Tenancy Types¶
What Tenancy Types are Supported by Omnistrate¶
Tenancy describes the boundary between customer deployments: pods, virtual machines, networks, Kubernetes clusters, or cloud accounts. A SaaS product can serve many tenants at any of these levels, including when each deployment has dedicated infrastructure.
In Omnistrate configuration, Multi-Tenancy specifically means pod-level infrastructure multi-tenancy: separate deployments run in separate pods while sharing VMs. For dedicated machines, networks, or Kubernetes clusters, choose Dedicated Tenancy and configure the required isolation level. These levels are suboptions, not additional tenancyType values. For dedicated customer accounts, use the BYOC deployment model.
| Tenancy type | Isolation suboptions | Configuration |
|---|---|---|
| Infrastructure Multi-Tenancy | Separate pods on shared VMs | OMNISTRATE_MULTI_TENANCY in the Compose specification; configure CPU and memory requests and limits. |
| Dedicated Tenancy | Dedicated VMs; optionally dedicated networks or Kubernetes clusters. Account-level isolation is available through BYOC. | OMNISTRATE_DEDICATED_TENANCY in the Compose specification for dedicated VMs, plus network or cluster-placement configuration. Use BYOC for dedicated customer accounts. |
| Custom Tenancy | Isolation defined by your deployment artifacts | CUSTOM_TENANCY for Helm, Kustomize, Terraform, and OpenTofu in the Service Plan specification; not a Compose tenancy option. |
Deployment models and tenancy are separate decisions. Hosted SaaS runs in your cloud accounts; BYOC runs in customer-owned accounts. Neither account ownership nor dedicated VMs alone implies a dedicated network or Kubernetes cluster.
For detailed guidance, see the Infrastructure Multi-Tenancy Overview and Dedicated Tenancy Overview.
What Does Sharing a VM Mean?¶
A VM (virtual machine) is a cloud machine with its own CPU and memory capacity. A pod runs an application component, such as your application or database, on that machine. A deployment is one provisioned copy of your service for a customer; it can contain several pods running across several machines.
- Shared machines: Pods from different deployments can run on the same VM and use its CPU and memory capacity. The pods are separate; sharing a machine does not mean sharing an application process or database.
- Dedicated machines: A deployment's workload VMs run only that deployment's pods. Other deployments cannot use those VMs, including other deployments belonging to the same customer.
The diagrams below show the same three deployments: D1 and D2 belong to Customer A; D3 belongs to Customer B. Each has its own application and database pods. Only the placement of those pods on machines changes.
Dedicated Tenancy¶
In Dedicated tenancy model, each deployment receives their own dedicated infrastructure stack (VMs). Your control plane, generated with Omnistrate, manages provisioning, scaling, and lifecycle operations. A tenant can have multiple deployments; the dedicated compute boundary is per deployment, not one shared stack per tenant.
Dedicated VMs¶
Choose OMNISTRATE_DEDICATED_TENANCY to give each deployment dedicated VMs and attached storage. VMs are not shared with other deployments, including other deployments belonging to the same tenant. The Kubernetes host cluster, network, and provider cloud account may still be shared unless you configure a broader boundary.
Dedicated Network¶
For Hosted SaaS, enable Customer Networks with the CUSTOM_NETWORKS feature. Each customer chooses or creates a network, and deployments using the same customer network can share that network and its host cluster. Use different networks when separate network boundaries are required. This does not create a dedicated cloud account.
For a customer-owned existing network, use BYO-VPC with BYOC instead.
Dedicated Kubernetes Cluster¶
To give each deployment a dedicated Kubernetes host cluster, combine Dedicated Tenancy with CUSTOM_DEPLOYMENT_CELL_PLACEMENT and maximumDeploymentsPerCell: 1. A deployment cell is the host-cluster placement unit; it is not a synonym for infrastructure multi-tenancy or in-app multi-tenancy. See Custom Deployment Cell Placement for configuration and limitations.
Dedicated Account¶
Use BYOC for dedicated-account isolation. Each customer supplies and owns the cloud account where their deployments run, while your generated control plane manages the service. Onboard a separate customer account for each required account boundary.
See the Dedicated Tenancy Deployment Guide for VM, network, Kubernetes cluster configuration and account-level isolation.
Dedicated VM Architecture¶
Each outlined machine below has CPU and memory reserved for one deployment. D1 and D2 use separate machines even though both belong to Customer A. Each deployment still has the same application and database components.
This example uses two VMs per deployment to make the boundary visible; actual VM and pod counts depend on resource sizing and replicas. Dedicated compute does not automatically dedicate the surrounding network, Kubernetes cluster, or cloud account.
How Dedicated Tenancy Works¶
- Compute isolation: Each deployment uses dedicated VMs and attached storage. Broader network, Kubernetes, and account boundaries require the corresponding configuration above.
- Customization: Expose supported instance types, storage settings, and network choices through your plan and deployment inputs.
- Managed lifecycle: Your generated control plane provisions, monitors, scales, upgrades, and maintains each deployment.
- Application security: Dedicated infrastructure complements application authorization and data isolation; it does not replace them or establish compliance by itself.
Dedicated tenancy is a good fit for plans that need dedicated compute capacity, deployment-specific lifecycle policies, or stronger infrastructure boundaries than shared VMs provide.
Configuration Example:
Infrastructure Multi-Tenancy¶
Select OMNISTRATE_MULTI_TENANCY for pod-level isolation on shared VMs. Each customer deployment has its own application pods; multiple deployments can be scheduled onto the same underlying machines. Customers do not need to share a single application process or database for this infrastructure model to work.
In this example, Machine 1 runs application pods from D1, D2, and D3; Machine 2 runs their database pods. Each machine's CPU and memory capacity is shared across those deployments. Matching colors identify pods from the same deployment, not a shared application or database.
This is an illustrative placement, not a requirement to use exactly two VMs or to separate application and database workloads onto different machines. Actual placement depends on capacity, resource requests, and replicas.
How Infrastructure Multi-Tenancy Works¶
Resource sharing and bin-packing: Your generated control plane places deployment pods on shared VMs based on their CPU and memory requests. Sharing VM capacity can improve utilization and reduce infrastructure cost.
Capacity and scaling: Configure deployment resource requirements and autoscaling for the plan. Shared capacity must accommodate normal operation, peak demand, and upgrades.
Isolation boundary: Separate pods and resource limits provide logical workload separation, not dedicated machines, networks, clusters, or accounts. Shared VMs can still experience contention. Your application must protect credentials, data, endpoints, and access between deployments.
Broader isolation: If the customer requires dedicated VMs, networks, or Kubernetes clusters, use Dedicated Tenancy with the appropriate suboption rather than this pod-level sharing mode. For dedicated customer accounts, use BYOC.
Configuration Example:
x-omnistrate-service-plan:
tenancyType: 'OMNISTRATE_MULTI_TENANCY'
services:
postgres:
deploy:
resources:
limits:
cpus: '0.50'
memory: 50M
reservations:
cpus: '0.25'
memory: 20M
You can also configure resource requests and limits as API parameters:
services:
postgres:
x-omnistrate-compute:
resourceRequestMemoryAPIParam: resourceRequestMemory
resourceRequestCPUAPIParam: resourceRequestCPU
resourceLimitMemoryAPIParam: resourceLimitMemory
resourceLimitCPUAPIParam: resourceLimitCPU
x-omnistrate-api-params:
- key: resourceRequestMemory
name: Memory Request
description: Memory Request for the Postgres instance
type: String
modifiable: true
required: true
export: true
defaultValue: 100M
- key: resourceRequestCPU
name: CPU Request
description: CPU Request for the Postgres instance
type: String
modifiable: true
required: true
export: true
defaultValue: 100m
- key: resourceLimitMemory
name: Memory Limit
description: Memory Limit for the Postgres instance
type: String
modifiable: true
required: true
export: true
defaultValue: 2048M
- key: resourceLimitCPU
name: CPU Limit
description: CPU Limit for the Postgres instance
type: String
modifiable: true
required: true
export: true
defaultValue: "4"
Architecture Selection: You can specify processor architecture using the platform attribute:
Use cases: Infrastructure multi-tenancy is suitable for cost-efficient plans and high deployment density when pod-level separation and shared machine capacity meet the customer's requirements. See the Infrastructure Multi-Tenancy Deployment Guide for setup and validation steps.
In-App Multi-Tenancy¶
In-app multi-tenancy means one application deployment serves multiple tenants and enforces separation inside the application, such as tenant-scoped authentication, authorization, and data access. It is independent of whether the deployment runs on shared or dedicated VMs.
An application may also group its tenants into bounded application-level cells. That application architecture is not the same as Omnistrate's deployment cells, which are Kubernetes host-cluster placement units, and neither term defines pod-level infrastructure multi-tenancy.
Custom Tenancy¶
Custom tenancy is used for Helm, Kustomize, Terraform and OpenTofu deployments where the tenancy model is defined within the deployment components themselves, rather than being controlled by Omnistrate.
How Custom Tenancy Works¶
Custom-Defined Tenancy: The tenancy model, isolation boundaries, and resource allocation are defined within your Helm charts, Terraform configurations, or Kustomize overlays.
Omnistrate Orchestration: Omnistrate handles the deployment, lifecycle management, and scaling of your components, but respects the tenancy rules you've defined.
Flexibility: You have complete control over how tenants are isolated, how resources are shared or dedicated, and how scaling occurs.
Component-Level Control: Different components within your service can have different tenancy models based on your specific requirements.
Use Cases: Custom tenancy is ideal when you have specific tenancy requirements that don't fit the standard multi-tenant or dedicated models, need fine-grained control over resource allocation, or are migrating existing infrastructure-as-code deployments to Omnistrate.
Choosing the Right Tenancy Type¶
| Tenancy Type | Isolation Level | Cost Efficiency | Management Overhead | Best For |
|---|---|---|---|---|
OMNISTRATE_DEDICATED_TENANCY | Dedicated VMs; optional network, Kubernetes cluster, or customer account boundaries | Lower utilization than shared VMs; depends on isolation level | Depends on isolation level | Dedicated capacity, enterprise plans, customer-specific infrastructure boundaries |
OMNISTRATE_MULTI_TENANCY | Separate pods on shared VMs | High deployment density | Shared capacity planning | Cost-efficient plans where pod-level isolation is sufficient |
CUSTOM_TENANCY | User-defined | Variable | Medium | Specific tenancy requirements, existing IaC, fine-grained control |
The choice of tenancy type depends on your specific requirements for isolation, cost, compliance, and operational complexity. You can also mix different tenancy types within the same service for different tiers or customer segments.