Google Cloud Professional Cloud Architect Managing and Provisioning a Solution Infrastructure Syllabus

What are the two main Cloud Interconnect options and their bandwidth granularity?

Dedicated Interconnect (direct physical connection, 10 Gbps or 100 Gbps circuits) and Partner Interconnect (through a service provider, from 50 Mbps up to 50 Gbps).

When would you choose Partner Interconnect over Dedicated Interconnect?

Choose Partner Interconnect when you can't meet Google in a colocation facility, need less than 10 Gbps, or want a provider to manage the physical connection.

What SLA can Dedicated Interconnect achieve, and what topology is required for 99.99%?

Dedicated Interconnect offers a 99.9% SLA by default and a 99.99% SLA when deployed with redundant connections across two metro availability zones (four connections).

What is the difference between Cloud Interconnect and Direct/Carrier Peering?

Interconnect provides access to internal RFC 1918 VPC IPs (private connectivity); Direct/Carrier Peering only provides access to Google public services (Google Workspace, APIs) and does not reach VPC internal IPs.

What is Cloud Router and what protocol does it use?

Cloud Router is a fully managed service that dynamically exchanges routes between your VPC and on-prem/peer networks using BGP (Border Gateway Protocol), enabling dynamic routing for VPN and Interconnect.

In Cloud Router BGP, what is a custom route advertisement (advertised routes) used for?

It lets you control which subnets/prefixes Cloud Router advertises to the on-prem peer — either default (advertise all subnets) or custom (advertise specific IP ranges).

What is a BGP ASN, and what ranges does Cloud Router support?

An Autonomous System Number identifies each BGP peer. Cloud Router supports private ASNs (64512-65534, 4200000000-4294967294) for Google's side and you assign a peer ASN to the on-prem router.

What is Network Connectivity Center (NCC) in Google Cloud?

NCC is a hub-and-spoke connectivity management model where a central hub connects multiple spokes (VPNs, Interconnects, router appliances, VPCs) to enable site-to-site and site-to-cloud data transfer.

In Network Connectivity Center, what is a 'spoke' versus a 'hub'?

A hub is the central management resource that provides connectivity between spokes; a spoke is a resource attached to the hub (HA VPN tunnel, VLAN attachment, router appliance, or VPC) representing a network endpoint.

What are the four standard Cloud Storage classes from hottest to coldest?

Standard, Nearline, Coldline, and Archive.

What are the minimum storage durations for Nearline, Coldline, and Archive Cloud Storage?

Nearline = 30 days, Coldline = 90 days, Archive = 365 days minimum storage duration (early deletion incurs a charge).

For Cloud Storage, what access pattern fits Nearline vs Coldline vs Archive?

Nearline: accessed less than once a month; Coldline: less than once a quarter (~90 days); Archive: less than once a year (long-term backup/DR).

How does Cloud Storage pricing trade off across classes as data gets colder?

Colder classes have lower at-rest storage cost but higher data retrieval/access costs and longer minimum-storage commitments; Standard has highest storage cost but no retrieval fee.

What is an Object Lifecycle Management rule in Cloud Storage and give two example actions?

It's a per-bucket policy that automatically takes action on objects meeting conditions (e.g., age, createdBefore). Example actions: SetStorageClass (transition to Coldline) and Delete.

What is the difference between Cloud Storage 'Object Lifecycle Management' and 'Object Versioning'?

Versioning keeps noncurrent (older) copies of objects when overwritten/deleted; Lifecycle Management automates transitions/deletions, including cleaning up noncurrent versions based on conditions.

What are the Persistent Disk types, and which offers the highest IOPS?

pd-standard (HDD), pd-balanced (general SSD), pd-ssd (performance SSD), and pd-extreme (highest IOPS with provisioned IOPS). pd-extreme/Hyperdisk offer the highest performance.

What is Hyperdisk and how does it differ from traditional Persistent Disk?

Hyperdisk is Google's next-gen block storage that decouples capacity, throughput, and IOPS so each can be provisioned and tuned independently, offering higher and more flexible performance than standard PD.

Name three Hyperdisk variants and their primary optimization.

Hyperdisk Balanced (general purpose), Hyperdisk Extreme (highest IOPS for demanding databases), and Hyperdisk Throughput (cost-effective high throughput for scale-out/analytics workloads).

How does Persistent Disk performance generally scale, and what is a key implication for sizing?

PD IOPS and throughput scale with disk size and (for some types) the VM's vCPU count, so you may need to increase disk size or VM size to reach a target performance level.

What is the difference between a zonal and a regional Persistent Disk?

A zonal PD exists in one zone; a regional PD synchronously replicates data across two zones in a region for higher availability and DR failover.

What is Filestore, and what protocol does it use?

Filestore is Google's fully managed NFS (network file system) service providing shared file storage for Compute Engine and GKE workloads via the NFSv3 protocol.

Name the main Filestore service tiers and a distinguishing feature of each.

Basic HDD/SSD (general file sharing), Zonal (high performance, scalable capacity), Enterprise (regional, high availability across zones, GKE-friendly), and Regional. Enterprise/Regional add multi-zone redundancy.

What is a retention policy (bucket lock) in Cloud Storage used for?

A retention policy enforces a minimum retention period during which objects cannot be deleted or overwritten; with a bucket lock it becomes immutable, supporting compliance (WORM).

What is an instance template in Compute Engine and a key constraint about modifying it?

An instance template is an immutable resource defining VM properties (machine type, image, disks, network) used to create instances/MIGs. It cannot be edited; you must create a new template to change configuration.

What is a Managed Instance Group (MIG) and what three capabilities does it provide?

A MIG creates identical VMs from an instance template and provides autoscaling, autohealing (health-check-based recreation), and automated/rolling updates plus load balancing integration.

What is the difference between a zonal and a regional MIG?

A zonal MIG places all instances in one zone; a regional MIG spreads instances across multiple zones in a region for higher availability.

What signals can Compute Engine autoscaling for a MIG be based on?

CPU utilization, HTTP(S) load balancing serving capacity, Cloud Monitoring custom metrics, and schedule-based scaling. A target value is set and the autoscaler adds/removes instances to meet it.

In MIG updates, what is the difference between a rolling update and a canary update?

A rolling update gradually replaces all instances with the new template (controlled by maxSurge/maxUnavailable); a canary update applies the new template to only a subset (e.g., target size) to test before full rollout.

What are the high-availability options for Cloud SQL?

Cloud SQL HA uses a regional configuration with a primary and a standby instance in a different zone, with automatic failover; read replicas can be added for read scaling (and cross-region DR).

What is AlloyDB and how does it relate to Cloud SQL/PostgreSQL?

AlloyDB is a fully managed, PostgreSQL-compatible database optimized for high performance and analytical (HTAP) workloads, offering better performance and scalability than standard Cloud SQL for PostgreSQL via a disaggregated storage architecture and columnar engine.

What scaling/consistency properties make Cloud Spanner unique?

Spanner is a globally distributed, horizontally scalable relational database providing strong consistency and high availability with up to 99.999% SLA (multi-region), using TrueTime for external consistency.

What is Bigtable best suited for, and what data model does it use?

Bigtable is a NoSQL wide-column store for large-scale, low-latency, high-throughput analytical and operational workloads (time-series, IoT, AdTech). It uses a single-key (row key) sorted, sparse table model with no SQL/joins.

How does Bigtable performance scale, and what is the rough throughput per node?

Bigtable scales linearly by adding nodes; each SSD node handles roughly 10,000 QPS (reads or writes) at ~6 ms latency. Capacity/performance grows with node count.

What are the two modes of Firestore and how do they differ?

Native mode (document database with real-time listeners, offline sync, strong consistency, for mobile/web apps) and Datastore mode (backward-compatible with Datastore for server workloads, no real-time/offline features).

In Bigtable, why is row key design critical?

Because data is sorted by row key and stored contiguously, poorly designed keys (e.g., sequential timestamps) cause hotspotting; well-distributed keys spread reads/writes evenly across nodes to avoid bottlenecks.

What is Terraform and what are its core files/commands for Google Cloud?

Terraform is HashiCorp's declarative infrastructure-as-code tool. It uses .tf config files and the google provider; core workflow is terraform init, plan, and apply, with state stored (ideally in a GCS backend).

Why store Terraform state in a Cloud Storage bucket, and what feature prevents concurrent corruption?

A GCS remote backend centralizes/shares state and enables state locking (via object generation/Cloud Storage) so concurrent applies don't corrupt state; it also keeps sensitive state off local machines.

What is Infrastructure Manager in Google Cloud?

Infrastructure Manager is Google's managed service that provisions infrastructure using Terraform configurations, automating Terraform execution with state management, IAM integration, and versioned deployments.

What is Config Connector and what does it enable?

Config Connector is a GKE add-on (Kubernetes addon) that lets you manage Google Cloud resources as Kubernetes objects (via CRDs/YAML), so you provision GCP infrastructure using kubectl and Kubernetes-native, declarative reconciliation.

What is Cloud Marketplace and what deployment benefit does it provide?

Cloud Marketplace offers pre-configured, vendor-provided solutions (VM images, Kubernetes apps, SaaS) that can be deployed into your project with one click, often using Deployment Manager/Terraform under the hood and integrated/unified billing.

What is the Google Cloud resource hierarchy from top to bottom?

Organization → Folders → Projects → Resources. IAM policies and organization policies set at a higher level are inherited by everything below.

What is the difference between IAM policies and Organization Policies?

IAM policies define who (identity) can do what (role) on a resource; Organization Policies (constraints) define what configurations are allowed on resources (e.g., restrict allowed regions, disable external IPs), regardless of identity.

What is a project's role in the GCP resource hierarchy and billing?

A project is the base-level organizing entity that contains resources, is the unit for enabling APIs, IAM, quotas, and billing; every resource belongs to exactly one project, and projects link to a billing account.

What is an organization policy constraint, and name two common examples?

A constraint is a rule applied via Organization Policy Service to restrict resource configurations. Examples: constraints/compute.vmExternalIpAccess (disable external IPs) and constraints/gcp.resourceLocations (restrict resource regions).

What is Shared VPC and what problem does it solve in the resource hierarchy?

Shared VPC lets a host project share its VPC network with multiple service projects in the same organization, so resources in different projects use a centrally managed network — separating network admin from project admins.

What is VPC Network Peering and one key limitation versus a single VPC?

VPC Peering privately connects two VPCs (even across projects/orgs) so they exchange traffic using internal IPs. Limitation: peering is non-transitive (peered VPCs don't share each other's other peers) and subnet IP ranges must not overlap.

For data retention compliance, how do Cloud Storage retention policies and lifecycle rules work together?

Retention policies enforce a minimum time objects must be kept (preventing deletion); lifecycle rules then automate transition to colder classes and eventual deletion once the retention period and age conditions are met.

What is the difference between Compute Engine sole-tenant nodes and standard shared-core provisioning?

Sole-tenant nodes dedicate an entire physical server to a single customer's VMs (for compliance/licensing isolation), whereas standard VMs share host hardware with other tenants.