PPN architecture overview
fix(infrastructure): add dated Correction callouts to 9 of 19 infrastructure/ articles, verified against canonical origin/main throughout — fabricated cgroups/virtio_balloon mechanism (3 articles), os-orchestration scaffold, wrong UDP port/packet format/crate names (2 articles), hardware-vs-filesystem WORM enforcement overclaim (compliance-relevant, escalated), telemetry payload overclaim, shared-not-per-provider API key storage; 7 verified clean, 2 already correctly self-corrected
@@ -44,7 +44,7 @@ The **[[genesis-protocol]]** governs first boot: a node generates its keypair fr The hypervisor layer is the compute substrate. **`os-infrastructure`** is the hypervisor layer that hosts the virtual machines running Totebox Archives and orchestration gateways (QEMU/KVM-hosted today; **intended** to become a bare-metal Type-I hypervisor under NetBSD/NVMM in Phase 2 and seL4/Microkit in Phase 3). It manages a **per-node resource pool**: memory via `virtio_balloon` (inflation reclaims guest RAM into the node pool; deflation returns it) and CPU via cgroups v2 `cpu.weight` per QEMU process. **`os-infrastructure`** is the hypervisor layer that hosts the virtual machines running Totebox Archives and orchestration gateways (QEMU/KVM-hosted today; **intended** to become a bare-metal Type-I hypervisor under NetBSD/NVMM in Phase 2 and seL4/Microkit in Phase 3). It manages a **per-node resource pool**: memory via `virtio_balloon` (inflation reclaims guest RAM into the node pool; deflation returns it) and CPU via cgroups v2 `cpu.weight` per QEMU process. (Correction, 2026-08-02, verified against canonical `origin/main`: no `virtio_balloon` or cgroups/`cpu.weight` code exists anywhere in `os-infrastructure` or `service-vm-fleet` — real `os-infrastructure/src/main.rs` is a bare-metal Multiboot2 boot stub doing framebuffer text + mDNS genesis handshake, with zero resource-pooling logic. Separately, `os-orchestration` — described elsewhere in this article as a stateless PSP aggregator — is a 2-line placeholder scaffold on canonical, matching the finding on [[os-orchestration-stateless-hub]]. Flagged, not resolved.) The pool is bounded to the physical node. Cross-node workload placement is the Totebox Orchestration layer's responsibility; once a VM is placed on a node, the hypervisor manages its local resource allocation.