On Wed, Jan 12, 2011 at 12:43 PM, Greg KH
On Wed, Jan 12, 2011 at 06:33:26PM +0100, Johannes Nohl wrote:
2011/1/12 Greg KH
: On Wed, Jan 12, 2011 at 05:29:58PM +0100, Johannes Nohl wrote:
I'd like to ask for feedback. How do you treat your SSD running openSUSE 11.3?
Like a normal disk, not worrying about any TRIM crud or anything else like that. You should not have to ever send those types of commands to the disk unless you feel like potentially causing problems :)
Great, but why does http://en.opensuse.org/SDB:SSD_discard_%28trim%29_support suggest to use a tool like wiper.sh then? I would prefer to not interact with the disk by intention at all.
I do not know at all. But realize, these disks are designed to work in systems that do not have TRIM support at all in them (i.e. some other operating system), so not running these commands on your Linux system is just fine.
thanks,
greg k-h
Greg, Windows 7 has near real-time trim support and lots of laptops run that by now. A kernel developer traced windows 7 with a SATA bus analyzer to see if Windows 7 was doing the same thing as the linux kernel or not. I believe to his surprise, Windows 7 was more sophisticated in its handling. As to linux kernel's "mount --discard" implementation, it is highly non-optimized and has no real world net gain use cases. In theory there are some NDA only pre-production units that show it as a net benefit. But I've been hearing that for a couple years now, and from the same people each time. Thus the need for wiper.sh in a linux environment. It's been in openSUSE since 11.2. opensuse 11.4 should have a new kernel based batched discard (2.6.37 required), but it offers little that wiper.sh doesn't already have. The latest util-linux has a small userspace app to call the new ioctl. A cron job or similar will still be need to call the new userspace app. The main advantage of the new kernel code is eventually the DM and MD portions of the block layer will add trim range translation support, but until that happens, wiper.sh is superior from a performance perspective because it handles bigger trim range payloads and it handles more translations! fyi: The performance difference between wiper.sh and the new kernel patched discard support is not enough for both to be supported for the long term in my opinion, but in a opensource world, you never know. Redhat people wrote the new kernel logic, so they will likely ensure it is supported long term. Greg -- To unsubscribe, e-mail: opensuse-project+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-project+help@opensuse.org