To protect your application data in a Google Distributed Cloud (GDC) air-gapped multi-zone
universe and maintain highly available applications, you can implement a data
protection strategy that is resilient to local outages or failures.
GDC provides data replication strategies for
[object storage](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform/pa-user/storage.md.txt#object_storage) and
[block storage](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform/pa-user/storage.md.txt#block_storage) so you can
maintain failover procedures for primary and secondary zones in your universe.

This document is for the following audience groups:

- Platform administrators, such as IT administrators, who are responsible for developing disaster recovery workflows.
- Application operators, such as application developers, who are responsible for developing and maintaining applications in a GDC universe.

For more information, see
[Audiences for GDC air-gapped documentation](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/resources/audiences.md.txt).

## Storage replication for disaster recovery

You can set up robust data protection for your application storage in a
multi-zone universe using *asynchronous data replication* for disaster recovery.
This approach involves copying data from a primary zone to a secondary zone at
periodic intervals. This mechanism keeps your data protected and accessible if
the primary zone experiences an outage.

Data replication for object storage uses dual zone buckets to automatically
replicate your data, and doesn't require manual intervention. For more
information about creating a dual zone bucket, see
[Create storage buckets](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform/pa-user/create-storage-buckets.md.txt#cli).

Data replication for block storage uses dual zone persistent volumes to
replicate your data, and requires a volume failover procedure. For more
information, see
[Replicate volumes asynchronously](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/application/ao-user/async-block-replication.md.txt).

After you configure data replication, your data follows a sequential *failover
procedure* when the primary zone is offline. The failover procedures are
distinct for block and object storage replication. However, both data
replication strategies follow these critical stages:

1. Verify the primary zone outage.
2. Stop the replication from the primary zone.
3. Promote the backup secondary zone to assume the role of the primary zone with manual intervention or a pre-configured failover.
4. Verify the operational status of the new primary zone.

Reach out to a member of the infrastructure operator group to confirm your two
zones are configured for asynchronous data replication.

The inherent delay that comes with asynchronous data replication means that this
setup is most useful for systems that require a low, but non-zero recovery point
objective (RPO). If your system requires minimal data loss, but can tolerate a
small predefined maximum amount of data loss measured in time, usually related
to data generated just immediately before a disaster event that could be
potentially unrecoverable, then asynchronous data replication is a valuable
feature to implement for your applications.

An example of a low non-zero RPO might be a financial trading platform with an
RPO of five minutes, where asynchronous data replication is set to copy trade
data to a secondary disaster recovery zone every two minutes:

- This is a *low RPO* scenario because the five minutes represents the minimum acceptable data loss window for the high-volume system.
- It's a *non-zero RPO* scenario because the inherent delay in asynchronous replication of two minute intervals means that there is a small window of time where data has not yet been copied, resulting in potential loss.

You must work with your infrastructure operator group to define your dual zone
asynchronous storage replication workflow, and verify the infrastructure's data
replication capabilities support your RPO requirements.

## Limitations

*Synchronous data replication* isn't supported in GDC
air-gapped. Synchronous data replication maintains strict consistency between
data in two zones by immediately duplicating all data written from a primary
zone to the secondary zone, which provides zero RPO in disaster scenarios.

## What's next

- [High availability for your apps](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform-application/pa-ao-operations/ha-apps/overview.md.txt)
- [Deploy a highly available VM application](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform-application/pa-ao-operations/ha-apps/deploy-ha-vm-app.md.txt)
- [Deploy a highly available container application](https://docs.cloud.google.com/distributed-cloud/hosted/docs/latest/gdcag/platform-application/pa-ao-operations/ha-apps/deploy-ha-container-app.md.txt)