On premises, in the cloud or hybrid: there is no single answer that fits the whole company. The choice is made application by application, on a few criteria that this guide reviews.
On-premises, cloud, hybrid: what do they mean?
On-premises infrastructure runs on servers you own, installed on your premises or in a room you rent. In the cloud, you rent resources from a provider and pay according to usage. Hybrid infrastructure combines the two: some applications stay on your servers, others run in the cloud, and the two environments are connected securely. Many organisations end up with a hybrid architecture, because their applications do not all have the same constraints.
The differences in one table
| Criterion | On premises | Cloud | Hybrid |
|---|---|---|---|
| Spending | Upfront investment, then upkeep | Monthly expense, based on usage | Both, depending on the application |
| Data location | On your premises | In the region you choose at the provider | Depends on how sensitive the data is |
| Scaling | By buying hardware | On demand | On demand for the cloud part |
| Operations | Your team or a managed service provider | Shared with the provider | Two environments to administer |
| Dependency | On the hardware and its warranties | On the provider and its pricing | Spread between the two |
1. Start from the applications, not the technology
List your applications, then note for each one: who uses it and from where, when it must be available, what data it holds, and which other applications it exchanges with. This inventory reveals groups: applications that gain from moving to the cloud, those that must stay on premises, and those better replaced.
2. Look at where the data must be
Data location is often the first criterion. In Morocco, law 09-08 governs the processing of personal data and its transfer abroad, under the supervision of the CNDP; for people located in the European Union, the GDPR applies. Some sectors add their own rules. Check these points with your legal adviser before choosing where to host.
3. Compare costs over several years
On-premises infrastructure requires an upfront investment, then upkeep costs: warranties, power, cooling, and hardware replacement at end of life. The cloud is paid by usage: it avoids the investment, but the bill follows consumption and has to be monitored. The comparison only makes sense over several years, counting operations time, licences and the cost of a possible exit as well.
4. Measure how much the load varies
An application whose load varies widely, by season or campaign, benefits from the cloud: capacity adjusts to demand. An application with a stable, predictable load can cost less on servers sized once and for all.
5. Take the network into account
A cloud application depends on each site’s internet connection. Check the available bandwidth, plan a backup link for critical sites, and measure response times for highly interactive applications or those handling large files. On a remote site, hybrid backhaul combines satellite, microwave and fibre.
6. Decide who runs it
On premises, everything is yours to manage: hardware, system, backups, security. In the cloud, the provider manages the lower layers, but configuration, access, backups and costs remain your responsibility. In both cases, operations can be handed to managed IT services.
7. Plan backup and recovery
Wherever you host, apply the 3-2-1 rule: three copies of the data, on two different media, one of them off site. The cloud does not replace a backup: an accidental deletion or an attack spreads there too. For each application, set the acceptable downtime and data loss, then test the restores. See our cybersecurity and recovery services.
8. Keep a way out
Open data formats, exportable backups, documented architecture: reversibility is prepared at the design stage. It applies to a cloud provider as much as to a hardware manufacturer.
What about artificial intelligence?
AI projects raise the same question. When data can be processed at a provider, the cloud avoids buying compute servers. When it must stay on your premises, you need to plan compute and storage resources on site. See our AI solutions.
Which option for which situation?
| Situation | Option to consider first |
|---|---|
| Variable load, teams spread across several sites | Cloud |
| Data that must stay on your premises | On premises |
| Legacy application that is hard to move | On premises, then rewrite or replace |
| Standard need, such as email or office tools | Online service |
| Several of these situations at once | Hybrid |
The checklist in short
- Each application is described: users, availability, data, exchanges.
- Data location has been checked for each one.
- Costs are compared over several years, exit included.
- Bandwidth and the backup link have been checked for each site.
- Operations are assigned: internal team or managed services.
- Backups follow the 3-2-1 rule and restores are tested.
- Reversibility is planned from the design stage.
To design and deploy the chosen architecture, see our IT infrastructure services, with hardware supplied by our IT hardware and licensing practice. If you choose the cloud, cloud migration and our cloud migration checklist take over. Our IT consulting team can formalise the choice. To talk it through, contact us.