Compute · zone mum-1a
Virtual machines from ₹0.236 an hour.
AMD EPYC cores, NVMe block storage, and 50 Gbps networking on every size, from a 2 vCPU burstable box to a 64 vCPU dedicated machine. Billed by the hour. Stop the VM and the bill stops with it.
$ curl -fsSL https://excloud.dev/install.sh | bash
$ exc compute create --name api-1 \ --instance_type t1a.small --wait ✓ running — api-1 · t1a.small · mum-1a $ exc compute list NAME TYPE ZONE STATE api-1 t1a.small mum-1a running
Instance types
Nine sizes across two classes, all rates listed below.
Both classes run on AMD EPYC with EBS NVMe disks and 50 Gbps networking. The difference is how the cores are sold: t1a shares them, m1a hands them to you outright.
t1a
Burstable. For Node.js, Go, Python and other workloads that mostly wait, then spike.
| Instance | vCPU | RAM | Rate |
|---|---|---|---|
| t1a.micro | 2 | 1 GiB | ₹0.236/hr |
| t1a.small | 2 | 2 GiB | ₹0.472/hr |
| t1a.medium | 2 | 4 GiB | ₹0.945/hr |
m1a
Dedicated. For Kubernetes, multi-tenant deployments, and databases that need the same performance at 3 p.m. and 3 a.m.
| Instance | vCPU | RAM | Rate |
|---|---|---|---|
| m1a.large | 2 | 8 GiB | ₹1.889/hr |
| m1a.xlarge | 4 | 16 GiB | ₹3.778/hr |
| m1a.2xlarge | 8 | 32 GiB | ₹7.556/hr |
| m1a.4xlarge | 16 | 64 GiB | ₹15.112/hr |
| m1a.8xlarge | 32 | 128 GiB | ₹30.224/hr |
| m1a.16xlarge | 64 | 256 GiB | ₹60.448/hr |
Hourly rates; a 720-hour month on a t1a.micro comes to ₹169.92. Full details on the compute pricing page.
Which class
The right class depends on what the CPU actually does.
A t1a shares physical cores with its neighbours, which is exactly why 2 vCPUs cost ₹0.236/hr. An m1a owns its cores outright. Both are good answers to different questions.
Pick t1a when the CPU mostly idles
Web servers, API backends, cron boxes, staging environments. Most application servers spend their lives waiting on the network or the database, and paying dedicated-core rates for waiting is a waste. Burstable cores soak up the spikes, and the whole 2 vCPU box runs you ₹169.92 a month.
Pick m1a when the CPU is the work
Kubernetes nodes, Postgres, anything multi-tenant, build runners. When sustained throughput matters, shared cores show up as noisy latency graphs. m1a gives you the whole core, every hour, and the sizes double cleanly from 2 vCPU/8 GiB up to 64 vCPU/256 GiB.
Picked wrong? exc compute resize moves a VM to a new type
without rebuilding it.
Disks & network
Storage, IPs, and egress, priced per unit.
Every VM boots from an EBS NVMe volume. Storage is ₹4/GB·mo, and if the default tier isn't enough you provision exactly the throughput and IOPS you want, priced per unit. Volumes resize online, and outlive the VM they're attached to.
On the network side: traffic in is free at any volume, IPv6 is free, and traffic out is a flat ₹1/GiB whether you move one GiB or a hundred thousand.
| Item | Spec | Rate |
|---|---|---|
| Block storage | EBS NVMe, resizable online | ₹4/GB·mo |
| Provisioned throughput | per MB/s, monthly | ₹1.79/MB/s |
| Provisioned IOPS | per IOPS, monthly | ₹0.04/IOPS |
| Public IPv4 | per address | ₹0.3/hr |
| Public IPv6 | per address | free |
| Ingress | any volume | free |
| Egress | flat, no tiers | ₹1/GiB |
Day two
A whole lifecycle from one binary.
The exc CLI covers create, stop, resize, snapshot, metrics, and serial logs, so the 2 a.m. incident doesn't require finding the console password. If you'd rather declare than type, the Terraform provider speaks the same API.
- AMD EPYC, NVMe disks, 50 Gbps networking on every size
- Public region and zone IDs documented for automation
- 99.5% instance SLA, in writing
- Online volume resize, snapshots, delete protection
$ exc compute resize --vm_id 42 \ --instance_type m1a.xlarge resizing api-1 → m1a.xlarge (4 vCPU · 16 GiB) ✓ running — api-1 · m1a.xlarge · mum-1a $ exc compute metrics --vm_id 42 cpu 12% · net 340 Mbps · disk 2.1k IOPS
Launch your first VM
A t1a.small for 24 hours runs you ₹11.33.
That's a useful ceiling for sizing a workload before you commit to anything larger. The console takes about two minutes to get you to a running instance.