>>>>> "re" == Richard Elling <richard.ell...@sun.com> writes:
re> Flash has been around for well over 25 years there is NOR flash and NAND flash, though. I think NOR is 25 years old, and MLC and SLC FLASH are both NAND right? NOR and NAND have completely different behavior and implementation, and even within NAND the number of tolerated write cycles varies wildly MLC vs SLC and vendor vs vendor. Also wasn't someone saying the cheapo USB sticks do wear leveling in 16MB chunks, so if one of the chunks is hotter than others you might blow it sooner than you expect based on device-wide write cycles * size / bandwidth? software people would assume the wear leveling chunk size is the entire device, otherwise what does ``level'' mean, but apparently the electrical engineer monkeys have a different idea. The quality or chunk-size of wear leveling could vary from one device to another. I think hard disks are a little different in their failure behavior after increasing 100x in capacity, too, though. re> Trivia: Sun has been shipping flash memory for nearly its re> entire history. are you talking about the firmware? because that's NOR FLASH which is completely different. I'm not saying don't use it, but this sounds too much like Apple telling us 400 megaBIT/s firewire is faster than 80 megaBYTE/s parallel-SCSI. re> It occurs to me that you might be too young to remember that re> format(1m) was the tool used to do media analysis and map bad re> sectors before those smarts were moved onto the disk ? ;-) yeah im old enough to remember. the smarts stayed redundantly in format long after it was moved into the disk. I thought one of those netapp .pdf's said they deliberately tell some of their SCSI/FC disks to stop doing reallocation and pass bad block errors up the stack. but aside from that all these SCSI disks do it, even the 5.25" ones. I'm old enough to remember that every SCSI Sun system I've used including even VME-based systems and Sun3/60's use SCSI disks which would do their own bad block remapping. I haven't used SMD disks. I used ST506 and ESDI disks in peecees, and with those you got a sheet of dot-matrix printout taped to the top of the drive by the manufacturer. The factory test for bad sectors with special controller boards that you don't have, to find marginal sectors you will miss if you do your own ``low-level format'' scan---although the disk layer does have to do remapping, and although you do scan for bad sectors during low-level format, with ST506 and ESDI disks you will not scan any bad sectors not marked on the printout unless your disk is failing, and you must use the printout to avoid marginal sectors.
pgpNnGjvgyNHJ.pgp
Description: PGP signature
_______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss