Glossary

Cloud storage replication

Cloud storage replication automatically creates and maintains copies of data across storage systems, locations or cloud providers, so data survives if the primary fails. It protects against losing a site. It does not protect against bad writes, because deletions and encrypted files replicate like anything else.

How does replication confirm a write?

Two modes differ in when the application is told a write succeeded. Synchronous replication confirms only after every copy has the data. There is no lag, and writes are slower by the round trip to the farthest copy. Asynchronous replication confirms first and ships the change afterward. Writes are fast, and a short window exists in which the replica trails the primary. That window is the recovery point: if the primary is lost, changes inside it are gone. How the two modes map onto distance is covered in multi-region storage.

Why is a replica not a safety net by itself?

Replication copies whatever happens at the source. When ransomware encrypts a bucket, the encrypted objects replicate within seconds to minutes, and the second copy is encrypted too. A mistaken bulk delete, or one made with a stolen key, is copied the same way. The replica was built to be identical, so it is.

Two controls on the target change that. Versioning keeps earlier versions of an object when a new write or delete arrives, so an old good version remains. S3 Object Lock prevents versions from being deleted or overwritten until a set date, so a replicated attack cannot remove them. Without both on the target, the arrangement is data replication for availability and not a route to recovery.

What does it cost to keep copies?

  • Every replica multiplies stored capacity: two copies are twice the storage bill, three copies three times.
  • Replication across regions or providers adds transfer charges for the data that moves, covered under egress fees.
  • Versioning adds capacity for every retained version, so change rate and retention length feed the estimate.

How are the mode and the target chosen?

The recovery point objective and the distance between sites choose the mode. A tight recovery point objective between nearby sites points to synchronous replication. A distant second region or a second provider points to asynchronous replication, with the lag accepted and measured. The second copy's place in a wider plan is the subject of disaster recovery, where restore tests matter as much as the copy.

Frequently asked questions

Does cross-region replication run in both directions?

It can be one-way, from primary to replica, or two-way. Two-way replication needs rules for conflicting writes and spreads deletions and damage in both directions.

Is replication the same as backup?

No. A backup keeps earlier restorable versions that stay available after a deletion. Replication keeps the current state in more than one place.

How is replication lag monitored?

By tracking the age of the oldest object not yet copied, or the count and size of objects pending. Alerts on that figure show when the recovery point is slipping.