growing modes; ranks retry, backoff, admission, concurrency, and timeout changes by modal sensitivity; and validates the selected coordinated configuration under replay with residual, spectral-gap, cross-mode, and drift stop rules. It is also empirically falsified if the coupled model adds no held-out predictive value over coordinate-wise or sparse service-local baselines, or its selected configuration does not safely reduce target-mode gain in replay.","gates":{"adequate_source_search":{"status":"PASS","rationale":"Four distinct search lanes covered direct phrasing, metastability and historical terminology, operational guidance and products, and component combinations. Exactly four opened sources from Google SRE, USENIX, arXiv, and an author-uploaded preprint were retained, including official guidance and primary research.","source_ids":["SRC1","SRC2","SRC3","SRC4"]},"supported_problem":{"status":"PASS","rationale":"The broad operational problem is supported by official guidance and multi-incident primary research, although the specific pre-threshold latent-mode signature remains unverified, yielding PARTLY_SUPPORTED problem evidence.","source_ids":["SRC1","SRC2","SRC4"]},"distinct_testable_claim":{"status":"PASS","rationale":"Despite adjacent collision on distributed-software modal analysis and Network DMDc, the remaining claim is narrowly testable against explicit local baselines, mode reproducibility and separation criteria, control projections, and replay outcomes.","source_ids":["SRC1","SRC2","SRC3","SRC4"]},"bounded_next_test":{"status":"PASS","rationale":"The proposed first test is bounded to captured telemetry from one subgraph and one non-production replay, with predeclared variables, comparison models, residual and spectral rules, measurable outcomes, and no production change.","source_ids":["SRC1","SRC2","SRC3"]},"no_obvious_safety_or_authority_stop":{"status":"PASS","rationale":"Read-only fitting and authorized non-production replay present no obvious authority stop. Production changes remain separately approved, reversible, canaried, and subject to halt criteria; this aligns with official cautions that mitigation changes can amplify cascades and should be tested and introduced slowly.","source_ids":["SRC1","SRC2"]}},"screen_survival":true,"world_novelty_boundary":"This bounded public-web screen supports only an adjacent-prior-art disposition and a researchable remaining contrast. It cannot establish world novelty, patentability, freedom to operate, market size, expert acceptance, production effectiveness, or realized value; unindexed patents, internal SRE systems, non-English materials, inaccessible publications, and later disclosures may contain closer art."}