Hey folks — looking for some quick feedback on a design using VAST block devices for Versity ScoutAM (metadata + primary cache layer, scale: hundreds of TB).
A few specific questions:
• Device sizing: While we can provision up to ~1 PB per device, is it better to aggregate multiple smaller devices with mdadm RAID0 for performance / parallelism? Any trade-offs to be aware of?
• Snapshots: How do VAST snapshots behave if we stripe across multiple block devices (e.g., mdadm RAID0)? Are snapshots still consistent / manageable at that layer?
• Shared FS (ScoutFS): ScoutAM uses ScoutFS (shared filesystem across multiple hosts). Is it supported / safe to mount the same VAST block device across multiple hosts in this setup? Any caveats around fencing / consistency?
Context: workload is metadata-heavy + cache layer, so mix of small IO + some throughput sensitivity.
https://github.com/versity/scoutfs/blob/main/README.md
Appreciate any pointers, best practices.