HCI compute with S3

This scenario extends the HCI compute model with object storage. Every node runs compute and storage; the S3 object (OS) and name (NS) services run across the cluster, and S3 gateways (GW) run on the worker nodes. Compute and S3 use separate tiers to allow compute and object-storage resources to be managed independently.

Cluster layout

			S3 GW, NS, OS	S3 GW, NS, OS	S3 GW, NS, OS
Compute worker	Compute worker	Compute worker	Compute worker	Compute worker	Compute worker
Compute controller	Compute controller	Compute controller
Virtuozzo Storage	Virtuozzo Storage	Virtuozzo Storage	Virtuozzo Storage	Virtuozzo Storage	Virtuozzo Storage
Management	Management	Management
Node1	Node2	Node3	Node4	Node5	Node6
Bond 1 (Storage) 2 × 25 GbE LACP Non-routable, MTU 9000	Storage · Internal management · OSTOR private (S3 internal)
Bond 2 (Overlay + Public) 2 × 10 GbE LACP VLAN-tagged, public via NAT	VM private (VXLAN) · VM public · S3 public · Admin panel · Compute API · SSH

Storage policy

Define two storage tiers, each with its own storage policy:

• NVMe tier for compute: all-flash storage for VM volumes, with the same EC 3+2 and host failure domain policy as in the HCI compute scenario.
• HDD tier with SSD cache for S3: high-capacity HDDs fronted by an SSD cache for object data, with a capacity-oriented EC policy.

Because both tiers use EC, the RAM values below include the per-node storage-client reservation described in CPU and RAM requirements.

Per-node hardware

Resource	Management node		Worker node
	Minimum1 The minimum values use the same base reservations as the HCI compute scenario, plus the S3 service reservation and erasure-coding storage-client reservation per node. The S3 reservation is small compared with the compute reservation, so the rounded minimums match the HCI compute scenario.	Recommended	Minimum	Recommended
CPU cores (physical)	16	32+	8	24+
RAM (ECC, required)	32 GB	96 GB+	16 GB	64 GB+
System disk	120 GB SSD	2 × 480 GB SSD (RAID-1)	120 GB SSD	2 × 240 GB SSD (RAID-1)
Metadata (MDS)	On system disk (combined), first 5 nodes		On system disk (combined), first 5 nodes
NVMe tier (compute)	≥1 NVMe SSD per node		≥1 NVMe SSD per node
HDD tier (S3)	HDDs + 1 SSD cache per 4–12 HDDs		HDDs + 1 SSD cache per 4–12 HDDs
Backend network	2 × 10 GbE (LACP)	2 × 25 GbE (LACP)	2 × 10 GbE (LACP)	2 × 25 GbE (LACP)
Frontend network	10 GbE	2 × 10 GbE (LACP)	10 GbE	2 × 10 GbE (LACP)

Sizing S3

S3 service reservations are usually smaller than compute reservations, but they must still be added separately. The S3-only scenario provides the formula and a worked example. In this scenario, the S3 services run alongside compute, so add the S3 per-node reservation on top of the compute figures above. The scenario table already includes the EC storage-client reservation; when sizing a different topology, add it separately.