CDL
Deep Dive
Regions, zones, global network, resource hierarchy, and how Google Cloud organizes its physical and logical infrastructure.
Google Cloud organizes its infrastructure into regions (geographic locations) that contain multiple zones (physically separated datacenters). Physical separation between zones protects against localized failures.
Independent geographic areas such as us-central1 (Iowa), europe-west1 (Belgium), or asia-east1 (Taiwan). Each region is in a distinct geographic area.
Key fact for the exam
Data does NOT move automatically between regions. Cross-region replication must be explicitly configured.
Examples
Physically separated datacenters WITHIN a region. Connected by a low-latency network (<5ms) but with independent infrastructure (power, cooling, network).
Key fact for the exam
For high availability, deploy resources across MULTIPLE zones in the same region. If one zone fails, the others keep working.
Examples
Points of Presence where Google's network connects with the internet. Used by Cloud CDN and Cloud Interconnect to reduce latency to the end user.
Key fact for the exam
Edge Locations are NOT for deploying VMs. They are only for CDN, network routing, and connecting with interconnect partners.
Examples
GCP vs AWS network difference
AWS VPC — Per region
A VPC in AWS exists in ONE single region. To connect VPCs from different regions you need VPC Peering or Transit Gateway.
GCP VPC — GLOBAL
A VPC in GCP is global. It can have subnets in ANY region. Resources in different regions communicate automatically over Google's private network.
Understanding the hierarchy is fundamental for IAM and governance. IAM policies are inherited down the hierarchy.
Organization
Represents the entire company. Root of the whole hierarchy. Linked to a Google Workspace or Cloud Identity domain. Policies applied here are inherited by everything.
Folders
Group projects. Represent departments, teams, or environments (prod/dev). Can be nested. IAM policies applied to a folder are inherited by all projects within it.
Projects
Base unit of isolation, billing, and management in GCP. Every resource belongs to a project. They have a unique Project ID, a name, and a number. Quotas, enabled APIs, and billing are configured per project.
Resources
The concrete services: VMs, buckets, databases, etc. They belong to a project. They can have their own IAM policies (more granular than at the project level).
IAM policy inheritance
IAM policies are inherited from parent to child: Organization → Folder → Project → Resource. If you grant the "Editor" role to a user at the Organization level, that user will have Editor permissions in ALL projects. Policies are additive — you cannot remove inherited permissions at a child level (for that, use IAM Deny policies).
| Criterion | Description | Example |
|---|---|---|
| Latency | Choose the region closest to your users for lower latency. Use Cloud CDN for static content. | Customers in Mexico → southamerica-east1 (São Paulo) or us-central1 (Iowa) |
| Availability | Deploy across multiple zones or regions for higher availability. Check the uptime history of each region. | High availability: us-central1 (3+ zones) or multi-region US |
| Compliance | Some data must remain in certain countries due to regulation (GDPR in EU, LGPD in Brazil, etc.). | Data of European citizens → europe-west1 (Belgium) or europe-west4 (Netherlands) |
| Pricing | Prices vary by region. us-central1 tends to be the most economical in the USA. | Dev/test workloads → us-central1 for lower cost |
| Service availability | Not all GCP services are available in all regions. New features launch first in us-central1. | New AI services → us-central1 gets availability first |
The CDL exam checks that you know these fundamental network terms. You don't need to be a networking expert — just understand what each concept does.
| Term | What it is | Relevance in GCP |
|---|---|---|
| IP Address | Unique numeric identifier of each device on the internet. IPv4 (e.g.: 192.168.1.1) or IPv6. | Every VM, Load Balancer, and GCP service has an IP. External IPs cost money if reserved but unused. |
| ISP (Internet Service Provider) | Company that provides internet access (Telmex, Claro, Movistar). Internet traffic passes through ISPs until it reaches Google. | Cloud Interconnect lets you bypass the public ISP with a direct connection to Google's network. |
| DNS (Domain Name Server) | Translates domain names (google.com) to IPs (142.250.69.46). "The phone book of the internet". | Cloud DNS manages DNS records with 100% SLA. Private zones for internal resolution within VPCs. |
| Latency | Time it takes for a data packet to travel from origin to destination. Measured in milliseconds (ms). | Google has global PoPs to reduce latency. Regions closer to the user = less latency. |
| Bandwidth | Maximum data transfer capacity per unit of time (Gbps, Mbps). "The width of the pipe". | Cloud Interconnect offers up to 200 Gbps. Google's network has more than 1 million km of fiber. |
| Fiber optics | Cable that transmits data as pulses of light. Much faster and lower latency than copper cables. | Google's backbone network is built on fiber optics and proprietary submarine cables. |
| Submarine cables | Fiber optic cables on the ocean floor connecting continents. Carry more than 95% of international traffic. | Google has 33+ proprietary submarine cables (Grace Hopper, Dunant, Equiano) to connect its datacenters globally. |
| Network edge / PoP | Points where a provider's network connects with the internet. Closer to the end user than the main datacenters. | Google has 100+ PoPs where Cloud CDN caches content and Cloud Interconnect connects with partner networks. |
Did you get this topic?
Test what you just learned
A company headquartered in Europe needs to guarantee that its European customers' data never leaves the European Union, complying with GDPR. How should it configure its infrastructure on Google Cloud?
Put what you learned to the test with original CDL exam questions.
Start CDL mock examUnlimited mock exams with the Pro plan
Log in to track your progress.