As a service producer, you can allow service consumers to provision resources
with private (RFC 1918) or public IP addresses. If service consumers want to use
private IP addresses, they must use [private services
access](https://docs.cloud.google.com/vpc/docs/private-services-access). However, service
consumers can only use private service access if your managed service
offers it. To offer private connectivity, you must complete a one-time
[onboarding process](https://docs.cloud.google.com/service-infrastructure/docs/enabling-private-services-access.md.txt#onboarding).

The onboarding process requires you to use the Service Networking
API and tenancy units. For detailed step-by-step instructions, contact your
Google representative.

## Overview

The following sections describe the components and general network topology
required to enable private services access for your managed service.

### Tenancy units

When a service consumer enables your managed service, the service create a
[tenancy unit](https://docs.cloud.google.com/service-infrastructure/docs/tenancy-units-tutorial) to formalize
a relationship between your Google Cloud organization and the service
consumer's project. Tenancy units isolate resources and billing costs between
different service consumers.

For each service consumer, you will have two tenancy units. One for your managed
service and another for the private access management service. The managed
service is the external service that you're offering to service consumers, and
the private access management service manages private connections with service
consumer VPC networks. These tenancy units must be in the same
Google Cloud organization where your managed service lives.

### Service Networking

Service Networking automates the private connectivity set up
(using VPC Network Peering) between you and the service consumer. You enable and
use Service Networking in the same project in which you created
the private access management service. This is a different project from the one
that contains your managed service.

When a service consumer creates a private connection with your managed service,
Service Networking creates a *Shared VPC host project* and a
*Shared VPC network* for you. The host project and network are created in a
predesignated Google Cloud folder in your organization. You specify this
folder name as part of the onboarding process. The project and network are
contained in a tenancy unit, so they are isolated and can only be used by that
service consumer.

After Service Networking creates the Shared VPC network,
Service Networking automatically creates a VPC
Networking Peering connection between the Shared VPC network and the service
consumer-specified VPC network.

Service consumers must also supply an allocated IP address range when they
create the private connection. This allocation reserves IP addresses that can
only be used by you, a service producer. For example, when a service consumer
provisions a resource, you use Service Networking to create
subnets in the Shared VPC network. For the subnet's IP address range,
Service Networking automatically selects a range from the
allocated range. This process prevents collisions between the Shared VPC network
and service consumer VPC network.

> [!NOTE]
> **Note:** Creating a subnet is a long running operation. You can [poll](https://docs.cloud.google.com/service-infrastructure/docs/polling-operations.md.txt) it regularly to check the status of the operation.

#### Shared VPC service projects

When your service provisions a service consumer's resource for the first time,
your managed service provisions it in a [Shared VPC service
project](https://docs.cloud.google.com/vpc/docs/shared-vpc), which is attached to the
Service Networking host project. This Shared VPC relationship
enables resources in the service project to use subnets in the Shared VPC
network.

Your managed service creates the service project in a tenancy unit and
predesignated folder, specified during the onboarding process. The folder and
tenancy unit are related to your managed service and are different from the ones
that Service Networking uses.

### Network topology

The following example shows a single service consumer that has private
connectivity to a single service producer. The service consumer provisioned two
resources in different regions. Because each resource is in a different region,
they are in different subnets.
[![](https://docs.cloud.google.com/static/service-infrastructure/images/service-networking-overview.svg)](https://docs.cloud.google.com/static/service-infrastructure/images/service-networking-overview.svg) Overview of Service Networking for service producers (click to enlarge)

- There are two Endpoints projects: one for the managed service
  and another one for the private access management service. These must be in
  the same Google Cloud organization.

- Within the Google Cloud organization, there are two folders, one for
  each of the Endpoints services. The private access management
  service folder contains a Shared VPC host project for the private connection.
  The managed service folder contains a service project for service consumer
  resources.

  - In each folder, the service consumer-related projects are contained in tenancy units. Both tenancy units are associated with `consumer-project-a`.
- Service consumers must initiate the private connection (which is also a
  VPC Network Peering connection). They must supply an allocated
  IP address range for the private connection from which the subnet IP addresses
  come. For more information about the service consumer steps, see [Configuring
  Private Services Access](https://docs.cloud.google.com/vpc/docs/configure-private-services-access).

  - If you offer multiple services, service consumers only need one private connection. All traffic to and from the service consumer goes through the Shared VPC host project.
- In a single service consumer project, multiple VPC networks can
  privately connect to your services. This requires a Shared VPC
  host project for each connected VPC network. However, all of
  those projects can be contained in the same `consumer-project-a` tenancy unit.

- In the host project, you must configure firewall rules and routes to enable
  connectivity to new resources. Because other services can use the same Shared
  VPC network, these rules can allow or deny connectivity between
  your different services.

## Onboarding process

The following list is a general outline of the onboarding process. You must
complete this process for each managed service that will offer private
connectivity. Contact your Google representative for more information.

1. Create a peering management service.

   This is a managed service that a service producer builds through
   the Service Management and Endpoints API. For more
   information, contact your Google representative.
2. Provide the following configuration information to your Google
   representative:

   - The minimum IP address range that service consumers must allocate when they connect to you, specified as an IPv4 prefix length. If you offer multiple services, you might want users to allocate a larger IP address range, such as `/16`.
   - The folder ID where your private access management service creates Shared VPC hosts projects. Use Resource Manager to find the folder ID.
   - The billing account that's associated with the organization where your private access management service creates Shared VPC host projects.
   - The principals (typically these are service account IDs) that manage the host project's network firewall rules.
3. Activate the Compute Engine API.

   For each Shared VPC host project, activate the
   `compute.googleapis.com` API, which you can do by using the
   Service Usage APIs or in the project configuration.

   After resources have been provisioned, configure firewall rules for the
   Shared VPC network in the host project. You must use the identity that you
   provided during the onboarding process to access the VPC network. If you
   offer other services, those services might use the same VPC network. Don't
   create rules that might inadvertently allow or deny traffic to other
   services.
4. Inform service consumers.

   Inform service consumers that they must establish a private connection. For
   more information, see [Configuring Private Services
   Access](https://docs.cloud.google.com/vpc/docs/private-access-options). Service consumers must provide
   the following information:
   - The name of their project and network where they want to establish private connectivity.
   - The Cloud region where the resource must be provisioned.

## What's next

- To enable Service Networking APIs, see [Getting Started with the
  Service Networking
  API](https://docs.cloud.google.com/service-infrastructure/docs/service-networking/getting-started).
- For information about Service Networking private connection limits, see [Quotas and
  Limits](https://docs.cloud.google.com/service-infrastructure/docs/quotas#networking).