Re:Raspberry Pi – UASP, Trim, and Boot Performance via USB
by londons_explore on Sunday September 20, 2020 @05:12PM
(#24537455)
These reports always comment on raw performance without checking if the drive actually meets the storage contract with the host.
That's like having a 'hall of fame' for the fastest marathon runners, but not having anyone keep an eye on if they took a taxi half the route...
These benchmarks should always have a section where a drive has its power pulled repeatedly while under heavy load. If, ever, a single sector becomes overwritten, rolled back, or unreadable, except those being written between a pair of write barriers, then the drive has failed the test, and all other performance results thrown out. That drive took the taxi.
That's like having a 'hall of fame' for the fastest marathon runners, but not having anyone keep an eye on if they took a taxi half the route...
These benchmarks should always have a section where a drive has its power pulled repeatedly while under heavy load. If, ever, a single sector becomes overwritten, rolled back, or unreadable, except those being written between a pair of write barriers, then the drive has failed the test, and all other performance results thrown out. That drive took the taxi.