Instance Types
An instance type picks the vCPU count, memory, baseline network bandwidth, and underlying hardware family for a VM. You pick one at create time and can change it later with exc compute resize.
List what’s available:
exc compute instancetype listPricing for each family: Compute pricing and GPU pricing.
Current families
| Family | What it is | Pick when |
|---|---|---|
t1a.<size> | Burstable, EBS-backed | Cheap dev boxes, runners, intermittent workloads |
m1a.<size> | Dedicated CPU, EBS-backed | Web apps, APIs, batch workers, anything steady-state |
g1.<size>-<gpu> / nv* | GPU instances | LLM inference, ML training, professional visualisation |
Naming: family prefix + dot + size.
- Sizes in
t1agomicro,small,medium. - Sizes in
m1agolarge,xlarge,2xlarge,4xlarge,8xlarge,16xlarge— doubling vCPU and RAM at each step. - GPU instances name the card in the suffix (e.g.
nv1a.4xlargeships an Nvidia RTX 6000 Pro Blackwell). See the GPU pricing page for the full catalogue.
Memory and network bandwidth scale with size within a family. Exact ratios live in exc compute instancetype list and on the
pricing pages.
Deprecated families
| Family | Replacement | Notes |
|---|---|---|
n1.<n>c (e.g. n1.2c, n1.16c) | m1a.<size> | EOL — use m1a for new VMs |
t1.<size> (e.g. t1.micro) | t1a.<size> | EOL — use t1a for new VMs |
Existing VMs on deprecated types keep running but can’t be created or resized into. See Compute bandwidth limits for the bandwidth tables, including the deprecated rows.
Picking a family
| Workload | Try |
|---|---|
| Web app, API, batch worker | m1a.large / m1a.xlarge / m1a.2xlarge |
| Cheap dev box, runners | t1a.micro / t1a.small |
| LLM inference, ML training | nv* GPU instances (see
GPU pricing) |
| Heavy database / cache | Larger m1a.* with provisioned-IOPS volumes |
Resize
Stop the VM, resize, start:
exc compute stop --vm_id 42
exc compute resize --vm_id 42 --instance_type m1a.2xlarge
exc compute start --vm_id 42The root and data volumes survive the resize unchanged.
Bandwidth caps
Each instance type has a maximum transmit bandwidth, listed in the Compute bandwidth limits page. Receive bandwidth is uncapped today.
Quotas
Your org has a per-account vCPU quota (see exc quota). To raise it, email support@excloud.dev with the workload context.