What Is It?
Cubes are lightweight virtual machines with built-in high-speed NVMe storage, designed for cost-effective workloads that don't require continuous maximum compute. Each Cube uses a fixed template (vCPU, RAM, NVMe size) that cannot be changed after provisioning. Built on shared resources, they can rival physical servers because the platform redistributes available performance among instances, which makes them ideal for development environments, testing, website hosting, and simple web applications.
Quick Facts
| Aspect | Details |
|---|---|
| Type | Template-based virtual machine with direct-attached NVMe |
| Templates | 9 pre-defined sizes (Basic and Memory families); the older Cubes XS to 4XL templates are closed to new deployments since 1 October 2024 |
| Storage | NVMe volume (60 GB - 960 GB) + up to 23 additional block storage devices |
| Resource Model | Fixed vCPU, RAM, and NVMe size at provisioning (immutable) |
| Encryption | AES-XTS 256-bit at rest (NVMe volumes created after the feature launch) |
| Networking | IPv4/IPv6, LANs, firewalls, flow logs |
| Management | Data Center Designer (DCD), Cloud API |
| Locations | Berlin, Frankfurt (FRA, FRA/1 and FRA/2), London, Worcester, Paris, LogroƱo, Newark, Lenexa; may be unavailable in some older Frankfurt data centers |
Template Specifications
| Template | vCPUs | RAM | NVMe Storage | Best For |
|---|---|---|---|---|
| Basic Cube XS | 1 | 2 GB | 60 GB | Minimal workloads, testing |
| Basic Cube S | 2 | 4 GB | 120 GB | Small applications, dev environments |
| Basic Cube M | 4 | 8 GB | 240 GB | Standard web apps, microservices |
| Basic Cube L | 8 | 16 GB | 480 GB | Medium workloads, databases |
| Basic Cube XL | 16 | 32 GB | 960 GB | Large applications, data processing |
| Memory Cube S | 2 | 8 GB | 120 GB | Memory-intensive small apps |
| Memory Cube M | 4 | 16 GB | 240 GB | Caching layers, in-memory databases |
| Memory Cube L | 8 | 32 GB | 480 GB | Analytics, memory-heavy workloads |
| Memory Cube XL | 16 | 64 GB | 960 GB | Large in-memory applications |
Basic templates: Balanced CPU-to-RAM ratio for general-purpose workloads.
Memory templates: Higher RAM-to-CPU ratio for memory-intensive applications.
What You Can Do
Quick Template-Based Provisioning
Select from pre-defined templates and have a Cube running in minutes via DCD or API. Templates fix vCPU, RAM, and NVMe size at creation, simplifying sizing decisions and accelerating deployment.
Expand Storage Capacity
Attach up to 23 additional Block Storage devices (Essential, Balanced or Performance classes). The NVMe volume occupies one device slot, leaving 23 available for expansion.
Suspend and Resume
Suspend a Cube to turn it off while keeping it saved: it releases its CPU, RAM and non-reserved IP addresses, and you can resume it later. Shut down the operating system first to prevent data loss. Note that billing continues during suspension.
Snapshot and Clone
Create snapshots of NVMe volumes or attached Block Storage for backup and cloning. Use snapshots to recreate identical Cubes or restore to previous states.
Flexible Boot Configuration
Select any attached device (NVMe, Block Storage, CD-ROM) or network as the boot volume.
Secure Data Protection
NVMe volumes created after the feature became available are encrypted at rest with AES-XTS 256-bit, with a unique key per volume. Secure deletion destroys the encryption key and zeros the metadata, so deleted data cannot be recovered.
Network Integration
Connect Cubes to LANs, enable IPv4/IPv6 dual-stack, configure firewalls, and enable Flow Logs for traffic analysis. Cubes work with the other VDC Networking services in their location, such as IP failover groups and Network Security Groups. NIC Multi-Queue is not available for Cubes.
Best For
| Scenario | Why It Fits |
|---|---|
| Development and testing environments | Cost-effective with quick provisioning and snapshot-based workflows |
| Small web applications and APIs | Fixed templates match typical web server requirements |
| Microservices and containerized workloads | Lightweight footprint with fast NVMe storage |
| Short-lived compute tasks | Template-based sizing accelerates deployment for temporary workloads |
| Memory-intensive applications | Memory Cube templates provide high RAM-to-CPU ratios |
| Variable-load workloads | Cost efficiency for applications with intermittent usage patterns |
Consider Alternatives If
| If You Need... | Consider | Why |
|---|---|---|
| Customizable CPU, RAM, or storage | Compute Engine | Flexible resource allocation and Live Vertical Scaling |
| Container orchestration | Managed Kubernetes | Automated container management and scaling |
Key Considerations
Billing & Costs
- Main billing: One hourly price per template covering vCPU, RAM and NVMe, from 0,007 EUR per hour (Basic Cube XS) to 0,114 EUR per hour (Memory Cube XL)
- When suspended: Billing continues for all resources
- Cost optimization: Snapshot the NVMe volume first, then delete unused Cubes to stop billing; recreate them from the snapshot later (the new Cube may get a different IP address)
- Additional costs: Attached Block Storage, data transfer, public IPs, snapshots, and operating system licenses (Windows Server or RHEL, billed per vCPU)
Limitations
- Immutable templates: vCPU, RAM, and NVMe size cannot be changed after provisioning
- Maximum devices: 24 total (the NVMe volume uses 1, leaving 23 for Block Storage and CD-ROM devices).
- NVMe constraints: Cannot unmount or delete NVMe volume independently (tied to Cube lifecycle)
- No Live Vertical Scaling of CPU or RAM: vCPU and RAM are fixed by the template; Live Vertical Scaling works only for NICs and block storage volumes
- Suspended billing: Suspension does not reduce costs; only deletion stops charges
- Template selection: Must choose appropriate template at creation (no resize option)
- High availability: Cubes rely on the application for failover, not the infrastructure (per-instance SLA 99.9% monthly); use IP failover groups to build your own HA setup
Management Options
- Data Center Designer (DCD): Drag-and-drop interface for Cube creation, configuration, snapshots, power management
- Cloud API: Programmatic access via REST endpoints (
/datacenters/{dcId}/servers) - Composite API call: Single request that creates the Cube together with its NVMe volume and image (required, because the NVMe volume cannot be added after provisioning)
- Template retrieval: Query available templates via
/templatesendpoint before provisioning