Hi folks! Thinking about running a WinXP instance in VirtualBox on OpenSolaris, using a 20 GB harddisk-file. However, I am worried about fragmentation... with the constant reading/writing that WinXP does... will not the fragmentation of the hardrive-file in ZFS be humongus?? Best regards, gigantex. -- This message posted from opensolaris.org
On Thu, 20 Aug 2009, Johan Eliasson wrote:> Thinking about running a WinXP instance in VirtualBox on > OpenSolaris, using a 20 GB harddisk-file. However, I am worried > about fragmentation... with the constant reading/writing that WinXP > does... will not the fragmentation of the hardrive-file in ZFS be > humongus??With the default block size of 128K bytes, the fragmentation can not possibly be "humungous". NTFS usually uses 4K or 8K block sizes and has a max of 64K. Unfortunately, if NTFS updates just one 4K block, zfs needs to COW a whole 128K block. If NTFS does sequential reads and writes and the NTFS filesystem is not fragmented, then whole sequential 128K blocks should be updated and zfs won''t be terribly fragmented either. So defrag your NTFS regularly, but keep in mind that doing so will bloat any zfs snapshots. Bob -- Bob Friesenhahn bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/ GraphicsMagick Maintainer, http://www.GraphicsMagick.org/
So a full NTFS defrag should result in just a long sequential ZFS write? -- This message posted from opensolaris.org
On Thu, 20 Aug 2009, Johan Eliasson wrote:> So a full NTFS defrag should result in just a long sequential ZFS write?That would depend on how the defrag algorithm works, how often NTFS issues a synchronous write request (cache flush), how much memory is installed on your Solaris system, and how fast the defrag works. Various people have stated here that VirtualBox intentionally does not pass through cache flush requests. Solaris will buffer writes up to 30 seconds so if you have enough RAM, then quite a lot of writes may be coalesced into a nice order. If defrag works fast enough, then it may give Solaris more to work with so that the volume is less fragmented. Bob -- Bob Friesenhahn bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/ GraphicsMagick Maintainer, http://www.GraphicsMagick.org/
On Fri, August 21, 2009 10:32, Bob Friesenhahn wrote:> Various people have stated here that VirtualBox intentionally does not > pass through cache flush requests.The current default value ignores flush requests from guest OSes, but this can''t tweaked via a parameter (11.1.3 Responding to guest IDE flush requests):> If desired, the virtual disk images (VDI) can be flushed when the guest > issues the IDE FLUSH CACHE command. Normally these requests are ignored > for improved performance. To enable flushing, issue the following command: > > VBoxManage setextradata VMNAME > "VBoxInternal/Devices/piix3ide/0/LUN#[x]/Config/IgnoreFlush" 0http://www.virtualbox.org/manual/UserManual.html#id2531504
On Fri, August 21, 2009 11:18, David Magda wrote:> The current default value ignores flush requests from guest OSes, but this > can''t tweaked via a parameter (11.1.3 Responding to guest IDE flush > requests):s/can''t/can be/