Amazon recently announced Elastic File System Replication to keep an up-to-date copy of a network file system in a second AWS region or within the same region.
The new feature is designed to support business continuity and meet compliance requirements as part of a disaster recovery strategy, targeting a Recovery Time Objective (RTO) and a Recovery Point Objective (RPO) of minutes. Jeff Barr, vice president and chief evangelist at AWS, explains how it works:
All replication traffic stays on the AWS global backbone, and most changes are replicated within a minute, with an overall RPO of 15 minutes for most file systems. Replication does not consume any burst credits and it does not count against the provisioned throughput of the file system.
Replication allows to store secondary data copies in different regions and customers can choose the destination region, and the storage class for the copy: Standard, storing data in three AZs within a region, and One Zone, storing data in a single AZ for a lower cost. Massimo Re Ferrè, principal technologist at AWS, comments:
This is a big step forward for customers exploring multi-region deployments.
and Drew Mingl, CDO at State of Utah, adds:
This will be more useful than people realise.
The replication is block-based and it is not crash-consistent: Amazon EFS tracks modifications to the blocks that are used to store files and metadata, and replicates the changes at up to 300 MB per second, a process that can be monitored with the CloudWatch metric TimeSinceLastSync.
A fail-over to the replica can be triggered deleting the replication rule, either from the source or the destination file system, and the read-only destination becomes a writable file system that can be used in the recovery process.
Source: https://aws.amazon.com/blogs/aws/new-replication-for-amazon-elastic-file-system-efs/
In a Reddit thread, many developers highlight that replication has been a missing feature for many years, and user jamsan920 explains:
We were attempting to use DataSync to meet some fairly aggressive RTOs for data stored on EFS and it was woefully inequipped to handle the amount of data we had without jumping through some major hoops - just under 2 TB of data which doesn't seem too significant. I'm really hopeful for this, as it was the one of the few gaps remaining to take our DR to a second region with the low RTO that we wanted.
The feature is currently available in a subset of AWS regions and customers pay for any cross-region or intra-region data transfer charges and the standard EFS storage fees for the original and replica file systems.