Virtual Machine Resource Profiles
A resource profile defines the compute capacity allocated to a VM, including vCPUs, memory, and optional GPU resources. Choose a resource profile based on your workload's performance requirements and characteristics.
Profile types
The evroc compute service offers three types of CPU-based profiles:
- General-purpose (a1a) - Balanced CPU and memory ratio (1:4 - in units of number of vCPUs to memory in GB), suitable for most workloads including web servers, development environments, and small databases
- Compute-optimized (c1a) - Higher CPU-to-memory ratio (1:2), ideal for compute-intensive applications like batch processing, scientific computing, and high-traffic web servers
- Memory-optimized (m1a) - Higher memory-to-CPU ratio (1:8), designed for memory-intensive workloads like in-memory databases, caching servers, and data analytics
Available CPU profiles
| Profile Name | vCPUs | Memory | Architecture |
|---|---|---|---|
| a1a.xs | 1 | 4 GB | amd64 |
| a1a.s | 2 | 8 GB | amd64 |
| a1a.m | 4 | 16 GB | amd64 |
| a1a.l | 8 | 32 GB | amd64 |
| a1a.xl | 16 | 64 GB | amd64 |
| a1a.2xl | 32 | 128 GB | amd64 |
| c1a.s | 2 | 4 GB | amd64 |
| c1a.m | 4 | 8 GB | amd64 |
| c1a.l | 8 | 16 GB | amd64 |
| c1a.xl | 16 | 32 GB | amd64 |
| c1a.2xl | 32 | 64 GB | amd64 |
| m1a.s | 2 | 16 GB | amd64 |
| m1a.m | 4 | 32 GB | amd64 |
| m1a.l | 8 | 64 GB | amd64 |
| m1a.xl | 16 | 128 GB | amd64 |
GPU profiles
GPU-equipped VMs are designed for machine learning training and inference, AI workloads, and high-performance computing. GPU VMs include local NVMe SSD storage for high-throughput data access.
| Profile Name | vCPUs | Memory | Architecture | GPU model | GPU quantity | Local disk |
|---|---|---|---|---|---|---|
| gn-l40s.s | 15 | 198 GB | amd64 | NVIDIA L40S | 1 | 3,800 GB |
| gn-l40s.m | 30 | 396 GB | amd64 | NVIDIA L40S | 2 | 7,600 GB |
| gn-l40s.l | 60 | 792 GB | amd64 | NVIDIA L40S | 4 | 15,200 GB |
| gn-b200.s | 26 | 262 GB | amd64 | NVIDIA B200 | 1 | 4 TB |
| gn-b200.m | 52 | 524 GB | amd64 | NVIDIA B200 | 2 | 8 TB |
| gn-b200.l | 104 | 1048 GB | amd64 | NVIDIA B200 | 4 | 16 TB |
| gn-b200.xl | 208 | 2096 GB | amd64 | NVIDIA B200 | 8 | 32 TB |
Deprecated profiles
The following profiles use the previous naming scheme and are deprecated. Existing VMs using these profiles continue to work, but new deployments should use the current naming scheme (a1a, c1a, m1a) shown above:
| Profile Name | vCPUs | Memory | Architecture |
|---|---|---|---|
| general.xs | 1 | 4 GB | amd64 |
| general.s | 2 | 8 GB | amd64 |
| general.m | 4 | 16 GB | amd64 |
| general.l | 8 | 32 GB | amd64 |
| general.xl | 16 | 64 GB | amd64 |
| general.xxl | 32 | 128 GB | amd64 |
| compute-optimized.s | 2 | 4 GB | amd64 |
| compute-optimized.m | 4 | 8 GB | amd64 |
| compute-optimized.l | 8 | 16 GB | amd64 |
| compute-optimized.xl | 16 | 32 GB | amd64 |
| compute-optimized.xxl | 32 | 64 GB | amd64 |
| memory-optimized.s | 2 | 16 GB | amd64 |
| memory-optimized.m | 4 | 32 GB | amd64 |
| memory-optimized.l | 8 | 64 GB | amd64 |
| memory-optimized.xl | 16 | 128 GB | amd64 |
Choosing a profile
Consider these factors when selecting a resource profile:
- Workload type - Match CPU, memory, and GPU requirements to your application's needs
- Performance requirements - Start with a smaller profile and scale up based on actual usage
- Cost - Larger profiles cost more; right-size your VMs to avoid over-provisioning
You can stop a VM and resize it to a different profile if your requirements change.
Next steps
- See the Functional Description for more details on VMs and resource profiles
- Learn how to create a VM with a specific resource profile