Compute¶
Nine compute services covering virtual, dedicated, and elastic patterns — from a single-vCPU dev VM to a fleet of bare-metal AI nodes.
-
Standard, memory-optimized, CPU-optimized flavors. Per-second billing.
-
Flavor matrix — Standard / Memory / CPU — sizes, ratios, vNUMA notes.
-
Single-tenant servers, NVMe-direct, 25/100 GbE.
-
Single-tenant hypervisor for licensing-bound workloads.
-
Carved-out capacity pool with predictable performance and quotas.
-
Point-in-time snapshots, golden-image catalog, cross-region copy.
-
Policy- and schedule-based scaling, integrated with Load Balancer.
Pick the right service¶
| Need | Pick |
|---|---|
| Generic web/app workload | Virtual Machines (Standard) |
| In-memory cache, JVM heap, analytics | Virtual Machines (Memory-optimized) |
| HPC, encoding, batch | Virtual Machines (CPU-optimized) |
| BYO licence, no noisy-neighbour | Dedicated Host |
| Database with extreme IO needs | Bare Metal |
| Kubernetes worker pool | Virtual Machines + Managed K8s |
| Predictable spend + capacity | VDC + commitment plan |
| Elastic load | Auto Scaling Group + Standard VMs |
Cross-cutting capabilities¶
- Per-second billing for VMs, with a 60-second minimum
- Live migration for hypervisor maintenance — no customer downtime
- Snapshot-based backups integrate with Backup-as-a-Service
- VPC-native — every compute resource is launched into a VPC
- Tags propagate to billing, IAM, and Cost Explorer