Eric W. Biederman
ebiederman at lnxi.com
Fri May 7 05:16:02 CEST 2004
Oleg Goldshmidt <pub at goldshmidt.org> writes:
> ebiederman at lnxi.com (Eric W. Biederman) writes:
> > It sounds like from the description there is no way to do windows diskless
> > without emulating a disk.
> I am not sure I understand what you mean. Isn't that what iSCSI is
> about: everything above that thinks there is a local SCSI disk...
There are some efficiencies to be gained by not having a remote disk,
and instead having a remote filesystem. From everything I have
heard that does not sound like it has been implemented for windows
with NT kernels.
> > I have serious issues with iSCSI because the only implementation I have
> > seen was tremendously complex and nasty looking. Which is a real downside
> > when compared to something simple like nbd.
> Well, I don't know what implementations you have seen. There are
> undoubtedly some ugly ones. That's an implementation issue, though:
> the basic idea looks quite clean to me. Not that there are no issues...
nbd is 600 likes of obviously correct code, which can put full disk
speed onto the wire. Nor have I seen any issues with it. That
makes tough competition for iSCSI in modest deployment locations.
I still need to read the spec to see how sane iSCSI looks on the wire.
> > How well does iBOOT work on the linux side.
> Better ;-)
> > If you have an iSCSI driver for Linux.
> We do.
Cool. If iSCSI looks sane I will suggest to the etherboot guys that
we implementing booting over it :)
More information about the coreboot