Skip to main content

Topology

A Gallium Cluster can be deployed in several topologies depending on your requirements. The default topology uses three identical hypervisors, while Enterprise Flex customers have access to additional configurations for different scale and availability needs.

Server Roles​

Every server in a Gallium Cluster is assigned one or more roles:

  • Control Plane — Manages cluster operations and stores configuration information.
  • Worker — Runs virtual machine workloads.
  • Backup Repository - An S3 compatible immutable storage repository running on dedicated physical disks for the purposes of storing backups

In the default three-hypervisor topology, every hypervisor holds the Control Plane and Worker role.

Three Hypervisors​

The default topology consists of three identical hypervisors, each running all three roles. The cluster operates in an N+1 failover capacity — if one hypervisor goes offline, the remaining two hypervisors continue running all workloads.

Three hypervisor topology

Four to Nine Hypervisors​

For larger deployments, a cluster can scale from four to nine hypervisors. In this topology, a subset of hypervisors are assigned the Control Plane role in addition to the Worker role, while the remaining hypervisors run as dedicated Workers only.

Four to nine hypervisor topology

Control Plane​

Either 3 or 5 hypervisors run the Control Plane roles. This count must be odd because configuration database requires a majority quorum — 3 members tolerate 1 failure, and 5 members tolerate 2 failures. The Control Plane role not automatically reallocated if a hypervisor goes offline.

Disk Pool Replication​

Disk pools in this topology can be configured with either 2-copy or 3-copy replication. The number of copies, combined with the total number of hypervisors, determines the cluster's failover capacity:

  • 4 hypervisors with 2-copy replication — The cluster can tolerate the loss of 1 hypervisor (N+1). Three hypervisors run Control Plane and Worker roles; the fourth is a dedicated Worker.
  • 6 hypervisors with 3-copy replication and 5 control planes — The cluster can tolerate the loss of 2 hypervisors simultaneously (N+2).

Two Hypervisors + Witness​

info

This topology requires an Enterprise Flex subscription.

Two full hypervisors run all three roles and handle VM workloads. A third lightweight Witness participates in cluster quorum and runs the control plane, but does not run VM workloads.

Two hypervisors plus witness topology

The Witness:

  • Can be a virtual machine rather than dedicated hardware.
  • Must be on the same Layer 2 network segment as the other hypervisors.
  • Must have a network latency of 40ms or less to the other hypervisors.

This topology can optionally use a dedicated storage network with cables connected directly between the two hypervisors for storage replication traffic. This removes the need for 10 Gbps switching infrastructure, as storage traffic bypasses the network switches entirely.

Two Hypervisors + Backup Repository​

info

This topology requires an Enterprise Flex subscription.

Similar to the Witness topology, two full hypervisors run all three roles and handle VM workloads. The third server acts as both a quorum participant and a backup storage target.

Two hypervisors plus backup repository topology

The Backup Repository has the same requirements as a Witness — it can be a VM and must be on the same Layer 2 segment within 40ms of the other hypervisors.

As with the Witness topology, a dedicated storage network can be configured with cables directly between the two hypervisors, removing the need for 10 Gbps switching infrastructure.