The stack

Every application sits on the same nine layers

ApplicationsDataRuntimeMiddlewareOperating systemVirtualisationServersStorageNetworkingthe program your users see and usewhat it stores: customers, orders, filesthe language engine: Node.js, Java, Python, .NETweb servers, queues and connectors between programsLinux or Windows, patched every monthslices one physical server into many virtual machinesprocessors and memory, bolted into a rackdisks, and replacing them when they failswitches, cables, firewalls and internet linkssoftware youwrite or installthe operating systemhardware andthe buildingEvery application sits on the same nine layers. In your own building, you run all nine.Each layer needs buying, installing, patching and someone on call. Most of that work is the same for every company,and none of it is the reason your application exists. The cloud’s question is: which layers will you hand over?

From the application down to the network cables: on your own premises you buy, run and patch all nine layers.

Service models

A service model is a line through the stack: above it is yours

On premisesyou own the buildingApplicationsDataRuntimeMiddlewareOperating systemVirtualisationServersStorageNetworkingIaaSinfrastructure as a serviceApplicationsDataRuntimeMiddlewareOperating systemVirtualisationServersStorageNetworkingPaaSplatform as a serviceApplicationsDataRuntimeMiddlewareOperating systemVirtualisationServersStorageNetworkingSaaSsoftware as a serviceApplicationsDataRuntimeMiddlewareOperating systemVirtualisationServersStorageNetworkingyou run itthe provider runs itthe line is where your work endsA service model is a line drawn through the stack: above it is yours, below it is the provider’s.IaaS hands over the hardware and virtualisation. PaaS also hands over the operating system, middleware and runtime.SaaS hands over everything, including the application. The three names come from the NIST definition.

IaaS, PaaS and SaaS differ only in how many layers the provider runs for you.

IaaS: infrastructure as a service

Rent the machine, run everything on it

ApplicationsDataRuntimeMiddlewareOperating systemVirtualisationServersStorageNetworkingWhat you getA virtual machine, a disk and a network address, in about a minute.Still your jobInstall, patch and secure the operating systemInstall the runtime, the middleware and your applicationScale it, back it up and watch itThe same thing on each providerAWSEC2AzureVirtual MachinesGoogle CloudCompute EngineIaaS rents you the machine. Everything you would do on your own server, you still do.Reach for it to move an existing system as it is (called lift and shift), or when software needs a particular operatingsystem setup. It gives the most control of the three models and leaves you the most work.

Infrastructure as a service gives you virtual machines, disks and networks. The operating system and everything above it are your work.

PaaS: platform as a service

Bring the code, the platform runs it

ApplicationsDataRuntimeMiddlewareOperating systemVirtualisationServersStorageNetworkingWhat you getA place that runs your code: you push it, the platform deploys, patches and scales it.Still your jobYour code, and its bugsYour data, and who can reach itSettings: instance size, scaling limits, secretsApplication platforms and managed databasesAWSElastic Beanstalk, RDSAzureApp Service, Azure SQLGoogle CloudApp Engine, Cloud SQLPaaS runs your code for you. You never log in to a server, and you never patch one.Reach for it for new web applications and APIs, when a team would rather ship features than maintain machines.Managed databases are PaaS too: you bring the tables and queries, the provider runs the database engine underneath.

Platform as a service deploys, patches and scales for you. You own the code and the data.

SaaS: software as a service

Finished software, you just use it

ApplicationsDataRuntimeMiddlewareOperating systemVirtualisationServersStorageNetworkingWhat you getFinished software in a browser, usually priced per user per month.Still your jobWhat data you put into itWho has an account, and what each person may doSettings: sharing rules, multi-factor sign-inExamplesOffice workMicrosoft 365, WorkspaceSalesSalesforceMessagingSlack, ZoomSaaS is someone else’s finished application. You configure it; you do not build or run it.Reach for it when the need is the same in every company: email, documents, customer records, payroll.Even here the top of the picture is not empty: your data and your users’ accounts are always yours to look after.

Software as a service is a complete application run by its vendor. Your data and your accounts are still yours.

SaaS: software as a service · Going deeper

Every “as a service” name puts the line somewhere

Short nameWhat is the serviceWhat the provider runs for youExamplesIaaSinfrastructuremachines, disks, networksEC2, Azure VMs, Compute EngineCaaScontainersa cluster that runs your imagesEKS, AKS, GKEPaaSplatforma runtime for your codeElastic Beanstalk, App Service, App EngineFaaSfunctionsone function, run per requestLambda, Azure Functions, Cloud Run functionsDBaaSdatabasea database engine, patched and backed upRDS, Azure SQL Database, Cloud SQLDaaSdesktopa Windows desktop streamed to any deviceWorkSpaces, Azure Virtual DesktopSaaSsoftwarethe finished applicationMicrosoft 365, Salesforce“As a service” means one thing every time: the provider runs it, and you rent the result.NIST defines only three models, and exams ask about those three. The other names mark where the line sits more exactly.Serverless is a property, not a layer: no servers for you to manage, automatic scaling, and a bill that is zerowhen nothing runs. Functions are the best-known serverless service; some databases and queues are serverless too.

CaaS, FaaS, DBaaS and DaaS are finer marks on the same scale. Serverless describes services with no servers to manage and no charge when idle.

Control and convenience

More done for you means less you can change

control: you choose every detailwork the provider does for youOn premisesyour own racksIaaSvirtual machinesEC2, Azure VMsContainersyou bring an imageEKS, AKS, GKEPaaSyou bring codeApp Service, App EngineFunctionsyou bring one functionLambda, Azure FunctionsSaaSyou bring usersMicrosoft 365Each step to the right hands the provider more work, and takes one more choice away from you.Containers and functions are newer steps on the same line: with containers you still choose the runtime, with functions yousupply only the code for one task. Take the step furthest right that still lets you do what you need.

From IaaS to SaaS you trade control for convenience. Containers and functions are steps on the same line.

Shared responsibility

The provider secures the cloud; you secure what you put in it

You: security IN the cloudYour dataApplications, identities and accessOperating system, network, firewallEncryption and traffic protectionThe provider: security OF the cloudSoftware that runs the services: compute, storage, database, networkingHardware and buildings: regions, zones, edge sites, power, guardsShared responsibility: the provider secures the cloud itself. You secure what you put in it.The words “of the cloud” and “in the cloud” are AWS’s, and its exam uses them exactly. Microsoft and Google publish the same split.The provider will not read your data to protect it, choose your passwords, or close a door you configured to be open.

The shared responsibility model: security of the cloud is the provider’s, security in the cloud is yours.

Shared responsibility by model

The line moves with the service model; three rows never do

On premisesIaaSPaaSSaaSInformation and datayouyouyouyouDevices: phones and PCsyouyouyouyouAccounts and identitiesyouyouyouyouIdentity and directory serviceyouyousharedsharedApplicationsyouyousharedproviderNetwork controlsyouyousharedproviderOperating systemyouyouproviderproviderPhysical hosts, network, buildingyouproviderproviderproviderThe top three rows never move: your data, your devices and your accounts are yours in every model.The bottom row is always the provider’s in any cloud. Only the middle moves, and “shared” means both sides have work:on PaaS the provider patches the platform, and you still write a safe application and set its network rules.

From IaaS to SaaS the provider takes on more, but data, devices and accounts stay with you in every model.

Shared responsibility by model · Going deeper · 1 of 2

Who patches what, service by service

ServiceThe provider patches and runsYou configure and maintainVirtual machine (EC2)hardware and the hypervisoroperating system patches, firewall rules,and all the software you installManaged database (RDS)hardware, operating system, database enginepatches, and the backups you switch ondatabase users, network access,and how long backups are keptObject storage (S3)hardware, software, and the durabilityof every stored objectwho can read each bucket, the encryptionchoice, and versioningFunctions (Lambda)hardware, operating system, languageruntime, and scalingyour code and its libraries, andthe permissions it runs withThe more managed the service, the shorter your list, and it never reaches zero.Exam questions name a service and a task and ask whose it is. Patching the operating system of a virtual machine: yours.Patching the operating system under a managed database or a function: the provider’s. Deciding who may access any of them: yours.AWS services are shown; the same split holds for the matching Azure and Google Cloud services.

For a virtual machine you patch the operating system. For a managed database, storage or functions the provider does, and access is still yours.

Shared responsibility by model · Going deeper · 2 of 2

Inherited, shared and customer-specific controls

InheritedThe provider does all of it.You inherit the result.physical security of buildingspower, cooling, fire suppressionSharedBoth do it, each fortheir own layer.patching: their hosts, your guestsconfiguration: their devices, your databasestraining: their staff, your staffCustomer-specificOnly you can do it.It depends on your application.which data goes in which regionhow your network is divided into zonesAWS sorts security controls into three kinds: inherited, shared and customer-specific.This wording is examined on the AWS Cloud Practitioner and architect exams. The examples are AWS’s own. “Shared” never meansthe provider does half of your job: it means the same kind of work exists on both sides of the line.

AWS’s three kinds of control: those you inherit, those both sides apply to their own layer, and those only you can decide.

What is always yours

Your data, your identities and your devices are always yours

Your datawhat you store, who may read it,whether it is encrypted and backed upYour identitiesaccounts, passwords, multi-factorsign-in, and each person’s permissionsYour devicesthe laptops and phones thatsign in to all of it99%of cloud security failuresthrough 2025 are thecustomer’s fault.Gartner’s predictionNo service model takes these three off your hands, and they are where breaches happen.Attackers rarely break the provider’s buildings or hypervisors. They find a storage bucket left public, a password with nosecond factor, or an access key pasted into public code. Every one of those is on the customer’s side of the line.So the safest reading of any exam question, and any real design: if it concerns data or access, it is yours.

Breaches in the cloud are almost always on the customer’s side: an open bucket, a weak sign-in, a leaked key.

What is always yours · Going deeper

The open bucket: nothing was hacked, the door was open

Providercustomer-recordsAccess: anyone on the internetAnyoneno password neededreads every fileThe provider’s part worked:✓disks encrypted and replicated✓building guarded✓service onlineThe setting was the customer’s:✗access left openA storage bucket set to public is the classic failure: nothing was hacked, the door was open.In 2019 a misconfigured firewall at the bank Capital One let an attacker reach storage holding records of about 100 millionpeople; the US regulator fined the bank 80 million dollars. Providers now block public access by default on new storage(AWS since April 2023), and still cannot stop a customer who deliberately switches that off.

A misconfiguration is a setting left unsafe. It is the most common cause of data exposure in the cloud, and it is the customer’s side of the model.

Compliance

A certified cloud certifies the floor, not what you build on it

The provider’s half: audited for youISO 27001SOC 2PCI DSSand dozens moreYour half: your audityour application and how it handles datayour settings: access, encryption, loggingyour people and your processesWhere the provider’s audit reports areAWSAWS ArtifactAzureService Trust PortalGoogle CloudCompliance Reports ManagerRunning on a certified cloud does not make your system certified. It certifies the floor you built on.You inherit the provider’s controls for the layers they run, download their reports as evidence, and are audited yourselffor everything above the line. A shop taking card payments on a PCI-certified cloud must still pass PCI for its own code.

Compliance is shared too: you inherit the provider’s audited controls and are audited yourself for your half.

Choosing a service model

Take the most managed option that fits

Does every company need this?email, documents, payroll, CRMyesSaaSbuy it, configure itnoIs it your own code?and standard runtimes will doyesPaaS or functionspush code, skip the serversnoNeed your own OS setup?old software, special drivers, licencesyesIaaSrent machines, run it yourselfStart from the most managed option that fits, and step down only for a reason.Every layer you keep is a layer you patch, scale and answer for at night. Real systems mix all three: email on SaaS,the web shop on PaaS with a managed database, and one old billing system on virtual machines.

SaaS for common needs, PaaS for your own code, IaaS when you must control the operating system. Most companies use all three.

Lesson complete

You finished “Service models and shared responsibility”

Nicely done.

Open in a new tab