This page provides an overview of the Bare Metal Rack HSM solution.

## Overview

Bare Metal Rack HSM is an infrastructure-as-a-service offering that lets
you deploy racks of customer-owned hardware security modules (HSMs) next to your
Google Cloud workloads. Your HSMs are deployed in PCI-compliant facilities to
meet your security, compliance, and low-latency requirements.

To support moving your workloads to the cloud, Google hosts your HSMs, providing
physical and network security, rack space, power, and network integration for a
monthly fee.

Bare Metal Rack HSM lets you contract directly with Google for placement of
your HSMs. HSMs are placed within specified colocation facilities and connect to
Google Cloud.

The Bare Metal Rack HSM solution is supported in colocation facilities with
active peering fabrics. These facilities meet and exceed Google's standards for
data center security and provide low-latency, highly available service.

### Comparison with Bare Metal HSM

Both Bare Metal Rack HSM and Bare Metal HSM let you host
your own HSMs in Google Cloud facilities. The primary difference between
the Bare Metal Rack HSM and Bare Metal HSM solutions is
scale. The following table summarizes the key differences between these solutions:

| Bare Metal HSM | Bare Metal Rack HSM |
|---|---|
| Google hosts your HSMs on a *per-device basis*. | Google hosts your HSMs on a *per-rack basis*. |
| You have logical access to your HSMs but *no physical access*. | You have logical access to your HSMs and can schedule *escorted physical access*. |
| Intended for small deployments of *10-15 HSMs* | Intended for large rack-level deployments of *100 or more HSMs* |

If you're not sure which of these solutions is right for your needs, reach out
to your account representative.

## Operational model

- Onboarding process
  - Contract: Minimum of 12 months. Premium Support is required.
  - Procurement and configuration: Your organization acquires, configures, and ships HSMs to Google.
  - Rack and stack and connect: Google deploys your HSMs and configures the Partner Interconnect connection.
  - Validation and handoff: Confirm the engineering solution and accessibility to the HSMs, test the solution, and sign off.
- Support model
  - Google provides support for rack and stack, hosting, smart hands, compliance, and Partner Interconnect connection.
  - Work with your HSM vendor for support for HSM software, licensing, tools, and troubleshooting.
  - You have physical access to your racks as needed.
- Decommission process
  - You file a request for decommissioning.
  - You must erase all data and initialize all HSMs to factory default.

## Compliance requirements

This offering is limited to HSMs that are FIPS 140-2 Level 3 certified or
better, and is not a generalized hosting or colocation service. The
Bare Metal Rack HSM solution is hosted in facilities that are fully
compliant with PCI-DSS, PCI-3DS, and SOC 1, 2, and 3. Google will support your
AOC for PCI-PIN, PCI-P2PE, and SOC compliance in all regions.

## Separation of responsibilities

It's your responsibility to obtain and provision HSMs and ship them to the
appropriate Google Cloud regions. The HSMs used are your choice, but they
must comply with the [HSM equipment requirements](https://docs.cloud.google.com/kms/docs/bare-metal-rack-hsm.md.txt#hsm_requirements).

Google pre-configures the racks, top-of-rack switches, and connectivity. The
switches are from different vendors for each pair of racks. For the
Bare Metal Rack HSM solution, you have your own dedicated racks and
switches. Google provides a racking service for your HSMs and works with you to
validate the Partner Interconnect connection. Each rack has
redundant power supplies.

### Accessing Bare Metal Rack HSMs

You have logical management access to your HSMs and are responsible for their
maintenance and management. You maintain full control of your HSMs.

Google doesn't have logical access to your HSMs, but provides and maintains the
racks, switching, and connection. Google has no access to the data or keys on
your HSMs.

Google provides a remote hands service. With notice, you can schedule an
escorted visit to the facility. You are responsible for your own compliance and
audit requirements.

At the end of your contract or the HSM's end of life, you submit a request to
decommission the HSMs and erase all data or restore the HSMs to factory
settings. After the HSMs are erased or reset and legal clearance is obtained,
the HSMs will be shipped back to you or destroyed if they can't be shipped back.

## HSM equipment requirements

This section details the physical requirements for HSMs and associated cables
for hosting HSMs in a Google facility.

The number of HSMs that may fit in a rack depends on the number of ports
available in the current model of the top-of-rack switch, the number of rack
units taken up the HSM model, and the power draw of the HSMs.

> [!NOTE]
> **Note:** You are provided with a dedicated rack that has 36 rack units available.

- Power

  - Dual AC power supplies (16A max per power supply).
- Power distribution

  - 208V line to line (for United States based locations).
  - Rack PDU providing C13 or C19 receptacles and outlets.
- Power cables (to be provided by you)

  - Rack PDU cable end should be C14 or C20 connector types.
  - 2 x 6 feet or 2 meter (preferred length) power cables.
- Network

  - Network interface controller: Dual 1g copper NICs (if applicable).
- Network cables (to be provided by you)

  - 2 x 6 feet or 2 meter (preferred length) CAT-5e or better patch cables.
- Physical dimensions

  - Rack depth: 42 inches deep.
  - Rack unit spacing: Standard EIA-310 19" rack mount with square hole mounts. You can occupy up to 4 rack units per HSM.
- Security

  - The HSMs must not be equipped with cameras or wireless networks such as Bluetooth.
  - The HSM must be FIPS 140-2 Level 3 certified or better.
- The HSM must be new equipment.

- The HSM must be fully remotely manageable.

There are no requirements for weight or cooling.

## Deployment overview

To qualify for a 99.99% uptime SLA, you must meet the following requirements:

- Deploy HSMs in a minimum of two zones---either two different [Google Cloud regions](https://docs.cloud.google.com/docs/geography-and-regions.md.txt) or, where available, two zones in the same region.
- Deploy a minimum of four HSMs per zone (at least two HSMs per rack in at least two racks).

You provide Google with the MAC address for each HSM network interface and its
assigned IP address. This information helps Google verify server-to-top-of-rack
cabling and aids troubleshooting during the deployment process.

> [!NOTE]
> **Note:** The HSM internal IP address allocations shouldn't overlap with IP addresses used in your on-premise network or in your Google Cloud project.

Network requirements will be discussed in more detail with your account
representative during the onboarding process.

### Network topology

A pair of racks at a single location is covered by a 99.9% uptime SLA.

A full deployment across two locations provides a 99.99% uptime SLA.

Applications should be designed to take advantage of this redundancy model. An
application should be able to fail over from zone 1 to zone 2 within a single
location, from HSM to HSM or from rack to rack.

Enabling the Global Routing feature lets HSMs at either location reach
Google Cloud resources in any region.

A single Partner Interconnect connection failure isn't an SLA
violation.

The following high-level diagram shows the required connectivity to achieve a
99.99% SLA on the service.

![Network topology for Bare Metal Rack HSMs](https://docs.cloud.google.com/static/kms/images/private_hsm_network.png)

- Each zone deployment contains a minimum of two racks for your use, and one switch per rack.
- The top-of-rack switches are provided by Google and are from different vendors.
- Each top-of-rack switch has a 10G [Partner Interconnect](https://docs.cloud.google.com/network-connectivity/docs/interconnect/concepts/partner-overview.md.txt) with redundant VLAN attachments for Partner Interconnect to redundant Cloud Routers.
- Each HSM should have a minimum of 2 1GE copper network interfaces with redundant connections to both top-of-rack switches. Both the management and data interfaces should have their own redundant connections to both top-of-rack switches.
- You provide the IP address allocations for the HSM networks.
- Top-of-rack switches advertise their locally attached subnets to the pair of Cloud Routers.
- You enable global dynamic routing in your virtual private cloud (VPC) to allow access to the HSMs from any Google Cloud region where you've deployed resources. Global dynamic routing is also required to qualify for 99.99% availability.
- BGP between the top-of-rack switches and the Cloud Routers in your project exchange reachability information to route between Google Cloud project resources and the HSMs.

### Networking requirements

You must complete the following steps for each set of racks in a zone to
enable your HSMs to be hosted with Google:

1. Create a redundant pair of Cloud Routers per zone using ASN16550. For
   detailed instructions, see [Creating Cloud Routers](https://docs.cloud.google.com/network-connectivity/docs/router/how-to/create-router-vpc-on-premises-network).

2. Create two redundant pairs of [VLAN attachments with
   Partner Interconnect](https://docs.cloud.google.com/network-connectivity/docs/interconnect/how-to/partner/creating-vlan-attachments) per zone using the
   Cloud Routers from the previous step. Create the attachments with the
   pre-activate option enabled. There should be a total of four attachments per
   zone. If the attachments were created without the pre-activation option
   enabled, you can [activate the connections](https://docs.cloud.google.com/network-connectivity/docs/interconnect/how-to/partner/activating-connections.md.txt) manually.

   For more information about Partner Interconnect and
   pre-activation options, see
   [Partner Interconnect overview](https://docs.cloud.google.com/network-connectivity/docs/interconnect/concepts/partner-overview.md.txt).
3. Enable global dynamic routing in the VPC network.

   - To achieve 99.99% availability, use the steps in [Establishing 99.99%
     Availability for Partner Interconnect](https://docs.cloud.google.com/network-connectivity/docs/interconnect/tutorials/partner-creating-9999-availability).
   - Deployments in a single zone have 99.9% availability until the second zone is available. For this case, see [Establishing 99.9% Availability
     for Partner Interconnect](https://docs.cloud.google.com/network-connectivity/docs/interconnect/tutorials/partner-creating-999-availability)
4. Configure firewall rules as needed to allow traffic between your premises
   and project resources.

## Location compatibility

Bare Metal Rack HSM is available in the following Cloud KMS locations:

| Geographical area | Location name | Location description | Zones per region |
|---|---|---|---|
| Americas | `us-central1` | Iowa | 1 |
| Americas | `us-south1` | Dallas | 1 |
| Americas | `us-east4` | Northern Virginia | 1 |
| Americas | `us-west1` | Oregon | 1 |
| Europe | `europe-west4` | Netherlands | 1 |
| Europe | `europe-west3` | Frankfurt | 1 |
| Americas | `southamerica-west1` | Santiago | 1 |
| Americas | `southamerica-east1` | São Paulo | 1 |
| Asia-Pacific | `australia-southeast1` | Sydney | 2 |
| Asia-Pacific | `australia-southeast2` | Melbourne | 1 |
| Middle East | `me-west1` | Tel Aviv | 2 |

> [!NOTE]
> **Note:** You can achieve a 99.99% SLA at a single region if the region contains two zones.

## Contact Google

This product is only available for customers with specific business and
technical requirements.

If you're interested in Bare Metal Rack HSM with Google, reach out to
your account representative for additional assistance.