| |||||||||||||
KS2007: Mini-summit reportsKS2007: Mini-summit reportsPosted Sep 6, 2007 9:44 UTC (Thu) by jengelh (subscriber, #33263)Parent article: KS2007: Mini-summit reports
>So "object-based storage" devices are on the horizon, though several years away still. These drives, for all practical purposes, implement the filesystem themselves on the disk.
Today, I can choose what filesystem I want, and if performance sucks, it is the filesystem to blame. I can choose to replace the filesystem if I desire so. Now they want a hardware-based filesystem. Suppose it sucks as much as VFAT. Statistically, there is always at least one manufacturer who does not get things right. Then I would be stuck with a slow FAT. Software-based filesystems (e.g. in the kernel) would probably continue to exist, but suffer from the hardware filesystem. And exchanging harddisks is quite pricey compared to re-mkfs'ing. That is a no-thanks for now.
(Log in to post comments)
KS2007: Mini-summit reports Posted Sep 7, 2007 12:15 UTC (Fri) by liljencrantz (subscriber, #28458) [Link] The article hints that not a complete hardware file system interface is in the works, only the parts related to reading/writing separate files. Anything else would be silly, since it would be impossible to implement the wildly varying file system semantics implemented by e.g. Unix and Windows.
KS2007: Mini-summit reports Posted Sep 11, 2007 5:25 UTC (Tue) by gdt (subscriber, #6284) [Link] The point of object-based storage isn't to make a single disk more efficient. It is to make a set of virtualised disks more efficient.
|
|||||||||||||