Skip to content

Practical guide · Deployment · Shared and Dedicated

Shared SaaS or Dedicated: how to choose a deployment

Compare operations, isolation, networking, data, models, changes, and support. Record requirements, conditions, and owners before choosing a deployment.

Who it is for

Technology, security, and operations leaders who need to choose a deployment.

What you take away

A matrix of requirements, candidate deployment options, acceptance conditions, and owners.

Shared SaaS uses an environment managed by Starya. Dedicated reserves an environment for the operation, with responsibilities defined in the contracted design. The choice depends on who operates it, isolation and network boundaries, data and model paths, and the conditions for changes and support. Record these requirements before deciding.

Who operates it, and what needs to be ready?

In Shared SaaS, Starya manages the infrastructure and organizes deployment and updates for the contracted environment. The company remains responsible for defining users, authorized sources, integrations, and usage criteria. This is a candidate deployment option when this operating model meets the requirements and the team can work with the available configurations and conditions.

In Dedicated, agree on who provisions, manages access, monitors, updates, and recovers each component. The deployment option alone does not determine whether the infrastructure will be managed by Starya, by the customer, or with shared responsibilities. In an assisted deployment on the company’s infrastructure, the customer’s team manages the resources and access under its responsibility, with support from Starya according to the scope.

Before choosing, confirm available capacity, connectivity, identities, integrations, observability, and technical owners. Dedicated on OCI, AWS, or on-premise is subject to technical assessment. Terraform and Helm can support provisioning and configuration; using them does not replace defining who maintains the environment.

Which isolation and network boundaries are needed?

For Shared SaaS, check how the environment separates organizations, restricts access, and connects authorized systems. A requirement for a private network, a specific region, or restricted access needs technical confirmation in the offering; the word SaaS does not answer these questions.

For Dedicated, describe the intended perimeter and which resources are exclusive to the operation: application, processing, storage, and networking. Confirm the dependencies that remain shared or external. Dedicated does not imply physical isolation, exclusive administration by the customer, or offline operation.

When connectivity is restricted, also test updates, models, monitoring, and support. If the required network cannot reach a necessary dependency, record the incompatibility and assess another design or a smaller scope before releasing the operation.

Where do data and model calls travel?

In both deployment options, trace information from input through storage, search, model, target system, copies, and support. Record region, purpose, access, retention, and permitted outbound flows. The application’s location does not guarantee that all processing stays in the same environment.

Shared SaaS routes and providers need to be checked in the contracted design. In Dedicated, confirm what stays in the reserved environment and which external services are used. If a model receives document excerpts, that outbound flow needs to be authorized even when the original files remain on the company’s infrastructure.

How do you agree on changes, support, and recovery?

In Shared SaaS, confirm how Starya communicates updates, which configurations the company controls, and how integration testing and failures are handled. In Dedicated, explicitly assign installation, change windows, updates, monitoring, and version rollback across the teams. A change to the model or data source may also require review.

For both, agree on a support channel, authorized technical access, incident communication, and owners for continuity. Define how the team checks a failure, restores components, and reconciles effects in target systems. Timelines, coverage, and service conditions need to be agreed; they cannot be inferred from the deployment option’s name.

How do you finalize the decision and contracting?

Fill in one matrix row for each requirement. A candidate only moves forward when there is a verifiable condition, an owner, and evidence that the condition has been met. Use the same scenario to test access, networking, data flows, updates, and recovery. Keep outstanding items visible instead of turning a preference into approval.

The commercial plan is independent of the deployment option. Enterprise and Custom are the current commercial offerings; scope needs to be agreed without assuming an automatic association with Shared SaaS or Dedicated. The contracting channel also does not determine where the operation runs. Bring the matrix to the technical and commercial discussion to confirm viability, responsibilities, and conditions.

Illustrative example · no customer data

A procedure lookup with an outbound data restriction

Fictional example: a company wants to look up internal procedures. The team prefers to reduce infrastructure administration but requires that only authorized excerpts be sent to an approved model. Another requirement is accessing a repository over a restricted network.

Shared SaaS remains a candidate if connectivity and the model route meet the requirements. Dedicated is assessed if the required perimeter calls for another design. Neither choice is approved until networking, access, and processing have been verified. The assignments below are proposals for the example, not commitments of an offering.

Procedure lookup decision matrix · fictional example
RequirementCandidate deployment optionAcceptance condition / outstanding itemOwner
Reduce infrastructure administrationShared SaaSConfirm managed activities, available configurations, and integration dependencies.Customer IT and Starya
Access the repository over a restricted networkShared SaaS or DedicatedValidate connectivity and isolation in the proposed design; network testing is still pending.Networking and security team
Process only authorized excerptsShared SaaS or DedicatedCheck the model destination, region, access, and retention; block unapproved outbound flows.Data and integration owner
Reserve resources for the operationDedicatedDefine which resources are exclusive and which dependencies are external; assess OCI, AWS, or on-premise.Architecture and infrastructure team
Update with a recovery optionShared SaaS or DedicatedAgree on communication or a change window and test version rollback and reconciliation.Application and operations owner
Receive support without broad access to dataShared SaaS or DedicatedAgree on a channel, coverage, and authorized temporary access; responsibilities still need approval.Security and support owner

Record the candidate, condition, and owner for each requirement of your operation. A pending row does not prove viability. If the network or processing restriction cannot be met, adjust the design or reduce the scope before contracting.

When the approach needs to change

If an approved model or a mandatory network limits the options, start with that dependency. If there is no team to take on the tasks assigned to the customer, review the distribution of responsibilities. If neither deployment option meets a mandatory requirement, keep the decision pending and validate a technical alternative; changing only the environment’s name does not resolve the restriction.

A point to watch: Choosing by the environment name without checking data flows, responsibilities, and dependencies.

The worksheet is ready to move forward when…

  • Each requirement has a candidate deployment option, an acceptance condition, and an owner.
  • Isolation, networking, data, and models have been assessed in the specific design.
  • Operations, changes, support, and recovery have agreed responsibilities.
  • Outstanding items and required tests are recorded before the deployment decision.

Compare the matrix with Deployment and data and the deployment guide. Bring the requirements and outstanding items to confirm the technical and commercial scope.

Your worksheet

Complete it with your team.

Record what you know and what still needs confirmation. Answers stay in memory and are not submitted.

Standalone HTML file with the full guide, your answers and an option to print / save as PDF. It does not inspect systems or verify controls.

References for further reading

  • Starya · Deployment and data ↗

    Comparison of responsibilities in managed SaaS and assisted deployment on the company’s infrastructure. Availability and scope depend on technical assessment.

  • Starya · Deployment guide ↗

    A roadmap for mapping components, data flows, dependencies, and owners. The scenario presented is illustrative.

Further reading

Continue exploring this context

Product and delivery

Enterprise and Custom: how to buy

Explore Enterprise, Starya’s business offering, and Custom for tailored requirements. Define scope, deployment model and terms in a proposal.

Deployment
Reference material

Continuity

Applied AI at Starya

Connect this worksheet to the products, deployment conditions and assessment of your operation.