What Is Cloud and Who Runs Each Layer?

Business owner and IT advisor reviewing a cloud responsibility diagram on a laptop

Cloud computing is a way for a business to use shared computing resources without owning all the underlying hardware. What is cloud, in practice, is a split of work. A provider runs a shared pool of servers, storage, and software. Your team still decides who can sign in, which data belongs there, and what happens when something breaks. NIST Special Publication 800-145 describes that model as on-demand access to a shared pool of computing resources that can be turned up or released with little hands-on work from the provider. The useful question is not the definition. It is which layer you still run.

Talk with IGTech365 about this article

Send a note and we will reply from this page.

This field is for validation purposes and should be left unchanged.
Name(Required)

See how business cloud infrastructure supports that split

What is cloud responsibility in each service model?

Cloud services come in three layers. NIST names them Software as a Service, Platform as a Service, and Infrastructure as a Service. Each one moves a different amount of the stack to the provider. None of them moves the business decisions.

Software as a Service is the layer most 10-to-150-person companies already use. Microsoft 365 mail, file sharing, and line-of-business apps are typical examples. You use the application. The provider runs the servers, storage, operating system, and the application itself. You still set users, sharing rules, and the few settings the app lets you change.

Platform as a Service sits under custom software. Your developers deploy an application the company built or bought. The provider runs the network, servers, operating system, and storage. Your team owns the application code and the settings of the place it runs.

Infrastructure as a Service is the closest to a server you rent. You provision processing, storage, and networks, then install what you need on top. The provider owns the physical building and the hardware pool. You own the operating system, the applications, and most of the configuration. That is also why an infrastructure project feels different from turning on a mailbox.

A Tampa law firm that already lives in Microsoft 365 is mostly on the software layer. Adding a case-management app the vendor hosts is still software, unless the firm is the one deploying the code. Standing up a file server or a line-of-business database in Azure is infrastructure, and the firm keeps the operating system. A clinic that wants a patient portal built for its own workflow is closer to the platform layer. The same company can hold all three. The labels only help if the project brief names one of them.

IGTech365 is a Tampa-based Microsoft partner, founded in 2015, that manages this stack for growing companies. The work covers Microsoft 365, cloud, security, backup, and the local systems those services still depend on. Healthcare, legal, and manufacturing clients are the usual fit, because those teams feel downtime and compliance gaps in the same week. The point of naming the layer is to stop a project from mixing all three into one vague “move to the cloud” request.

What changes when you move one layer down

Drop from software to platform, or from platform to infrastructure, and the weekly work changes with it. On software, a busy week is new hires, shared links, and a mailbox that will not open. On a platform, someone has to know which application version is live and who can publish the next one. On infrastructure, someone patches the operating system, watches disk space, and decides when the virtual machine is allowed to grow.

That is also where quotes go wrong. A subscription that looks cheap on the software row can hide an infrastructure row nobody staffed. The reverse happens too. A company keeps paying for server care on a workload that is already a finished application. Ask the vendor, or your own IT lead, to point at the row before you approve the spend.

Who still owns access, data, and recovery?

The split above is the shared responsibility line, and it does not move just because a brochure says the provider is secure. In every model, the business keeps three jobs.

  • Access. You decide who has an account, which roles they get, and when a former employee loses them. A provider can offer the switch. It cannot know your staffing changes.
  • Data. You decide what is allowed in the service, how it is classified, and who it may be shared with. Location controls, when the contract offers them, are still a business choice.
  • Recovery. You decide how fast a mailbox, file library, or server must come back, and you test that the copy actually restores. Provider retention is not the same as a backup you control.

NIST’s measured-service idea is the other half of that ownership. Usage is metered and can be reported, which is useful only if someone on your side reads it. A service that looks unlimited on day one can become a surprise invoice after a busy season. Healthcare, legal, and manufacturing teams feel that fastest, because their demand and their compliance duties both move.

Access is a staffing problem, not a feature list

Most cloud trouble at this size starts with an account that outlived the employee. A provider will not know that a nurse, a paralegal, or a shift lead left on Friday. Your side has to remove the account, pull shared mailboxes back, and check what that person could still open. Multi-factor sign-in only helps if it is turned on for the people who touch client files, patient records, or production data.

Role design matters as much as the login. A partner who can export the whole document library is a different risk from a receptionist who can book appointments. Write the roles in plain language first. Then match them to the product. If the product cannot express the role, fix that gap before go-live. Do not leave it as a setting to clean up later.

Data and recovery need an owner with a calendar

Data ownership is the list of what may live in the service and what must stay on systems you already control. A manufacturer may keep machine programs and quality records on local storage, and use the cloud for mail and shared drawings. A firm may keep the practice app in a hosted service and still need a copy it can restore without opening a ticket. Write that list down. A verbal “we are in the cloud” is how a file ends up in the wrong place.

Recovery needs a date on the calendar, not a sentence in a proposal. Pick one mailbox, one shared library, and, if you still have one, one server. Restore each of them to a point you choose. Record how long it took and who did the work. Provider retention, legal hold, and a backup you control are three different promises. Only the test tells you which one you actually have.

The monthly usage read belongs on that same calendar. Storage that grew because someone synced a laptop folder, a test database left running, or a shared mailbox nobody archives will show up there. IGTech365’s clients often want IT to disappear into the background. That only happens after a person, inside the company or at the partner, owns this review.

If the open question is whether the servers themselves should move, that is a different decision from this split. Start with whether a small business should move its servers to the cloud, then come back to who will run the layer you keep.

How do the layers compare for a growing business?

Use the table as a buying filter, not a product catalog. The right row is the one that matches the job. Mail for the whole staff is not the same purchase as a server you still patch.

Question SaaS PaaS IaaS
What you get. A finished application. A place to run your application. Compute, storage, and network you configure.
What you still run. Users, data, and app settings. The application and its configuration. Operating system, apps, and most settings.
What the provider runs. The full stack under the app. Network, servers, OS, and storage. The physical pool and facilities.
Typical business use. Email, files, and line-of-business apps. A custom or industry application. A server or network you are not ready to retire.
Main thing to verify. Identity, sharing, and restore tests. Who patches the app and the platform. Who patches the OS and watches the bill.

A company can use all three at once. Microsoft 365 for mail, a hosted line-of-business app, and one remaining Windows server is a normal mix for a firm in the 10-to-150-employee range. The mistake is assigning one owner, one backup plan, and one monthly review to all three.

What is cloud used for in a growing business?

Once the layer is clear, the cloud stops being a destination and starts being a way to keep work moving. Staff reach mail and files from the office, home, or a job site without a server sitting under a desk. A seasonal spike in a warehouse or a clinic can add capacity and release it later, which is the elasticity NIST describes, not a promise of free growth.

That support still depends on the systems around the cloud service. Identity has to match the people you employ. The network has to reach the users. Backup has to cover the data the provider does not promise to keep. IGTech365’s managed IT work is built around that whole path: monitoring, Microsoft 365, security, and backup for companies that cannot afford a stalled day. The same team also covers low-voltage cabling and VoIP support when the office itself is part of the problem. A cloud project should not stall on a dead drop or a phone that cannot transfer.

A normal week after the layer is named

Picture a 40-person professional firm with one internal coordinator and a managed IT partner. Monday, a new hire needs mail, files, and the practice application, and a departing employee needs those same doors closed. Wednesday, someone checks storage growth and a license that is assigned to a shared mailbox instead of a person. Friday, a restore of one folder proves the backup from Sunday still opens. None of that is a migration. It is the work the cloud leaves with you.

A plant or a clinic has a different week, and the same split. The line or the front desk has to stay up when a workstation fails. The cloud service helps only if identity, the network path, and the restore copy were already sorted. Adding capacity for a seasonal rush is useful. It does not replace a person who can say which system is allowed to grow and which one is not.

Read that as the practical meaning of the five characteristics in NIST Special Publication 800-145, not as a glossary. On-demand self-service means a lead can request capacity without opening a facilities project. Broad network access means staff get in from a desk, a home office, or a phone. Resource pooling means you are sharing hardware with other customers and should not assume a private room in the data center. Rapid elasticity is the seasonal bump. Measured service is the bill and the usage report. If any one of those is missing, you may still have a remote server. You do not yet have the cloud model NIST defined.

For the physical and virtual pieces under those services, read what cloud infrastructure includes. This page stays on the responsibility line. That page stays on compute, storage, networking, and the plan before a move.

Diagram separating provider-run cloud layers from the access, data, and recovery duties a business keeps
Each cloud layer shifts a different part of the stack, and the business keeps access, data, and recovery in every layer.

What should you settle before a cloud change?

A short list beats a long strategy deck. Settle these five points before you approve a project or a new subscription.

  1. Name the layer. Say whether the purchase is SaaS, PaaS, or IaaS. If the answer is “the cloud,” the scope is not ready.
  2. Name the owner. One person or one partner owns accounts, sharing rules, and the monthly usage review.
  3. Name the restore target. Write how many hours of mail, files, or a server the business can lose, then test a restore.
  4. Name what stays local. Some workloads belong on systems you already run. Hybrid is a design choice, not a failure to modernize.
  5. Name the next step. A migration plan is separate from this decision. Scope it only after the layer and the owner are clear.

IGTech365 starts infrastructure work the same way: look at what is working, what is failing, and where the business is trying to go, then build the roadmap. That assessment is a better first meeting than a tool demo. Leadership there comes from operations, so the question is how the workday runs, not which logo sits on the server. When the roadmap includes a move, cloud migration services cover Microsoft 365, Azure, files, and apps, with a plan, a migration, and support after the cutover. Ongoing support after the move sits with managed IT services, so the responsibility split does not end on launch day.

Questions worth asking in the first meeting

Bring these to whoever will run the layer you keep. They are plain on purpose. A fuzzy answer is a sign the scope is still a slogan.

  • Which of the three layers is this purchase, and what does our side still patch or configure?
  • Who removes access on the day someone leaves, and how is that checked?
  • Which data stays on systems we already run, and who approves an exception?
  • What can we restore ourselves, and when was the last test?
  • Who reads the usage report each month, and what number would make us call a stop?

You do not need a perfect score on day one. You need named owners. A company with 10 to 150 employees rarely has a full in-house IT department, which is why the partner has to pick up the weekly work instead of handing over a diagram. If the answers point at a move, schedule that project on its own. If they point at cleanup inside Microsoft 365 or a server you still run, do that first. Both are legitimate outcomes of asking what is cloud for this business, rather than for the industry in general.

If you want the plain-language definition, the practical uses, and the hybrid context, use the existing guide on what the cloud is. Come back here when the decision is who runs which layer.

Want a second set of eyes on which cloud layer you still run? Call IGTech365 at (866) 365-7798 or send a note through the contact page.

Frequently asked questions

What is cloud in one sentence?

Cloud is on-demand access to a shared pool of servers, storage, and applications that you can scale up or release, while your business still owns access, data, and recovery.

Does cloud mean the provider handles security?

No. The provider secures the service it runs. You still control accounts, sharing, and the backups you need beyond the provider’s retention.

Is Microsoft 365 the same as moving our servers?

No. Microsoft 365 is software as a service for mail, files, and apps. Moving a server is usually infrastructure as a service, and your team still patches and configures more of that stack.

Do we have to pick only one cloud model?

No. Most growing companies use a mix. The work is to name an owner, a restore target, and a monthly review for each layer you keep.

About the Author: Josh Holcombe is a forward-thinking IT leader and the driving force behind IGTech365, where he helps organizations modernize their technology, strengthen cybersecurity, and unlock operational efficiency. With a reputation for delivering innovative, business-focused IT solutions, Josh specializes in guiding companies through digital transformation in a way that is both practical and results-driven. Known for his ability to align technology with real-world business outcomes, Josh has worked with organizations across industries to streamline workflows, improve system reliability, and reduce risk.

To top