Compute Services

Business Value: CCP delivers a complete compute portfolio — virtual machines, containerized workloads, and bare metal — all provisioned through one self-service portal, governed by one RBAC system, and tracked by one metering engine. Teams provision the right compute for the right workload without operations team intervention.

Compute Service Portfolio

Service	Phase	Description
Virtual Machine (VM)	MVP1	On-demand VMs with configurable flavors, OS images, network attachment, and storage
Container as a Service (CaaS)	MVP1	Managed Kubernetes clusters with self-service deployment, scaling, and lifecycle management
Bare Metal as a Service (BMaaS)	MVP1	Dedicated physical servers provisioned on-demand for workloads requiring maximum performance

Virtual Machine (VM)

CCP provides self-service virtual machine provisioning backed by OpenStack (v2023.2) through the Cirrus Cloud Platform (CCP) IaaS orchestrator. Users select from pre-configured flavors and images, attach network interfaces and storage volumes, and deploy production VMs in minutes.

VM Capabilities

• Flavor Catalogue: Administrators define flavors (vCPU, RAM, root disk combinations) in the Admin Console; users select from available options in their cell
• OS Image Library: Standard OS images (Linux distributions, Windows Server) managed centrally; tenants can register custom images
• Network Attachment: VMs attach to tenant VPC subnets with configurable security groups and public IP assignment
• Storage Integration: Root volumes on OpenStack block storage; additional data volumes attachable post-deployment
• Availability Zone Placement: Users can specify AZ placement for latency, redundancy, and compliance requirements
• Instance Actions: Console access, start, stop, reboot, resize, snapshot, and deletion via portal or API
• GPU Support: GPU-enabled flavors available for AI/ML and graphics workloads on appropriate hardware

VM Lifecycle

1. Select Flavor & Image — Choose from catalogue of pre-defined compute flavors and OS images
2. Configure Network — Attach to VPC subnet; optionally assign public IP; configure security group rules
3. Attach Storage — Root volume pre-attached; additional block volumes configurable at launch or post-deployment
4. Deploy — Portal triggers orchestration workflow; VM becomes available in minutes
5. Manage — Monitor, resize, snapshot, or decommission through the portal; all actions tracked for metering

Container as a Service (CaaS)

CCP delivers Kubernetes-based containerized workloads through Cloud Orbiter, the Kubernetes orchestrator component. The platform manages the full Kubernetes cluster lifecycle — provisioning, upgrades, scaling, and decommissioning — through the self-service portal.

CaaS Capabilities

• Cluster Provisioning: Self-service creation of Kubernetes clusters; Cloud Orbiter manages control plane and worker node lifecycle
• Cluster Controller and Agent: The Cluster Controller centralizes Kubernetes API access; agents deployed on each cluster establish secure outbound connections to the controller
• Application Deployment: Deploy containerized applications directly from the portal; supports Helm charts, Kubernetes manifests, and container registry integration
• Container Registry: Built-in container registry for air-gapped deployments; push and pull custom images within the platform
• Observability: Integrated cluster metrics via observability-ui; CPU, memory, pod health, and deployment status visible in the portal
• Terminal Access: Browser-based Kubernetes shell (orbiter-term) for direct cluster operations without external kubectl configuration
• Scaling: Horizontal pod autoscaling and manual worker node scaling through the portal

Kubernetes Cluster Management

CCP uses the Coredge Kubernetes Platform (CKP) as the underlying Kubernetes distribution, providing enterprise-grade cluster management with:

• Production-ready Kubernetes with CKP distribution
• Cluster agent architecture for secure multi-cluster management
• Centralized kubeconfig and API proxy via Cluster Controller
• Per-cluster monitoring integration with Prometheus and Grafana

Bare Metal as a Service (BMaaS)

For workloads that require direct hardware access — databases requiring maximum I/O performance, HPC workloads, GPU compute, or applications where virtualization overhead is unacceptable — CCP provides Bare Metal as a Service through MaaS (Metal as a Service, v3.4.9) integration via the baremetal-plugin microservice.

BMaaS Capabilities

• Self-Service Provisioning: Users request bare metal servers through the portal; automated provisioning via MaaS handles OS installation and configuration
• Server Catalogue: Platform administrators define server types in the Admin Console; users select from available bare metal flavors
• OS Deployment: Golden OS images deployed automatically; server moves from powered-off to workload-ready without manual intervention
• Network Configuration: Bare metal servers integrated into tenant VPC networking; management and workload traffic segregated
• Storage Attachment: NetApp-backed block and file storage attachable to bare metal nodes; high IOPS volumes for database and HPC workloads
• Lifecycle Management: Commission, decommission, and re-provision through the portal; full metering of node-hours from allocation to release

BMaaS Use Cases

• High-performance database servers requiring direct disk access
• HPC and AI/ML workloads requiring maximum compute density
• Network functions and appliances requiring dedicated hardware
• Legacy applications that cannot be virtualized
• Security-sensitive workloads where shared hypervisor infrastructure is not acceptable

Platform Scalability

Dimension	Capacity
Virtual Machines	Up to 50,000 VMs per deployment
Container Pods	Up to 200,000 pods per deployment
Scalability	Worker nodes added on demand without downtime
IaaS Orchestrator	OpenStack v2023.2
Bare Metal Provisioning	MaaS v3.4.9

Compute Metering

All compute resources are metered from the moment of allocation:

• VM-Hours: Tracked from VM creation to deletion; billed per vCPU and RAM allocation
• Container Cluster-Hours: Tracked per cluster from provisioning to decommission
• Bare Metal Node-Hours: Tracked from server allocation to release; precise timestamp-based billing
• Quota Enforcement: Quotas enforced at Tenant and Cell level in real time — new compute requests blocked when quota is reached