DEV Community

Wakeup Flower
Wakeup Flower

Posted on

S3 cross region replication

The situation

  • Data size: 1 PB (massive).
  • Source: S3 bucket in us-west-1.
  • Destination: S3 bucket in us-east-1.
  • Constraints:

    • Already in AWS (data is in S3).
    • On-prem DC restrictions: can’t use Snowball.
    • It’s a one-time copy, not continuous replication.

  1. aws s3 sync command
  • Pros: Straightforward, flexible, and works across Regions.
  • Cons: Requires significant time and compute (you need EC2 instances or similar to run it), and moving 1 PB over the public internet (or even across AWS backbone) will take a long time. Could be expensive in terms of data transfer costs.
  1. Amazon S3 Batch Replication (S3 Replication with batch operations)
  • Pros:

    • Runs inside AWS, no need to pull/push from external compute.
    • Leverages cross-Region replication features.
    • Scales automatically, optimized for huge datasets (PB scale).
    • You can configure one-time replication and then delete the rule after the copy finishes.
  • Cons: Slightly more setup effort than s3 sync.


✅ Correct answer

  • Best solution: S3 Batch Replication (one-time) — optimized for PB-scale cross-Region copies within AWS.
  • Why not s3 sync: It’s slower, less scalable, and requires external compute resources.
  • Why not S3TA: It only accelerates internet uploads/downloads, not bucket-to-bucket copies.

Top comments (0)