Capacity, availability, and consensus modeling tools.
Deterministic architecture calculators built by our principal infrastructure practice for capacity planning, SLA error-budget governance, sovereign GPU cluster sizing, and multi-region quorum placement. All calculations run 100% locally in your browser—no proprietary parameters leave your workstation.
1. Composite SLA, Error Budget & Quorum Availability Calculator
Model effective system availability across serial single-region dependencies, warm standby failover windows, and zero-RTO active-active multi-region meshes, bounded by realistic common-mode failure ceilings.
A = 1 - (1 - A_reg)^N) predict unrealistic seven- or
eight-nines numbers because they assume statistically independent failures. In
production systems, bad database migrations, global BGP/DNS misconfigurations, or
bad CI/CD rollouts introduce correlated common-mode risk. Our model applies serial
tier coupling for single-region deployments, deducts annualized RTO convergence windows
for warm-standby failovers, and enforces a 99.9995% (~2.6 min/year)
common-mode ceiling. For implementation details, read Designing Deterministic
Active-Active Kubernetes Clusters.
2. Sovereign LLM VRAM, KV-Cache & GPU Cluster Sizer
Calculate exact memory footprints (model weights + Grouped-Query Attention KV-cache + CUDA activation workspace), valid Tensor-Parallel GPU topologies, memory-bandwidth decode throughput, and rack power envelopes for private on-premise LLM deployments.
N_params * B_prec * 1.08 for CUDA context and
NCCL buffers) plus dynamic KV-cache (2 * L * N_kv_heads * d_head * C_len * Batch * B_kv).
We reserve 12% HBM headroom to prevent vLLM / TensorRT-LLM preemption thrashing under
long-context RAG spikes, and round accelerator counts to valid power-of-two
Tensor-Parallel degrees (1, 2, 4, 8...) so attention heads partition evenly
across NVLink domains. Dive deeper in Production Architecture for
Sovereign On-Premise LLM Inference.
3. Global Multi-Region Quorum & Fiber RTT Estimator
Compute deterministic Raft/Paxos write commit latencies and pairwise round-trip times across EMEA, Americas, and APAC hubs using geodesic fiber-corridor distances and refractive silica propagation physics.
~200 km/ms (refractive index n ≈ 1.468).
Real terrestrial and subsea cable ducts add a 1.38x routing tortuosity
factor over great-circle Haversine distance, plus ~2.6 ms for DWDM
transponders, BGP routers, mTLS framing, and kernel socket serialization.
| Region Node | Geodesic Dist. | Est. Fiber RTT | Consensus Role |
|---|
Why Tbilisi + Frankfurt + Dubai forms an optimal EMEA quorum triangle
Placing three consensus voter regions across Tbilisi, Frankfurt, and Dubai yields physical jurisdictional separation across three distinct power grids and subsea/terrestrial fiber corridors while maintaining majority write commits around ~30–42 ms—avoiding the 115+ ms transatlantic penalty of US-anchored quorums.
Read the Financial Ledger Guide →Commission a formal feasibility & TCO architecture review
Export your specification from any calculator above and attach it to a technical RFP. Our principal systems engineers validate hardware BOMs, PCIe/NVLink topologies, network BGP peering, and compliance boundaries within 24 hours under mutual NDA.
Submit Technical Specification →