On Sun, Nov 8, 2009 at 11:20 AM, Joe Auty <j...@netmusician.org> wrote:
> Tim Cook wrote: > > On Sun, Nov 8, 2009 at 2:03 AM, besson3c <j...@netmusician.org> wrote: > >> I'm entertaining something which might be a little wacky, I'm wondering >> what your general reaction to this scheme might be :) >> >> >> I would like to invest in some sort of storage appliance, and I like the >> idea of something I can grow over time, something that isn't tethered to my >> servers (i.e. not direct attach), as I'd like to keep this storage appliance >> beyond the life of my servers. Therefore, a RAID 5 or higher type setup in a >> separate 2U chassis is attractive to me. >> >> I do a lot of virtualization on my servers, and currently my VM host is >> running VMWare Server. It seems like the way forward is with software based >> RAID with sophisticated file systems such as ZFS or BTRFS rather than a >> hardware RAID card and "dumber" file system. I really like what ZFS brings >> to the table in terms of RAID-Z and more, so I'm thinking that it might be >> smart to skip getting a hardware RAID card and jump into using ZFS. >> >> The obvious problem at this point is that ZFS is not available for Linux >> yet, and BTRFS is not yet ready for production usage. So, I'm exploring some >> options. One option is to just get that RAID card and reassess all of this >> when BTRFS is ready, but the other option is the following... >> >> What if I were to run a FreeBSD VM and present it several vdisks, format >> these as ZFS, and serve up ZFS shares through this VM? I realize that I'm >> getting the sort of userland conveniences of ZFS this way since the host >> would still be writing to an EXT3/4 volume, but on the other hand perhaps >> these conveniences and other benefits would be worthwhile? What would I be >> missing out on, despite no assurances of the same integrity given the >> underlying EXT3/4 volume? >> >> What do you think, would setting up a VM solely for hosting ZFS shares be >> worth my while as a sort of bridge to BTRFS? I realize that I'd have to >> allocate a lot of RAM to this VM, I'm prepared to do that. >> >> >> Is this idea retarded? Something you would recommend or do yourself? All >> of this convenience is pointless if there will be significant problems, I >> would like to eventually serve production servers this way. Fairly low >> volume ones, but still important to me. >> >> > Why not just convert the VM's to run in virtualbox and run Solaris directly > on the hardware? > > > That's another possibility, but it depends on how Virtualbox stacks up > against VMWare Server. At this point a lot of planning would be necessary to > switch to something else, although this is possibility. > > How would Virtualbox stack up against VMWare Server? Last I checked it > doesn't have a remote console of any sort, which would be a deal breaker. > Can I disable allocating virtual memory to Virtualbox VMs? Can I get my VMs > to auto boot in a specific order at runlevel 3? Can I control my VMs via the > command line? I thought Virtualbox was GUI only, designed for Desktop use > primarily? > > This switch will only make sense if all of this points to a net positive. > > Why are you running VMware server at all if those are your requirements? Nothing in your requirements explain why you would choose something with the overhead of VMware server over ESX. With those requirements, I'd point you at Sun xVM. In any case, while I can't answer all of your questions as I don't use Virtualbox: yes, you can control VM's from the command line. VMware server is designed primarily for Desktop use, hence my confusion with your choice. --Tim
_______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
