Skip to content
Technology transformation expertise firm · ARDIANE Group Innovate, secure, succeed

Cloud and platforms

Retaining several cloud providers: multicloud strategy put to the test of an executive committee

In brief

A multicloud strategy deliberately spreads an organisation's uses across several cloud providers, for a named reason and at the price of an accepted complexity. Management Square reviews here the questions senior management would do well to ask before retaining one more provider: rationale, cost, portability, applicable texts and governance.

Page contents

Several cloud providers often coexist within an organisation before it has settled on any multicloud strategy. An acquisition brought its own, a business department subscribed to software as a service, the data team chose another provider’s platform. De facto multicloud precedes the decision.

A multicloud strategy consists in deciding which uses go to which provider, for what reason, and what complexity the organisation agrees to pay in exchange. It is justified by a named reason: a location or qualification requirement for certain data, a service that no other provider offers, a concentration of risk to reduce, a legacy to take on. The mere fear of dependence is not enough, because each additional provider requires extra skills, an extra security foundation and extra cost tracking. The questions that follow are those an executive committee would do well to ask before signing with one more provider.

What problem should the multicloud strategy solve?

“No longer depending on a single provider” describes a concern. A reason designates what the organisation wants to be able to do. Each reason calls for a different form of multicloud, and some call for none at all.

Reason put forward What it requires in practice Is an active second provider needed?
Being able to leave the main provider Extractable data, substitutable components, exit clauses No: a prepared and tested exit is enough
Protecting a class of data A qualified or in-house environment for that data alone Yes, for that scope
Using a service specific to one provider That service, and a controlled interconnection with the rest of the system Yes, for that use
Staying in service if a provider is unavailable The same application deployed and tested at two providers Yes, and it is the most expensive form
Carrying weight in commercial negotiation Workloads that can really be moved, and a provider that knows it Not necessarily
Taking on a legacy (acquisition, a subsidiary’s choice) A decision: converge, or keep under a common framework It is already there

The first reason comes up often. It is addressed by a reversibility plan, whose content we describe in our page on cloud migration. Adding a provider does not make the first one easier to leave.

Chosen multicloud or inherited multicloud: what do we already have?

Before discussing a target, the committee needs an inventory. Which infrastructure and platform providers does the organisation use, for which applications, for what spending, and on whose decision? Software as a service is counted separately: dependence there runs through the contract and the data, with no technical foundation to operate.

Multicloud refers to the use of several cloud providers. Hybrid cloud combines a cloud and infrastructure operated in-house. The two often coexist, and the models for distributing workloads between environments are presented in our page on sovereign and hybrid cloud.

The inventory reveals two populations. Some presences have a reason that someone can state. Others result from a local initiative or an acquisition. For the latter, the alternative is simple to formulate: converge towards a provider already retained, or keep it while extending the common rules to that environment.

What does one more provider cost?

The price of the services is only the visible part. The first hidden cost is skills. Each provider has its own identity model, its network, its services and its vocabulary. Mastery of one transfers only partly to another, and spreading the same people across several environments lowers the level of expertise on each.

Next comes the foundation. Account organisation, access management, logging, default security rules: these foundations are built and maintained at each provider. The same security rule must produce the same effect everywhere, whereas it is written differently in each environment.

Exchanges between providers weigh in turn. An application whose components are spread out pays for outbound data transfers and suffers the latency of the journey. The soundest distribution keeps with the same provider whatever communicates constantly.

Finally, spending becomes dispersed. Commitment discounts are calculated on a volume, and a spread volume gives less leverage. Without common tagging and allocation rules, nobody can say what a product costs. An established FinOps practice then becomes a prerequisite.

How far should portability go?

Portability is decided application by application, on a three-level scale. At the first level, the data can be retrieved in a format usable elsewhere: this is the minimum to require everywhere. At the second, the application can be redeployed at another provider at the cost of an estimated and documented effort. At the third, it already runs at two providers and can switch from one to the other.

The third level requires sticking to what the providers have in common: containers, orchestration with Kubernetes, open source databases, infrastructure described as code with a tool such as Terraform. It rules out each provider’s own managed services, which are often the reason a provider was chosen. It also assumes regular tests, because a switchover never rehearsed remains a hypothesis.

The useful question for the committee therefore concerns scope: which applications justify the third level, and who funds the extra cost? For the others, the second level, kept up to date, generally meets the need.

What do the texts say about dependence and exit?

Regulation (EU) 2022/2554, known as DORA, applies to financial entities. Its Article 6(9) allows them to define “a holistic ICT multi-vendor strategy”, showing key dependencies on providers and explaining the rationale behind the mix retained. It is an option: the text says “may”. Article 29 requires, before entering into a contract for services supporting critical or important functions, an assessment of concentration risk and of the benefits and costs of alternative solutions, “such as the use of different ICT third-party service providers”. Article 28(8) imposes exit strategies for those same services. The regulation therefore requires an analysis and a prepared exit. It does not oblige the use of several providers.

Regulation (EU) 2023/2854, known as the Data Act, devotes its Chapter VI to switching between data processing services. Article 25 lists what the contract must provide for, including data porting and the provider’s assistance during the transition. Article 29 prohibits, from 12 January 2027, charging the customer switching charges; Article 31 excludes from this rule certain services custom-built for a particular customer. Article 34 extends some of these requirements to the simultaneous use of several data processing services. Its paragraph 2 allows, in that case, data egress charges, provided they do not exceed the egress costs incurred.

In France, a requirement for hosting on a SecNumCloud qualified service, the qualification framework of ANSSI (the French national cybersecurity agency) for cloud service providers, may lead to a second provider for part of the data. The qualification covers a specific service, and the other uses remain elsewhere.

Who will hold the whole together over time?

A multicloud strategy holds through a few written rules and through those accountable for them. The placement rule states which type of use goes to which provider and who grants an exception. It has a designated owner.

Cross-cutting functions (identities, monitoring, deployment) are not duplicated per provider. Where a platform team exists, it is the one that hides the differences behind common paths, following the logic of platform engineering.

Finally, the committee gives itself the means to check: the share of applications whose placement is justified in writing, the number of open exceptions and how long they have been open, the share of spending attributed to an owner at each provider, the date of the last exit test for the applications concerned.

The next meeting can open with the list of providers already in use, with for each one the spending, the applications hosted and the reason for its presence. The lines whose reason remains empty show where to begin.

FAQ

Frequently asked questions

Another question? Ask us

What is a multicloud strategy?

A multicloud strategy is the decision to spread an organisation's uses across several cloud providers, according to a written rule: which type of use goes to which provider, and for what reason. It differs from de facto multicloud, born of acquisitions or local initiatives, through the existence of this rule, of an owner accountable for it and of functions common to all environments.

What is the difference between multicloud and hybrid cloud?

Multicloud refers to the use of several cloud providers. Hybrid cloud combines a cloud with infrastructure that the organisation operates itself, in its own data centre or at a hosting provider. The two situations frequently coexist: an organisation may keep a data centre, use a main provider and reserve a qualified service for its most sensitive data.

Does multicloud protect against dependence on a provider?

Not on its own. Using two providers for different applications creates two dependencies instead of one. Dependence is reduced when the data can be extracted, the provider-specific components are inventoried and the exit has been tested. This reversibility work can be carried out with a single provider, and it still has to be done with several.

Does a multicloud strategy cost more than a single provider?

It adds costs that do not appear on the services invoice: skills to maintain on each provider, a security foundation to build in each environment, data transfers between providers and commitment discounts calculated on smaller volumes. These costs are justified when a precise reason offsets them, for example a requirement covering certain data or a service that no other provider offers.

Does the DORA regulation require financial entities to adopt multicloud?

No. Article 6(9) of Regulation (EU) 2022/2554 allows financial entities to define a multi-vendor strategy, without obliging them to. Article 29 requires them to assess concentration risk and alternative solutions before entering into a contract for ICT services supporting critical or important functions. Article 28 imposes exit strategies for those same services. The text requires an analysis and a prepared exit.

A project to bring to completion?

A first conversation is enough to understand your challenge, identify the right people and tell you frankly what we can bring.