[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-82677-en":3,"doc-seo-82677-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},82677,3848291630094,"Emma Wilson","https://eur-avatar.wpscdn.com/davatar_085a072bc5b1113ac321206ff7593b45",8,"Research & Report","First-Principles Infrastructure Modeling for Machine Learning Systems","Machine learning has evolved into critical infrastructure, spanning devices from microcontrollers to massive datacenter fleets, making cross-system reasoning difficult. Empirical profiling demands access to target hardware, while cycle-accurate simulation is too slow, and current analytic tools lack full-stack architectural coverage and runtime correctness checks. MLSYS · IM introduces a first-principles framework that formalizes systems “physics” using a dimensionally strict Python engine with unit integrity and provenance-tracked inputs.","MLSYS · IM  \nFirst-Principles Infrastructure Modeling for Machine Learning Systems  \nVijay Janapa Reddi  \nHarvard University  \n[mlsysbook.ai/mlsysim](mlsysbook.ai/mlsysim)  \narXiv :2607 .02558v1 [ cs .DC] 28 Jun 2026  \nAbstract  \nAs machine learning shifts from laboratory curiosity to critical infrastructure, the systems that sustain it span an extraordinary range, from sub-milliwatt microcontrollers to multi-gigawatt datacenter fleets. Reasoning across this range is hard: empirical profiling requires the target hardware in hand, while cycle-accurate simulation costs hours per configuration, leaving no tool for rapid, full-stack architectural reasoning. We present MLSYS · IM (Machine Learning Systems Infrastructure Modeling), a first-principles analytical framework that formalizes the “physics of systems” into a dimensionallystrict Python engine. MLSYS · IM is built on a demand– supply abstraction that decouples computational demand from silicon supply and environmental context, and it enforces unit integrity at runtime so the silent conversion errors that plague ad-hoc modeling cannot occur. Every input is drawn from a typed, provenancetracked registry, so no number enters an analysis without a documented source. On this engine we codify a taxonomy of 22 “Systems Walls” resolved by 28 composable models and solvers, enabling sub-second design-space exploration that identifies binding constraints and synthesizes ideal hardware specifications across the entire ML systems lifecycle.  \n1 Introduction  \nMachine learning has become infrastructure (Sutton, 2019) . Training a frontier model now requires orchestrating tens of thousands of accelerators across datacenter fabrics where memory ceilings, network bandwidth, power delivery, and regional carbon intensities interact in non-obvious ways (Dean et al., 2012; Shoeybiet al., 2019) . The pace of this scaling is accelerating: frontier models have grown from billions to trillions of parameters in under five years, and the infrastructure cost of a single training run now rivals that of a small datacenter (DeepSeek-AI, 2025) .  \nYet the hardware required to develop intuition for these systems is prohibitively scarce: a student cannot requisition a 100,000-GPU cluster to explore how topology affects AllReduce latency, and a researcher cannot easily sweep parallelism strategies across hardware generations. This creates a growing reasoning gap. The systems are getting more complex, but the tools to think about them have not kept pace.  \nConsider a concrete example. A team deploying LLaMA-3 70B for interactive serving must answer: How many H100 GPUs are needed to meet a 50 ms time-to-firsttoken SLA at 95th-percentile latency? The answer depends on at least seven interacting constraints, each a “wall”(a hard bound imposed by physics, economics, or algorithmic scaling; see Table 1) . We say a wall binds when it is the tightest constraint, the single bottleneck that limits end-to-end performance. The binding constraint is the wall whose relaxation would yield the largest throughput improvement; all other walls are slack.1 For the LLaMA-3 serving example, the model’s 70 billion parameters require ∼140 GB in FP16, exceeding a single GPU’s 80 GB HBM capacity (Wall 2: Memory) . Tensorparallel sharding across two GPUs introduces NVLink synchronization overhead (Wall 14: Communication) . Continuous batching with PagedAttention determines KV-cache memory utilization (Wall 5: Batching) . The decode phase is memory-bandwidth-bound at 3.35 TB/s per device (Wall 4: Serving) . Tail latency under load follows Erlang-C queueing dynamics (Wall 7: Tail Latency) . And the fleet’s total cost of ownership constrains what is economically viable (Wall 17: Capital) . A similar multi-wall analysis applies to training: determining the optimal parallelism strategy for a 512-GPU run involves compute throughput, memory capacity, communication overhead, scaling laws, reliability, and economics simultaneously. No single eq","cbCaieopJZ7ohWAl","https://ap.wps.com/l/cbCaieopJZ7ohWAl","pdf",1210220,1,45,"English","en",105,"# Introduction\n## Scaling Challenges and the Reasoning Gap\n## Multi-Wall Constraint Analysis\n## Limits of Existing Tools\n## MLSYS · IM Framework Overview","[{\"question\":\"Why is reasoning across ML hardware and infrastructure considered difficult?\",\"answer\":\"Empirical profiling requires the target hardware, while cycle-accurate simulation is too slow for rapid exploration. Analytic tools also struggle to cover the full stack and may not enforce correctness of units and relationships.\"},{\"question\":\"What problem does the “Systems Walls” concept address?\",\"answer\":\"It treats multiple interacting constraints as distinct “walls” and identifies which wall is binding, meaning it is the tightest bottleneck that limits end-to-end performance.\"},{\"question\":\"How does MLSYS · IM prevent common modeling errors?\",\"answer\":\"It enforces dimensional and unit integrity at runtime and requires typed, provenance-tracked registry inputs so numbers used in analyses have documented sources, avoiding silent conversion mistakes.\"}]",1784182228,113,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"first-principles-infrastructure-modeling-for-machine-learning-systems","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/first-principles-infrastructure-modeling-for-machine-learning-systems/82677/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why is reasoning across ML hardware and infrastructure considered difficult?","Question",{"text":75,"@type":76},"Empirical profiling requires the target hardware, while cycle-accurate simulation is too slow for rapid exploration. Analytic tools also struggle to cover the full stack and may not enforce correctness of units and relationships.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What problem does the “Systems Walls” concept address?",{"text":80,"@type":76},"It treats multiple interacting constraints as distinct “walls” and identifies which wall is binding, meaning it is the tightest bottleneck that limits end-to-end performance.",{"name":82,"@type":73,"acceptedAnswer":83},"How does MLSYS · IM prevent common modeling errors?",{"text":84,"@type":76},"It enforces dimensional and unit integrity at runtime and requires typed, provenance-tracked registry inputs so numbers used in analyses have documented sources, avoiding silent conversion mistakes.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]