@Patrik We do have performance benchmarking numbers for HyperScale X but our observed metrics are only around rebuild of a disk. Both N12 and N24 servers have shown similar rebuild performance of a disk, 300GB/hour. Now rebuild of a node is nothing but rebuild of all disks and while all disks are rebuilt in parallel, we expect some overhead in parallel processing. We do not have an observed metric on what that overhead is, but taking your example of a 16 TB drive and N24, one failed disk could roughly take roughly 48 hours and a failed node could roughly take 48-72 hours (not observed but an approximation) due to parallel rebuild of all disks. Note that when a node fails, meta data (DDB, index, File system catalog) too is rebuilt automatically, but that should be lower than that of data rebuild time.

~Pavan Bedadala