From Demand Windows to Rack Capacity and Lifecycle Cash Cost
How dated power constraints and whole-window rack reservations affect capacity and cash cost in a frozen synthetic planning study.
Infrastructure research and open-source tools
Papers, methods and the evidence behind each finding.
These studies use declared synthetic scenarios. Code and test results make the calculations inspectable; they do not establish a deployed service or operational qualification.
Papers
Regional capacity and managed inference cost floors
A frozen synthetic study of regional eligibility, peak reserve and managed inference cost floors, with reproducible methods and tests.
Network capacity is sometimes compute capacity
The paper assigns every accelerator-second to one of four fates. It expresses an infrastructure change as the number of extra accelerators that would deliver the same productive work. All results are analytical or simulated.
Open-source tools
Compare infrastructure choices with simulation tools and validators. Inspect the assumptions, run the code and check the results.
GPU Cluster Networking
Training workers wait when they cannot exchange results quickly enough. Compare communication costs and workload placement in a modeled GPU cluster.
GPU Cluster Scheduling
Compare how scheduling policies use the same GPU pool under power limits, failures and changing demand.
Chaos Fidelity Standard
Check whether a failure experiment tests the part of the system it claims to test.
pip install ai-cluster-chaos-fidelity chaos-fidelity my-experiment.yamlReliability Economics
Compare the cost of saving checkpoints with the work lost after failures, including spare capacity and recovery choices.
Governed Autonomy
Check evidence and independent approval before increasing software control over infrastructure. This tool evaluates records; it does not execute changes.
pip install governed-autonomy-gate promotion-gate my-record.yamlCompute↔Power Placement
Compare keeping a workload in place, splitting it between sites or moving it. Account for power costs, available capacity, communication and deadlines.
Edge↔Neocloud Placement
Check which locations meet a workload's stated network, power, latency and data-location requirements.
pip install edge-continuum-placement edge-placement evaluate -w frontier-pretrain -t tower # exit 1AI-RAN↔Neocloud Resiliency
Compare simulated downtime under different power, connectivity and timing failures, with alternative resilience measures.
pip install airan-neocloud-resiliency airan-resiliency ladderCooling & the PUE Ladder
Compare cooling options, rack power limits and facility power overhead under declared conditions.
pip install cooling-pue-ladder pue-ladder gate --density 120 --rung contained-air # exit 1The Span Admission Contract
Check a proposed workload against declared network constraints. The result includes missing evidence alongside the decision.
pip install span-contract spancontract validate envelope.jsonOptical Circuit Intent
Check a proposed optical-circuit change against network limits and workload timing requirements. The tool does not change the network.
pip install optical-circuit-intent ocintent ladderSlice-Packing a Torus Pod
Explore how failures and maintenance affect connected allocations in a torus network, whose opposite edges connect.
pip install slice-packer-torus slicepacker cordon --shape 16,16,16
Inspect the evidence
The joined-capacity model in the usable-compute repository has an evidence manifest recording assumptions, sources and scope. It distinguishes scenario choices from empirical measurements.