Prepare and test your skills
Prepare and test your skills
Hybrid and multicloud connectivity is the set of links that connect an on-premises data center to Google Cloud and to other clouds. These links carry traffic between environments, so bandwidth, latency, packet loss, and jitter directly affect application performance. Google Cloud provides Cloud VPN (opens in a new tab), Cloud Interconnect (opens in a new tab), and Cross-Cloud Interconnect (opens in a new tab) for these connections, with Cloud Router (opens in a new tab) providing dynamic routing across them. The choice depends on transfer volume, need for private transport, encryption needs, and reliability targets.
HA VPN is a high-availability VPN solution that connects an on-premises network to a Virtual Private Cloud (VPC) network through an IPsec VPN connection over the public internet. HA VPN can provide an availability service-level agreement (SLA) of 99.99% or 99.9% depending on topology and configuration. Traffic is encrypted with IPsec, which makes HA VPN suitable when encryption or private IP addresses are required for smaller or occasional transfers.
When you create an HA VPN gateway, Google Cloud automatically chooses two external IP addresses, one for each gateway interface, with each address from a unique pool to support high availability. Each gateway interface supports multiple tunnels, and you can create multiple gateways for higher throughput or multiple subnets. When you delete the gateway, Google Cloud releases the addresses for reuse. In API and gcloud commands HA VPN gateways are called VPN gateways, and you do not create forwarding rules for them.
HA VPN uses an external VPN gateway resource to store information about the peer VPN gateway on the on-premises or other-cloud side. HA VPN over Cloud Interconnect combines IPsec encryption with Interconnect capacity so traffic never traverses the public internet. When deployed over Cloud Interconnect, HA VPN interfaces can use regional internal IP addresses, but only when associated with VLAN attachments.
Exam tip: An HA VPN gateway configured with only one active interface and one external IP address does not provide an availability SLA.
Dedicated Interconnect (opens in a new tab) provides direct physical connections between an on-premises network and the Google network without traversing the public internet. The connection is provisioned between Google equipment and customer routing equipment in a common colocation facility, which is the Google point of presence (PoP) where the two networks meet. Because it offers deterministic performance with an SLA and high capacity, Dedicated Interconnect is more cost-effective than internet bandwidth for large, sustained transfers and long-term setups with high outbound volume.
Using Dedicated Interconnect requires meeting Google in a supported colocation facility and providing your own routing equipment, typically located in that facility. Network devices must support IPv4 link-local addressing, Link Aggregation Control Protocol (LACP) even for a single circuit, External Border Gateway Protocol version 4 (EBGP-4) with multi-hop, and 802.1Q VLANs. The physical circuits must use one of the following options:
The maximum supported fiber length for these optics is 10 km. If the on-premises router is more than 10 km from the Google demarcation, you must complete the connection with a lit wavelength, dense wavelength-division multiplexing (DWDM), or another managed circuit. When you create a VLAN attachment, you associate it with a Cloud Router that creates a Border Gateway Protocol (BGP) session to the on-premises peer router and receives the routes it advertises. Each colocation facility supports VLAN attachments in only a subset of Google Cloud regions, and 100-Gbps circuits are not available in every location.
Partner Interconnect extends an on-premises network to VPC networks through a supported service provider rather than a direct customer-owned cross-connect. The partner has already established connections to the Google network, so you select from the partner's available locations when you set up the VLAN attachment. Partner Interconnect capacity is normally less than 10 Gbps per connection, so consider location and required capacity when choosing between Dedicated and Partner Interconnect.
Routing responsibility for Partner Interconnect depends on connectivity type. With Layer 2 connections, you configure the BGP session between your on-premises router and Cloud Router. With Layer 3 connections, the service provider manages the BGP session between their router and Cloud Router. If you use Partner Interconnect to reach third-party providers and need network-layer encryption for security or compliance, you must add IPsec encryption, typically with HA VPN over Cloud Interconnect.
Cross-Cloud Interconnect provides private and secure connectivity between Google Cloud and supported cloud providers in specific locations. The service offers line-rate performance with high-availability options of 99.9% and 99.99%, which helps lower total cost of ownership without managing cross-connects and routing equipment in remote PoPs. If network-layer encryption in transit is required, Cross-Cloud Interconnect supports Media Access Control security (MACsec) for encryption.
Choose Cross-Cloud Interconnect when you need dedicated multicloud capacity without managing hardware and colocation relationships. For long-term multicloud setups with high outbound volumes, it can optimize performance and may reduce outbound data transfer charges for qualifying traffic. You must choose a location supported by the remote cloud provider, and each location supports VLAN attachments in only a subset of Google Cloud regions.
Cloud Router is the Google Cloud service that uses Border Gateway Protocol (BGP) to provide dynamic routing for hybrid connectivity. Cloud Router is used together with Cloud Interconnect and Cloud VPN gateways, including HA VPN. Instead of maintaining static routes on each side, Cloud Router exchanges routes with the external router over BGP sessions so both the VPC network and the on-premises or other-cloud network learn reachable prefixes automatically.
The external router is the device that provides access to the peered network, either an on-premises router or a virtual router hosted by a remote cloud provider. For Dedicated Interconnect, you manage the on-premises router and establish BGP sessions between that router and Cloud Router. For Partner Interconnect, responsibility follows the Layer 2 versus Layer 3 model. When a link, tunnel, or path fails, BGP withdraws unreachable routes and Cloud Router directs traffic to remaining healthy paths, which enables failover and supports active/active HA VPN designs.
The optimal hybrid connection depends on transfer volume, need for private transport, encryption requirements, and reliability targets. For occasional or low-volume exchanges that are not critical, transfer over the internet with public IP addresses is simplest and still provides relatively low latency and high bandwidth. For smaller transfers that need encryption or private IP addresses, use Cloud VPN with a managed VPN service on the other side.
A decision tree branching on multicloud need, volume, encryption, and capacity to choose Internet, HA VPN, Dedicated Interconnect, Partner Interconnect, or Cross-Cloud Interconnect.
| Option | Transport | Choose when |
|---|---|---|
| HA VPN | Encrypted IPsec over public internet, or over Interconnect | Transfers need encryption and high availability with 99.9% or 99.99% SLA, or IPsec is required over Partner Interconnect |
| Dedicated Interconnect | Direct physical connection in a colocation facility | Sustained transfers are typically 10 Gbps or more, deterministic performance and SLA are required, and the team can manage hardware and colocation |
| Partner Interconnect | Private connection through a service provider | Larger-than-VPN transfers need dedicated capacity below about 10 Gbps per connection without building direct cross-connects |
| Cross-Cloud Interconnect | Managed private connection between clouds | Multicloud traffic needs line-rate private connectivity with 99.9% or 99.99% availability without managing remote routing equipment |
For long-term hybrid or multicloud use with high outbound volumes, prefer Cloud Interconnect or Cross-Cloud Interconnect to optimize performance and potentially lower outbound charges. When connecting to Google Cloud resources, use Cloud Interconnect rather than Direct Peering or Carrier Peering, unless access to Google Workspace applications requires peering.
Workload portability is the ability to move an application between computing environments without significantly modifying the application or its operational model. Unified management is the practice of operating distributed clusters with consistent deployment, configuration, security, and observability from a central control point. In Google Cloud, containers and Kubernetes provide the common runtime that makes portability possible, while Google Kubernetes Engine (GKE) (opens in a new tab) provides managed Kubernetes and unified control.
Containers and Kubernetes create a common runtime layer that abstracts differences between on-premises, edge, Google Cloud, and other clouds. Containers decouple applications from host infrastructure, so software runs reliably when moved from one environment to another. Kubernetes handles orchestration, deployment, scaling, and management of containerized applications and provides the same services and APIs in cloud and private environments. Because this abstraction is higher than working directly with virtual machines (VMs), teams need less environment-specific work and can develop and operate workloads consistently.
Kubernetes is open source governed by the Cloud Native Computing Foundation and supports container runtimes that support the Container Runtime Interface (CRI), are industry-adopted, and are not tied to a specific vendor. When workloads run on Google Cloud, GKE avoids the effort of installing and operating Kubernetes by providing it as a managed platform. GKE Autopilot is a mode that manages cluster configuration, including nodes, scaling, security, and other preconfigured settings. Using a managed platform lets operations staff shift focus from maintaining infrastructure to maintaining applications.
GKE deployment options are the supported ways to run containerized applications at scale across on-premises, edge, Google Cloud, and other clouds. GKE includes managed open source components to secure workloads, enforce compliance policies, and provide network observability and troubleshooting. On-premises or edge clusters connect to Google Cloud through defined connectivity options, and Google Distributed Cloud connected (opens in a new tab) can simplify hybrid setup and operation for container workloads. Although some components can sustain temporary interruption, a software-only Google Distributed Cloud deployment should not run disconnected as a normal mode.
| Option | Where clusters run | What it provides |
|---|---|---|
| GKE hybrid environment | On-premises with Google Cloud | Cloud-like experience on-premises and unified path for transitioning applications |
| GKE Multi-Cloud | Other public clouds | Consistent governance, operations, and security posture across clouds |
| Google Distributed Cloud connected | On-premises or edge locations | Simplified hybrid operation for container workloads connected to Google Cloud |
Centralized operation ties these options together. A centralized management plane and monitoring solution in the cloud manages multiple locations efficiently, while consistent continuous integration and continuous deployment (CI/CD) pipelines deploy the same binaries, packages, or containers across environments. Common identity between environments lets systems authenticate across boundaries, and traffic between environments is encrypted in transit with VPN tunnels, Transport Layer Security (TLS), or both.
Fleet management (opens in a new tab) in GKE is the method for operating multiple clusters as logical groupings of similar environments. A fleet groups clusters so security, configuration, and service management stay consistent even when clusters run on-premises, at the edge, or in another cloud. This grouping lets administrators apply the same policies and observe multi-cluster applications from one management model instead of configuring each cluster separately. The result is unified provisioning, deployment, management, and monitoring that hides differences between environments.
Building multi-cluster architecture without managed tooling creates operational complexity. That complexity grows further when container-level security control is required. GKE reduces that burden by providing fleet-level controls for compliance and configuration alongside observability for troubleshooting across distributed resources.
A service mesh provides container-level security control, policy enforcement, and observability for traffic between services. GKE and related products use this capability to create a zero trust distributed architecture across hybrid and multicloud environments. In that architecture, services on-premises or in another cloud communicate across environments through end-to-end mutual TLS (mTLS) for secure service-to-service communication. Traffic therefore flows from a service in one environment directly to a service in another environment with identity-based authentication and encryption, rather than relying on network location alone.
Exam tip: A portable workload can move with minimal changes but still perform differently, because compute, networking, security capabilities, and proximity to dependent services differ by environment.
Limits to workload portability are the practical differences that remain even when Kubernetes hides many intricacies between environments. True portability requires consistent deployment and management plus consistent visibility, configuration, and security, without conflicting with being cloud-first. Even when those conditions are met, behavior is not identical across environments and data movement must be planned.
Data movement planning therefore shapes portability decisions, including storage type, capacity, accessibility, and restrictions for regulated data classes. Teams must identify business use cases, the data that powers them, and how storage, sharing, and access across environments affect performance and consistency before moving a workload.
Cross-cloud data and identity integration connects storage, analytics, and identities across Google Cloud, AWS, Azure, and on-premises environments into one architecture. Data is copied with Storage Transfer Service (STS) (opens in a new tab) when it must move, is analyzed in place with BigQuery Omni (opens in a new tab) when it should stay where it is, and is protected by consistent networking and identity federation (opens in a new tab). The goal is robust connectivity, portable data, and the same security controls on every side of the boundary.
BigQuery Omni is a multicloud analytics capability that runs BigQuery compute power locally to storage on Amazon Web Services (AWS) or Microsoft Azure. A query issued in BigQuery is executed by compute in the AWS or Azure region near the data, and the query can read data stored in Amazon Simple Storage Service (Amazon S3) or Azure Blob Storage as external sources. Data flows from the S3 bucket or Blob Storage container to the local Omni compute and then query results return to BigQuery, so teams can break down data silos without first copying entire datasets. Choose BigQuery Omni when the requirement is cross-cloud analysis of data that remains in its home cloud.
Storage Transfer Service (STS) is a fully managed, highly scalable service that automates transfers of large amounts of data between object and file storage systems. Transfers flow from the source system to a destination Cloud Storage bucket, and supported sources include Cloud Storage, Amazon S3, Azure storage services, on-premises network file system (NFS) storage, and public HTTP URLs. STS supports full and incremental copies, automatic retry of transient errors, scheduled recurring transfers, and monitoring of file counts, speed, and error samples. Choose STS when data must actually reside in Google Cloud for processing, backup, or migration.
The choice of transfer tool depends on source location and available bandwidth:
| Source | Scenario | Suggested product |
|---|---|---|
| Another cloud to Google Cloud | Any network condition, Google provisions connection between data centers | Storage Transfer Service |
| Cloud Storage bucket to different bucket | Copy without transformation | Storage Transfer Service |
| Private data center to Google Cloud | Enough bandwidth, small to medium or ad hoc needs | gcloud storage command |
| Private data center to Google Cloud | Enough bandwidth, large scale up to petabytes and billions of files, scheduled or managed | Storage Transfer Service for on-premises data |
| Private data center to Google Cloud | Not enough bandwidth to meet deadline | Transfer Appliance |
The gcloud storage command is the standard tool for small to medium transfers, streaming program output to Cloud Storage, and syncing a directory with low latency. Storage Transfer Service for on-premises data uses software agents in the data center to crawl NFS directories and create matching objects in Cloud Storage. Transfer Appliance is offline hardware shipped to the site for capture and then shipped back to Google when bandwidth is limited, unavailable, or too costly.
Exam tip: BigQuery Omni queries data where it already lives on AWS or Azure, while Storage Transfer Service copies data from AWS, Azure, or on-premises into Cloud Storage.
The handover networking pattern is the recommended way to ingest data from transactional systems in a private environment into Google Cloud. Data flows from the on-premises system to Pub/Sub queues or Cloud Storage buckets, and those queues or buckets then serve as sources for data-processing pipelines. If analytical results must be fed back to transactional systems, the handover pattern is combined with the gated egress pattern, which controls outbound flow from Google Cloud back to the private environment.
Once data lands in Google Cloud, fully managed services build pipelines and protect data. Cloud Data Fusion graphically builds extract, transform, and load (ETL) and extract, load, and transform (ELT) pipelines, while Dataflow defines and runs complex transformations using the Apache Beam model. Sensitive Data Protection discovers, classifies, and de-identifies sensitive data such as personally identifiable information (PII) by masking, encrypting, or replacing it. For long-term, high-volume exchange between clouds, Cross-Cloud Interconnect (CCI) provides high-bandwidth dedicated connectivity in supported locations, with connectivity-layer encryption from VPN tunnels, high-availability (HA) VPN over Cloud Interconnect, or media access control security (MACsec).
Storage Transfer Service for on-premises data depends on correctly deployed agents and operational controls in the data center. After setup, an operator provides a source directory, destination bucket, and time or schedule in the Google Cloud console, and the service recursively crawls subdirectories to create same-named objects in Cloud Storage. Agents are fault tolerant, self-updating, and self-healing, so the transfer continues with remaining agents if one agent goes down. Successful operations follow five practices:
Identity federation is the architecture for letting identities managed by an external identity provider (IdP) access Google Cloud without creating separate Google accounts for each user. Four patterns divide this work: Cloud Identity or Google Workspace federation, Workforce Identity Federation without System for Cross-domain Identity Management (SCIM), Workforce Identity Federation with SCIM, and Hybrid Cloud Identity and Workforce Identity Federation. The correct pattern depends on service portfolio, data residency, and integration needs such as Gemini Enterprise with Microsoft 365.
Service portfolio matters because Google services differ by service model and authorization model. Software as a Service (SaaS) services such as Gmail, Google Ads, or the Gemini Enterprise app are fully managed for a broad audience, while Platform as a Service (PaaS) or Infrastructure as a Service (IaaS) services such as BigQuery or Compute Engine serve technical users who build workloads. Authorization is either Identity and Access Management (IAM), used by most Google Cloud services for fine-grained access, or service-specific authorization, used by services such as Google Ads, Looker, Google Workspace, and Google Maps. Data residency also differs: Cloud Identity and Google Workspace may store service data such as names, attributes, and group memberships across Google data centers, while Workforce Identity Federation stores service data in Google Cloud regions.
Use Dedicated Interconnect for sustained transfers typically 10 Gbps or more when deterministic performance with a service-level agreement is required and the team can manage routing equipment in a supported colocation facility. Use Partner Interconnect when larger-than-VPN transfers need private dedicated capacity normally below 10 Gbps per connection without building direct cross-connects, because the partner has already established connections to the Google network.
BigQuery Omni runs BigQuery compute in Amazon Web Services (AWS) or Microsoft Azure regions to query data stored in Amazon Simple Storage Service (Amazon S3) or Azure Blob Storage without copying entire datasets. Storage Transfer Service (STS) copies data from sources such as Amazon Simple Storage Service (Amazon S3), Azure storage services, on-premises network file system storage, and public HTTP URLs into a destination Cloud Storage bucket. Choose BigQuery Omni when the requirement is cross-cloud analysis of data that remains in its home cloud, and choose Storage Transfer Service (STS) when data must actually reside in Google Cloud for processing, backup, or migration.
Cloud Router uses Border Gateway Protocol (BGP) sessions to exchange routes with the external router so both the Virtual Private Cloud (VPC) network and the on-premises or other-cloud network learn reachable prefixes automatically instead of maintaining static routes on each side. When a link, tunnel, or path fails, Border Gateway Protocol (BGP) withdraws unreachable routes and Cloud Router directs traffic to remaining healthy paths, which enables failover and supports active/active HA VPN designs.
Choose Cross-Cloud Interconnect (CCI) when you need dedicated multicloud capacity without managing hardware and colocation relationships. It provides private and secure connectivity between Google Cloud and supported cloud providers in specific locations with line-rate performance and high-availability options of 99.9% and 99.99%, but you must choose a location supported by the remote cloud provider and each location supports VLAN attachments in only a subset of Google Cloud regions.
An enterprise runs microservice workloads across Google Kubernetes Engine (GKE) on Google Cloud, on-premises VMware environments, and AWS. The organization requires a centralized management strategy to achieve the following goals:
Which combination of GKE Enterprise capabilities should the cloud architect recommend?
Last updated