Regions and Availability Zones
39 Regions, each with at least three isolated zones
A Region is a geographic area. An Availability Zone is one or more isolated data centres inside it.
AWS is built in layers of place. A Region is a geographic area, such as Ireland or Northern Virginia. Inside each Region are Availability Zones. Each zone is one or more data centres with its own power, cooling and networking. The zones are kilometres apart, joined by private fibre: far enough that one disaster does not reach two of them, close enough to copy data quickly. Today there are 39 Regions and 124 zones, and every Region has at least three.
Regions and Availability Zones · Going deeper
Names, distances and global services
Most services are regional. IAM, Route 53, CloudFront and Organizations are global.
High availability
Two zones and a load balancer survive a data centre failure
High availability on AWS means at least two Availability Zones.
Why do zones matter? Because of this. Users reach a load balancer, which spans two Availability Zones. Zone a loses power. Its servers and its database are gone. The load balancer notices, and sends every request to zone b, where identical servers and a standby database are running. Users see nothing. That is high availability, and on AWS it always means more than one Availability Zone.
Multiple Regions
A second Region needs one of four reasons
Use several Regions for disaster recovery, business continuity, low latency or data sovereignty.
Several zones give availability. Several Regions are a bigger step, and the exam guide names four reasons to take it. Disaster recovery: a copy elsewhere if a whole Region is disrupted. Business continuity: keep trading through anything. Low latency for end users: serve people from a Region near them. Data sovereignty: keep data inside a country because the law says so.
Multiple Regions · Going deeper
How to choose a Region
Choose by compliance, latency, service availability, then price.
Check yourself
A company must keep serving customers even if an entire data centre fails. What is the minimum design?
Edge locations
750+ small sites exist only to be near users
Edge locations cache content and answer DNS close to users. They are not Regions.
There is one more layer, and it is the most numerous. AWS has 39 Regions but more than 750 edge locations, drawn here to the same scale. A viewer far from your Region asks for a video. The request goes to the nearest edge location, a few milliseconds away, which already holds a copy. Only if it does not, it fetches from the origin. CloudFront caches at the edge, Route 53 answers DNS there, Global Accelerator takes traffic in there, and WAF and Shield filter there.
Edge locations · Going deeper
Local Zones, Wavelength, Outposts and clue words
Outposts is AWS in your building; Local Zones are AWS near a city.
Ways to access AWS
Console, CLI and SDK are three doors to the same API
The console for people, the CLI for scripts, SDKs for code: all call the same APIs.
How do you tell AWS what to do? The Management Console is a website: you click, and you see. The command line interface runs commands, which can be saved as scripts. The software development kits let your own programs call AWS in Python, Java, JavaScript and other languages. All three do the same thing underneath: they call the AWS APIs, which operate the services. Infrastructure as code is a fourth way, where a template makes the calls.
Infrastructure as code
A template builds the same environment every time
AWS CloudFormation builds a stack from a template: repeatable, consistent and reviewable.
Clicking through the console is fine once. It is a poor way to build the same thing three times. Infrastructure as code writes the environment down as a template: a server, a database, a bucket. The template lives in version control like any other code. CloudFormation reads it and builds everything in it as one unit, called a stack. Run it again for test, and again for production, and the three environments are identical.
Infrastructure as code · Going deeper
Tools for provisioning and operating
One-time: the console. Repeatable: CloudFormation, scripts and pipelines.
Check yourself
A team must create the same environment in five accounts, and prove each one matches. Which approach?
Deployment models
Cloud, on-premises, or hybrid with the two connected
The three deployment models are cloud, hybrid and on-premises.
Finally, where does an application run? There are three deployment models. Cloud: everything is on AWS. Hybrid: some systems stay in your own data centre and some are on AWS, joined by a VPN or by Direct Connect. On-premises, sometimes called private cloud: everything is in your own data centre. Hybrid is where most companies are during a migration.
Check yourself
Which statement about edge locations is correct?
Lesson complete
You finished “AWS global infrastructure and ways to deploy”
Answer the questions along the way to see your score here.