{
  "schema_version": "0.6.0",
  "paper_id": "altoom-out-of-memory-root-cause",
  "page_url": "https://naser.github.io/research-publications/papers/altoom-out-of-memory-root-cause/",
  "title": "AltOOM: A Data-driven Out of Memory Root Cause Identification Strategy",
  "title_variants": [],
  "authors": [
    "Pranjal Chakraborty",
    "Naser Ezzati-Jivan",
    "Vahid Azhari",
    "François Tetreault"
  ],
  "author_details": [
    {
      "name": "Pranjal Chakraborty",
      "orcid": null,
      "profile_url": "https://dblp.org/pid/367/0073.html"
    },
    {
      "name": "Naser Ezzati-Jivan",
      "orcid": "https://orcid.org/0000-0003-1435-6297",
      "profile_url": "https://naser.github.io/"
    },
    {
      "name": "Vahid Azhari",
      "orcid": null,
      "profile_url": null
    },
    {
      "name": "François Tetreault",
      "orcid": null,
      "profile_url": "https://dblp.org/pid/288/0256.html"
    }
  ],
  "publication": {
    "year": 2023,
    "venue": "IEEE International Conference on Big Data",
    "type": "conference paper",
    "publication_date": "2023-12-15",
    "online_date": null,
    "print_date": "2023-12-15",
    "volume": null,
    "issue": null,
    "pages": "1637-1646",
    "article_number": null,
    "publisher": "IEEE",
    "issn": [],
    "isbn": [],
    "crossref_type": "proceedings-article"
  },
  "publication_type": "research paper",
  "status": "published_metadata_record",
  "canonical_source_url": "https://doi.org/10.1109/bigdata59044.2023.10386937",
  "source_record_id": "altoom-a-data-driven-out-of-memory-root-cause-identification-strategy-ecff57f6a1",
  "identifiers": {
    "doi": "10.1109/bigdata59044.2023.10386937"
  },
  "abstract": null,
  "abstract_source": "AltOOM published-paper PDF reviewed on 2026-08-06; abstract not reproduced because reuse permission for the publisher version was not established.",
  "abstract_available": false,
  "scholar_eligibility": {
    "eligible": false,
    "basis": "not-eligible",
    "note": "The page is a discovery record; it does not claim Google Scholar article-host eligibility."
  },
  "description": "AltOOM combines early memory-pressure forecasting with selective process-level profiling to identify the process most responsible for an impending out-of-memory event.",
  "evidence_level": "metadata-or-abstract-reviewed",
  "evidence": {
    "source_basis": "metadata-only",
    "coverage": "bibliographic metadata",
    "summary_origin": "AI-assisted catalog editorial summary",
    "review_status": "catalog-reviewed; paper-author approval pending",
    "verified_on": "2026-08-09",
    "sources": [
      {
        "note": "DBLP/DOI bibliographic record for 10.1109/bigdata59044.2023.10386937"
      },
      {
        "note": "author identity matched to Naser Ezzati-Jivan in the local research catalog"
      },
      {
        "note": "AltOOM PDF pp. 1, 4-10: perf/sar collection, 34 metrics, 85% threshold, prediction models, burst ranking, evaluation tables, limitations, and future work"
      },
      {
        "note": "local AltOOM PDF hash verified in pdf-evidence/extraction-manifest.json"
      }
    ]
  },
  "summary": {
    "core_contribution": "AltOOM combines early memory-pressure forecasting with selective process-level profiling to identify the process most responsible for an impending out-of-memory event.",
    "problem": "Linux's reactive OOM killer can terminate a high-memory process even when a lower-memory process is the one whose memory usage is growing; resource-constrained systems need earlier warning and lower-overhead process attribution.",
    "method": "AltOOM samples 34 system-level signals at 0.5-second intervals - 28 virtual-memory statistics, three memory-related system calls (brk, sbrk, and mmap), and three kernel events (kmalloc, mm_page_alloc, and vmscan) - using perf and sar. It labels pressure at %memused >= 85%, compares SVM, vanilla DNN, and bidirectional-LSTM predictors, filters 34 features to 15, then uses burst-collected process-level allocation signals and moving-average growth ranking after an alert.",
    "findings": "In the reported evaluation, the feature-filtered DNN reaches 0.82 accuracy for (n,k)=(4,3) and 0.81 for (3,3); the abstract reports 85% memory-pressure forecasting accuracy. AltOOM process identification reaches 0.56-0.83 as the burst count increases from 3 to 7, versus 0.42 for the Linux OOM killer. A Firefox PDF-preview case reports 0.83 and 0.79 forecasting accuracy for (4,3) and (3,3), respectively.",
    "limitations": "The method can miss gradual memory buildup when only three timestamps (1.5 seconds) are observed. Fixed-rate monitoring creates overhead, and the evaluation centers on generated pressure scenarios plus a Firefox case rather than broad production-device coverage.",
    "future_work": "Evaluate adaptive sampling of rates, metrics, and metric groups, and implement actions such as controlling or adjusting the responsible process, terminating it, or restarting the system."
  },
  "tags": [
    "resource-analysis",
    "root-cause-analysis",
    "system-tracing",
    "predictive-monitoring"
  ],
  "keywords": [
    "out-of-memory",
    "OOM diagnosis",
    "data-driven RCA",
    "resource analysis",
    "memory pressure forecasting",
    "process-level profiling",
    "Perf",
    "sar"
  ],
  "versions": [
    {
      "id": "published-version",
      "label": "Published version",
      "relation": "version-of-record",
      "title": "AltOOM: A Data-driven Out of Memory Root Cause Identification Strategy",
      "url": "https://doi.org/10.1109/bigdata59044.2023.10386937",
      "pdf_url": null,
      "status": "published",
      "canonical_for_citation": true
    }
  ],
  "access": {
    "status": "published_metadata_record",
    "note": "The DOI is the canonical citation target; this catalog does not redistribute the publisher PDF.",
    "license": null
  },
  "resources": {
    "code": null,
    "data": null,
    "slides": null,
    "demo": null
  },
  "citation_guidance": {
    "when_to_cite": "Cite this paper when studying proactive memory-pressure forecasting or process-level out-of-memory root-cause identification.",
    "points": [
      "Perf/sar-based collection of VM statistics, memory-related system calls, and kernel events for memory-pressure prediction.",
      "Burst-based process profiling and moving-average ranking for identifying the responsible process.",
      "Comparison with the Linux OOM killer and reported forecasting and process-identification accuracy in simulated and Firefox cases."
    ],
    "canonical_version_id": "published-version"
  },
  "provenance": {
    "metadata_verified_on": "2026-08-09",
    "metadata_source": [
      "DBLP/DOI bibliographic record for 10.1109/bigdata59044.2023.10386937",
      "author identity matched to Naser Ezzati-Jivan in the local research catalog",
      "AltOOM PDF pp. 1, 4-10: perf/sar collection, 34 metrics, 85% threshold, prediction models, burst ranking, evaluation tables, limitations, and future work",
      "local AltOOM PDF hash verified in pdf-evidence/extraction-manifest.json"
    ],
    "summary_written_by": "AI-assisted",
    "summary_verified_by": "metadata/abstract-grounded catalog review; full-text review and author approval pending",
    "linked_preprint_record": null,
    "author_order_note": null
  },
  "batch": {
    "phase": 2,
    "batch_label": "expanded forty-paper release",
    "status": "included_in_expanded_catalog",
    "selected_at": "2026-08-09"
  }
}