Skip to main content

Private Compute Cluster

The Private Compute Cluster is the flagship infrastructure project in this wiki. It combines cloud orchestration, distributed storage, virtualization, and software-defined networking into a private platform for running virtualized workloads.

System Intent

The goal is to design a private cloud that exposes cloud-like infrastructure primitives without depending entirely on a public cloud provider. It should demonstrate strong systems taste: clean network boundaries, resilient storage, understandable operations, and enough automation that the platform can be rebuilt and explained.

Architecture Layers

LayerTechnologyPurpose
OrchestrationOpenStackVM lifecycle, identity, images, networking, APIs
ComputeKVM/QEMUHardware virtualization and guest execution
StorageCephDistributed block/object storage and resilience
NetworkNeutron + SDNTenant networks, routing, isolation, overlays
OperationsRunbooks + observabilityDeployment, debugging, and reliability

Reference Architecture

flowchart TB
user[Operator / Developer] --> horizon[OpenStack Horizon / API]
horizon --> keystone[Keystone Identity]
horizon --> nova[Nova Compute]
horizon --> glance[Glance Images]
nova --> kvm[KVM / QEMU Hypervisors]
nova --> neutron[Neutron Networking]
nova --> ceph[Ceph Block Storage]
neutron --> tenant[Tenant Networks]
neutron --> router[Virtual Routers]
ceph --> osd[OSD Nodes]
ceph --> mon[Monitor / Manager Nodes]

Design Questions

QuestionCurrent Direction
How are tenants isolated?Neutron networks, routers, and security groups.
How is VM storage made resilient?Ceph-backed volumes and image storage.
How are failures diagnosed?Runbooks, service health checks, and topology diagrams.
How does this become repeatable?Documented deployment steps first, automation after patterns stabilize.

Current References

Next Improvements

  • Add polished architecture diagrams
  • Add deployment runbooks
  • Add failure-mode analysis
  • Add capacity-planning notes
  • Add networking deep dives