Shared responsibility

Three rows are always yours, three are always Microsoft's

On-premisesIaaSPaaSSaaSInformation and dataDevices (phones and PCs)Accounts and identitiesIdentity and directory infrastructureApplicationsNetwork controlsOperating systemPhysical hostsPhysical networkPhysical datacentreyouMicrosofthalf and half: sharedYour data, your devices and your accounts stay yours in every column. The buildings never are.Only the middle four rows move, and they move to Microsoft as you go from IaaS to PaaS to SaaS.

Your data, devices and accounts are always yours. The physical layers are always Microsoft's. The middle depends on the service type.

Shared responsibility · Going deeper

IaaS, PaaS and SaaS on Azure

You manageAzure examplesUse it whenIaaSthe operating system and all above itVirtual Machines, networks, disksmoving servers as they arePaaSyour code and your dataApp Service, Azure SQL Databasebuilding apps without serversSaaSyour data, users and settingsMicrosoft 365, Dynamics 365you want ready-made softwareMost control, most workLeast control, least workIaaS · PaaS · SaaSAsk what the customer still has to patch. The operating system: IaaS. Only the code: PaaS. Nothing: SaaS."Lift and shift" in a question means IaaS. "Focus on the application, not the infrastructure" means PaaS.Software that people simply sign in to and use is SaaS.

Ask what you still patch: the operating system for IaaS, only your code for PaaS, nothing for SaaS.

Cloud models

Public, private or hybrid is a question of whose hardware

Public cloudyouothersothersMicrosoft's hardware, shared bymany customers over the internetPrivate cloudyouHardware used by one organisation,in its own or a hosted datacentreHybrid cloudyouyouBoth, connected, with workloadsand data moving between themThe model is about who owns the hardware and who else is on it.Public: no hardware to buy, capacity on demand, least control. Private: full control and full cost, often chosen fora legal or legacy reason. Hybrid: keep some systems where they are and use the public cloud for the rest.Using two or more public clouds at once is called multi-cloud. Azure Arc manages servers in all of these from Azure.

Public: Microsoft's shared hardware. Private: one organisation's own. Hybrid: both, connected.

Cloud models · Going deeper

Which model does the statement describe?

StatementPublicPrivateHybridNo capital spending needed to scale up✓Applications can be set up and removed quickly✓You pay only for what you use✓You have complete control over resources and security✓Hardware must be bought to start up and maintained✓You choose where each application runs✓Meets a rule that some data must stay on your own site✓✓Most flexibility✓The exam gives a statement and asks which model it describes. These eight cover it.Statements follow the comparison in Microsoft Learn's cloud concepts module.

Eight statements the exam matches to public, private or hybrid.

Check yourself

SaaSAccounts and identities?Applications?Network controls?Operating system?Physical datacentre?

A company moves its email to Microsoft 365, a SaaS product. Which of these is still the company's job?

Paying

You stop buying for the peak and pay for the use

Demand over a yearbought up front, for the peakJanDecCapExBuy the grey box before you start.Over the year, 40% of it is used.OpExPay each month for the purple bar only.Nothing up front; stop and the bill stops.The cloud is a consumption-based model: you pay for what you use, when you use it.Capital expenditure (CapEx) is a one-off purchase of something you then own. Operational expenditure (OpEx) isspending on a service as you go. Moving to the cloud turns CapEx into OpEx, removes the guess about how muchcapacity to buy, and lets you stop paying for what you no longer need.

Bought up front, only 40% of this capacity is used. The consumption-based model bills the bars, not the box.

Paying · Going deeper

Datacentre against cloud, row by row

Traditional datacentreConsumption-based cloudPaidup front, for hardware you ownas you go, for what you usedKind of spendingcapital expenditure (CapEx)operational expenditure (OpEx)Capacityguessed in advance, for the peakadded and removed as demand changesIdle capacitypaid fornot paid forRunning the buildingsyour cost: power, cooling, staffincluded in the priceTo stopyou still own the hardwaredelete the resource and the cost endsServerless is the same idea at its limit: no server to size, and no charge while nothing runs.Cloud pricing also lets you commit ahead for a discount (reservations) or buy spare capacity cheaply; Lesson 8covers those as cost factors.

CapEx is bought up front and owned. OpEx is paid as you go.

Benefits

Scale up for a bigger machine, out for more machines

Scale up: a bigger machine2 CPU8 CPUvertical scaling: more CPU or memory in one machineScale out: more machines2 CPU2 CPU2 CPU2 CPU2 CPUhorizontal scaling: more copies of the same machineand backwhen demandfallsScalability is changing capacity to meet demand: up for a bigger machine, out for more of them.Both drawings end with 8 CPUs. Scaling out also survives the loss of one machine, which scaling up does not, soit helps availability as well. Scaling back in when demand falls is what keeps the bill following the demand;doing it automatically is called elasticity.

Scaling up is vertical: a bigger machine. Scaling out is horizontal: more machines.

Benefits · Going deeper · 1 of 2

The benefit words, one by one

BenefitIt meansAzure delivers it withHigh availabilitythe service is up when it is neededuptime promised in a service level agreementScalabilitycapacity changes to meet demandscaling up or out, by hand or automaticallyReliabilityit recovers from a failure and carries ona design spread over regions; automatic failoverPredictabilityyou can foresee performance and costautoscaling and load balancing; cost toolsSecurityyou choose the level of control you needfrom full control (IaaS) to managed patchingGovernanceresources meet your company's rulestemplates, policy and auditingManageability of the cloudhow resources are runautoscale, templates, alerts, self-healingManageability in the cloudhow you reach and control themportal, command line, APIs, PowerShellThe study guide names these benefits in pairs. Know each word well enough to tell it from its neighbour.Availability is being up. Reliability is getting back up. Predictability is knowing ahead what speed and cost to expect.

Availability is being up, reliability is getting back up, predictability is knowing ahead.

Benefits · Going deeper · 2 of 2

What an uptime promise allows

Downtime a promise allows in a 30 day month99%7.2 hours99.9%43.2 minutes99.95%21.6 minutes99.99%4.3 minutesA service level agreement (SLA) is a promised percentage of uptime. Each extra nine cuts the downtime tenfold.Bars are to scale: 30 days is 43,200 minutes, and the allowed downtime is that times the missing percentage.If Azure misses the promise you can claim a credit on the bill. The promise is higher when you spread a workloadover availability zones than when it sits on one machine.

A promise of 99.9% still allows 43.2 minutes of downtime a month.

Check yourself

Thursday2 CPUFriday evening2 CPU2 CPU2 CPUSaturday2 CPU

A shop's website slows down every Friday evening. The fix shown here is called what?

Serverless

Serverless bills the work, not the waiting

00:0006:0012:0018:0024:00A virtual machine, always onbilled for 24 hoursThe work that actually arrivesserverless: billed for the 4 hours it ranServerless means no server for you to size or patch, and no charge while nothing is running.There are still servers; Azure runs them and adds or removes them as work arrives. Azure Functions runs a pieceof code when an event happens and bills for the time it ran. It suits work that comes in bursts. A steady, busyworkload can cost less on a machine that is always on.

An always-on machine bills 24 hours. Serverless bills the 4 hours in which work arrived.

Serverless · Going deeper

Clue words for cloud concepts

When the question saysThe answer isBecause"no upfront cost", "pay only for what you use"Consumption-based model (OpEx)the bill follows the use"buy servers", "one-off purchase"Capital expenditure (CapEx)paid up front and owned"move existing servers unchanged"IaaSyou keep the operating system"develop without managing infrastructure"PaaSyou bring code and data"sign in and use", "email", "Microsoft 365"SaaSyou bring users and data"data must stay on our own site, the rest can move"Hybrid cloudboth, connected"add more instances when busy"Scaling out (horizontal)elastic when automatic"stays up", "99.9%"High availability, an SLApromised uptime"recover from failure"Reliabilityget back up"code runs only when triggered"Serverless (Azure Functions)no idle chargeArea 1 is vocabulary. Each of its questions turns on one of these words.

Area 1 is vocabulary. Ten phrases cover most of its questions.

Check yourself

Company ACompany B

Two companies pay for the same computing in different ways. Which one is using operational expenditure?

Lesson complete

You finished “Cloud concepts, as Azure asks them”

Answer the questions along the way to see your score here.

Open in a new tab