Service models, shared responsibility, availability
14 min
Model
Customer manages
Provider manages
Example
IaaS
OS, runtime, apps, data
Facilities, hardware, hypervisor
Amazon EC2, Azure VMs
PaaS
Code, data, app settings
Above plus OS, runtime, patching
Azure App Service, Google App Engine
SaaS
Users, data, settings
The whole application
Microsoft 365, Salesforce
FaaS (serverless)
Functions and their triggers
Servers, runtime, scaling; billed per execution
AWS Lambda, Azure Functions
Shared responsibility: the provider secures the cloud (facilities, hardware, hypervisor); the customer secures what it puts in the cloud. Data, identities and access are always the customer's.
A public storage bucket or a weak IAM policy is the customer's fault in every model, including SaaS.
Term
Meaning
Region
A geographic area with its own set of data centers
Availability zone (AZ)
One or more isolated data centers in a region; spread instances across AZs to survive a data-center failure
Cloud bursting
A private or on-premises workload overflows into public cloud at peak
Edge computing
Processing near the data source to cut latency
Multicloud tenancy
Workloads spread across two or more providers
DR term
Meaning
RPO
Maximum acceptable data loss, measured in time (how old the last good copy may be)
RTO
Maximum acceptable time to restore service
Hot site
Running and current; failover in minutes; highest cost
Warm site
Hardware ready, data somewhat behind; hours
Cold site
Space and power only; days; lowest cost
RPO looks back (data). RTO looks forward (downtime).