A company is deploying a new web application and needs shared storage for Linux application servers. The dataset is up to 100 GB. Peak activity lasts 3 hours daily and requires 225 MiB/s of read throughput. The solution must be Multi-AZ and replicate a copy of the data to another AWS Region for disaster recovery with an RPO under 1 hour. Which design meets these needs?

EFS is a fully managed, scalable, and highly available file system suitable for Linux workloads, meeting the Multi-AZ requirement. With a 100 GB dataset, EFS's default throughput mode would provide 50 MiB/s (50 MiB/s per TiB for Standard storage class), which is insufficient. Provisioning 75 MiB/s of throughput for a 100 GB file system is cost-effective and provides the required 225 MiB/s peak throughput (as EFS throughput scales with data size, and provisioned throughput adds to this base). EFS replication provides an RPO under 1 hour for disaster recovery. FSx for Lustre is optimized for high-performance computing and is generally overkill and more complex for a web application's shared storage, and AWS Backup for DR might not meet the RPO. gp3 EBS with Multi-Attach is limited to a single Availability Zone and is not a shared file system across multiple instances in different AZs. FSx for OpenZFS is also a high-performance file system, and DataSync on a schedule might not consistently meet the RPO under 1 hour, depending on the data change rate and schedule.