A research team stores datasets in S3 and uses a Linux compute cluster for simulations. Jobs require a shared POSIX file system with very high aggregate throughput and low-latency parallel access. The team wants integration with the S3 datasets. Which service is the best fit?

1. A single sc1 EBS volume attached to all cluster nodes across different Availability Zones.
2. Amazon FSx for Windows File Server as an SMB share for the Linux simulation's required POSIX workload.
3. Use an EFS file system without evaluating its throughput configuration or the cluster's parallel-I/O requirements.
4. Amazon FSx for Lustre with an appropriate S3 data repository association.