Using tethered endpoints

When deciding how to configure your IBM Cloud® Object Storage instance, consider how the endpoints reflect your needs for resiliency and access.

When you use a Cross Region bucket, it is possible to direct your accesses to a tethered endpoint associated with a specific Cross Region metropolitan area, rather than connecting to the nearest available Cross Region metropolitan area. In contrast to the geo endpoint, when you send requests to a tethered end point there is no automated failover if that region becomes unavailable. Applications that direct traffic to a tethered endpoint must implement appropriate failover logic internally to achieve the availability advantages of the Cross Region storage.

One reason for using a tethered endpoint is to control where data ingress and egress occurs while still distributing the data across the widest possible area. Imagine an application running in the us-south region that wants to store data in a US cross-region bucket but wants to ensure that all read and write requests remain in the Dallas area:

1. The application creates a client using the https://s3.private.dal.us.cloud-object-storage.appdomain.cloud endpoint.
2. The Object Storage service in Dallas suffers an outage.
3. The application detects a persistent failure trying to use the tethered endpoint.
4. The application recognizes the need to fail over to a different tethered endpoint, such as San Jose.
5. The application creates a new client using the https://s3.private.sjc.us.cloud-object-storage.appdomain.cloud endpoint.
6. Connectivity is resumed, and access can be re-routed to Dallas when service is restored.

When sending requests to a tethered endpoint there is no automated failover if that region becomes unavailable.

For contrast, imagine another application using the normal US cross-region endpoint:

1. The application creates a client using the https://s3.us.cloud-object-storage.appdomain.cloud endpoint.
2. The Object Storage service in Dallas suffers an outage.
3. All Object Storage requests are automatically rerouted to San Jose or Washington until service is restored.