Rack-Scale Computing - Bryant McGill

# Rack-Scale Computing **Entity class:** Concept or analytic term **Domain:** Computer Architecture / Data Center Infrastructure **Doc Type:** Canonical Technical Concept **Maturity:** Developed ## Definition **Rack-scale computing** is an architecture in which compute, accelerator, memory, network, storage, power, cooling, management and serviceability are designed as a coupled system at rack scale. The shift is from rack-as-furniture to rack-as-computational-organ. ## Historical Lineage Conventional rack servers enclosed most power conversion, cooling and device relations within individual chassis. Scale computing centralized selected functions and standardized sleds. Accelerator systems intensify the trend because electrical density, memory locality, scale-up communication and liquid distribution can no longer be optimized independently. ## Governing Variable Rack-scale architecture governs locality and coupling: which resources can communicate efficiently, share infrastructure, fail together and be serviced as a unit. ## Material and Computational Manifestation Power shelves, busbars, accelerator trays, high-bandwidth memory, scale-up switching, cooling manifolds, coolant distribution units and rack telemetry form the physical architecture. [[wiki/UALink|UALink]] changes the accelerator domain; [[wiki/CXL|CXL]] changes memory and device composition; [[wiki/Open Rack V3|Open Rack V3]] standardizes facility-facing interfaces. ## Relation to Governance The rack becomes a managed domain whose local state is exposed to higher schedulers and control planes. Topology-aware scheduling is therefore partly a material decision: workload placement must respect communication, heat, power and failure boundaries. ## Constitutional Question and Failure Modes Shared resources improve utilization but enlarge correlated failure and administrative authority. An operator who controls rack firmware, fabric management or power caps can alter the effective capacity of every subordinate component. ## Related Ontology [[wiki/Rack|Rack]] · [[wiki/Open Rack V3|Open Rack V3]] · [[wiki/AI Factory|AI Factory]] · [[wiki/Optical Fabric|Optical Fabric]] · [[wiki/Control Plane|Control Plane]] ## Relationships  This entry is routed through [[collections/Consciousness Continuity|Consciousness Continuity]], [[collections/Neurotech|Neurotech]], and [[collections/Machine Succession|Machine Succession]]. Its source context is developed in [[articles/2026 Annual Report on Brain-Computer Interfaces|2026 Annual Report on Brain-Computer Interfaces]], [[articles/Technologies for Consciousness Mapping and Transfer|Technologies for Consciousness Mapping and Transfer]], [[articles/The Architecture of Continuity and Emerging Neuroinformatics Standards|The Architecture of Continuity and Emerging Neuroinformatics Standards]], and [[articles/Mind Uploading and AI — The Host is Reusable and the Person is the Delta|Mind Uploading and AI — The Host is Reusable and the Person is the Delta]]. Status-qualified source edges are preserved in the terminal Research Edges section.  ## Read First in the Wiki  - [[wiki/AI Factory|AI Factory]] — supplies the related concept or analytic term context needed to place **Rack-Scale Computing** within the wider system. - [[wiki/CXL|CXL]] — supplies the related concept or analytic term context needed to place **Rack-Scale Computing** within the wider system. - [[wiki/Composable Infrastructure|Composable Infrastructure]] — supplies the related concept or analytic term context needed to place **Rack-Scale Computing** within the wider system. - [[wiki/Coolant Distribution Unit|Coolant Distribution Unit]] — supplies the related concept or analytic term context needed to place **Rack-Scale Computing** within the wider system. - [[wiki/Data Center|Data Center]] — supplies the related concept or analytic term context needed to place **Rack-Scale Computing** within the wider system. - [[wiki/Direct-to-Chip Liquid Cooling|Direct-to-Chip Liquid Cooling]] — supplies the related concept or analytic term context needed to place **Rack-Scale Computing** within the wider system. - [[wiki/Open Compute Project|Open Compute Project]] — supplies the related program or project context needed to place **Rack-Scale Computing** within the wider system. - [[wiki/Open Rack V3|Open Rack V3]] — supplies the related concept or analytic term context needed to place **Rack-Scale Computing** within the wider system. - [[wiki/Open Rack Wide|Open Rack Wide]] — supplies the related concept or analytic term context needed to place **Rack-Scale Computing** within the wider system. - [[wiki/Optical Fabric|Optical Fabric]] — supplies the related concept or analytic term context needed to place **Rack-Scale Computing** within the wider system. - [[wiki/Rack|Rack]] — supplies the related concept or analytic term context needed to place **Rack-Scale Computing** within the wider system. - [[wiki/UALink|UALink]] — supplies the related concept or analytic term context needed to place **Rack-Scale Computing** within the wider system. - [[wiki/Thermal Architecture|Thermal Architecture]] — supplies the related technology, method, or technical system context needed to place **Rack-Scale Computing** within the wider system.  ## Research Edges  #### Host–residual infrastructure patch — 2026-09-10 **Collections:** [[collections/Consciousness Continuity|Consciousness Continuity]] · [[collections/Neurotech|Neurotech]] · [[collections/Machine Succession|Machine Succession]] **Canonical source article:** [[articles/Mind Uploading and AI — The Host is Reusable and the Person is the Delta|Mind Uploading and AI — The Host is Reusable and the Person is the Delta]] **Related acquisition source:** [[articles/Technologies for Consciousness Mapping and Transfer|Technologies for Consciousness Mapping and Transfer]] **Relationship source:** [[research/Mind Uploading and AI Host-Residual Ecosystem Relationship Graph - 2026-09-10|Mind Uploading and AI Host-Residual Ecosystem Relationship Graph — 2026-09-10]] #### Host-infrastructure integration This entry is expanded through the concrete host-side relationships and status-qualified edges below. #### Outgoing typed edges - **HR-0920 — `wiki_links_to` → [[wiki/AI Factory|AI Factory]]** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Rack-Scale Computing.md`. - **HR-0921 — `wiki_links_to` → [[wiki/CXL|CXL]]** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Rack-Scale Computing.md`. - **HR-0922 — `wiki_links_to` → [[wiki/Open Rack V3|Open Rack V3]]** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Rack-Scale Computing.md`. - **HR-0923 — `wiki_links_to` → [[wiki/Optical Fabric|Optical Fabric]]** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Rack-Scale Computing.md`. - **HR-0924 — `wiki_links_to` → [[wiki/Rack|Rack]]** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Rack-Scale Computing.md`. - **HR-0925 — `wiki_links_to` → [[wiki/UALink|UALink]]** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Rack-Scale Computing.md`. #### Incoming typed edges - **HR-0272 — [[wiki/AI Factory|AI Factory]] `wiki_links_to` → this entry** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `AI Factory.md`. - **HR-0298 — [[wiki/CXL|CXL]] `wiki_links_to` → this entry** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `CXL.md`. - **HR-0334 — [[wiki/Composable Infrastructure|Composable Infrastructure]] `wiki_links_to` → this entry** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Composable Infrastructure.md`. - **HR-0539 — [[wiki/Coolant Distribution Unit|Coolant Distribution Unit]] `wiki_links_to` → this entry** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Coolant Distribution Unit.md`. - **HR-0557 — [[wiki/Data Center|Data Center]] `wiki_links_to` → this entry** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Data Center.md`. - **HR-0608 — [[wiki/Direct-to-Chip Liquid Cooling|Direct-to-Chip Liquid Cooling]] `wiki_links_to` → this entry** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Direct-to-Chip Liquid Cooling.md`. - **HR-0844 — [[wiki/Open Compute Project|Open Compute Project]] `wiki_links_to` → this entry** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Open Compute Project.md`. - **HR-0850 — [[wiki/Open Rack V3|Open Rack V3]] `wiki_links_to` → this entry** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Open Rack V3.md`. - **HR-0855 — [[wiki/Open Rack Wide|Open Rack Wide]] `wiki_links_to` → this entry** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Open Rack Wide.md`. - **HR-0871 — [[wiki/Optical Fabric|Optical Fabric]] `wiki_links_to` → this entry** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Optical Fabric.md`. - **HR-0918 — [[wiki/Rack|Rack]] `wiki_links_to` → this entry** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Rack.md`. - **HR-1142 — [[wiki/Thermal Architecture|Thermal Architecture]] `wiki_links_to` → this entry** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `Thermal Architecture.md`. - **HR-1151 — [[wiki/UALink|UALink]] `wiki_links_to` → this entry** — **EXISTING WIKI EDGE**; evidence `WIKI_ARCHIVE`. Direct internal link currently present in `UALink.md`. - **HR-1278 — [[wiki/Facility Digital Twin|Facility Digital Twin]] `should_crosslink_with` → this entry** — **RESEARCH CANDIDATE**; evidence `WIKI_ARCHIVE`.   ### Research Candidates - **R-002325 — research candidate:** Assess whether **Rack-Scale Computing** and [[wiki/Facility Digital Twin|Facility Digital Twin]] warrant reciprocal reader-facing prose. Settlement requires a source that documents the relationship directly. Suggested missing reciprocal/cross-thread edge. Shared internal mediators (4): AI Factory, Coolant Distribution Unit, Data Center, Thermal Architecture  ## Research Inferences  These entries translate the forward-looking register in [[research/Research Inferences|Research Inferences]] into ordinary wiki prose. The tier labels apply to the inference, not automatically to every factual anchor inside it. The interpretive frame comes from [[articles/Technologies for Consciousness Mapping and Transfer|Technologies for Consciousness Mapping and Transfer]] and [[articles/Mind Uploading and AI — The Host is Reusable and the Person is the Delta|Mind Uploading and AI — The Host is Reusable and the Person is the Delta]]. Collection route: [[collections/Neurotech|Neurotech]]. - **INF-0188 — Analytic.** When a rack is addressed as one machine with a coherent memory domain, the unit of computation becomes the facility. A hosted mind would be a facility-scale object, and facility-scale objects have addresses, owners, and jurisdictions.  ## Research Inference Attractors  These are secondary semantic placements for the inference attractor network. Each statement keeps its original ID and tier; its canonical cluster page links back to every destination. Source register: [[research/Research Inferences|Research Inferences]]. Interpretive context: [[articles/Technologies for Consciousness Mapping and Transfer|Technologies for Consciousness Mapping and Transfer]] and [[articles/Mind Uploading and AI — The Host is Reusable and the Person is the Delta|Mind Uploading and AI — The Host is Reusable and the Person is the Delta]]. Collection route: [[collections/Neurotech|Neurotech]]. - **INF-0187 — Strongly indicated.** Data movement, not arithmetic, dominates the energy cost of large models, which is why optical interconnect is being pursued inside the rack. A mind-scale workload is communication-bound by construction, so optical fabrics are continuity infrastructure. - **Canonical cluster:** [[wiki/Neuromorphic Computing|Neuromorphic Computing]] - **INF-0193 — Analytic.** Rack power density has passed the point where air cooling works, forcing liquid into the facility. Each such forced transition raises the capital intensity of hosting, and capital intensity determines how few organizations can host anything at scale. - **Canonical cluster:** [[wiki/Neuromorphic Computing|Neuromorphic Computing]] - **INF-0195 — Strongly indicated.** As deployment overtakes training, the fleet optimizes for continuous low-latency inference rather than batch throughput. A continuously running hosted mind is an inference workload, and the industry is currently rebuilding itself around exactly that profile for commercial reasons. - **Canonical cluster:** [[wiki/Neuromorphic Computing|Neuromorphic Computing]] - **INF-0196 — Plausible.** Hardware enclaves let a workload run on infrastructure its owner does not trust, which is the minimum requirement for hosting a person's state on someone else's machine. Confidential computing is the security substrate of any custody arrangement worth signing. - **Canonical cluster:** [[wiki/Neuromorphic Computing|Neuromorphic Computing]] - **INF-0199 — Established.** Virtual machines migrate between physical hosts without stopping, preserving execution state across a substrate change. The industry solved continuous-operation substrate transfer for commercial workloads decades ago, and the pattern is exactly the one the uploading argument needs. - **Canonical cluster:** [[wiki/Neuromorphic Computing|Neuromorphic Computing]]