Cloud decisions are often framed as a choice between old and new. That is not useful. The practical question is where each workload should run, given its performance, data, cost, connectivity and support requirements. For many organisations, the answer is a mixture.

Key takeaways

  • Decide workload by workload, not with a single organisation-wide rule.
  • Connectivity, latency and data volume can outweigh licence savings.
  • Cloud cost needs active management, not just a monthly invoice.
  • Exit and portability should be considered before migration.

Start with the workload

List the applications and services in use. For each one, record the users, data volume, performance sensitivity, integration points, legal or regulatory constraints and current support arrangement.

A file server, a payroll system, a CAD application and a public website have different needs. Treating them as one category leads to poor decisions.

When cloud is a good fit

Cloud is often a good fit for email and collaboration, websites, development and test environments, backup, and workloads with variable demand. It can reduce hardware refresh cycles and make access from multiple locations easier.

It can also provide resilience that would be expensive to build locally, provided the design uses the available regions, backup and identity controls properly.

When on-premise is a good fit

On-premise can be the better choice for workloads with large data volumes, low-latency requirements, specialised hardware, or a need to keep data in a specific location. It can also be economical when utilisation is high and the facility is already well managed.

The key word is managed. An on-premise platform that is poorly cooled, powered or backed up is not more reliable than a cloud service simply because it is local.

Why hybrid is common

A hybrid model keeps some workloads local while using cloud services for others. Identity, email and backup may sit in the cloud; a line-of-business system or large data store may remain on-premise.

Hybrid adds integration and support boundaries. Directory synchronisation, network paths, backup coverage and responsibility for each component need to be documented.

Connectivity and performance

Cloud workloads depend on the internet and the path to the provider. Check bandwidth, latency, contention and failover. A link that supports email and browsing may not support a production application or large file transfers.

Consider what happens during an outage. If the organisation cannot operate without the service, a resilient connection or a local fallback may be required.

Cost and licensing

Cloud consumption is flexible, but it is not automatically cheaper. Costs can rise through oversized instances, forgotten storage, unnecessary services, data transfer and duplicated licences.

Build a cost model that includes migration, running costs, support, training and exit. Review actual consumption against the model after go-live and adjust.

Data protection and exit

Understand where data is stored, who can access it, how it is backed up and what happens if the service is discontinued. Review the provider's terms rather than assuming they match your obligations.

Plan how data and services would be moved if you changed provider or brought the workload back on-premise. Portability is a negotiating position as well as a technical property.

A practical decision process

Assess the workload, shortlist viable options, model cost and risk, run a pilot where possible, and agree the support model before migration. Document the decision and the assumptions.

Revisit the decision when usage, cost or requirements change. Cloud, on-premise and hybrid are operating choices, not permanent identities.

Note: Cloud suitability depends on the specific workload and organisation. This guide provides a decision framework, not a recommendation to adopt or avoid a particular platform.

Need a second opinion?

Tell us what you are trying to fix, build or buy. We will give you a clear view of the options and the next step.

Contact Stevicom