We are in the process of migrating from a single on campus data center
to a pair of collocation facilities.

The on campus data center has a TSM 5.5.4.0 server running under SLES 9
Linux in a mainframe LPAR. Z/OS runs in another LPAR. The Linux system
has the usual Linux file systems, file systems for the TSM database,
and file systems for the TSM recovery log on ECKD disk. File systems
for disk storage pools are on FCP disk. The Linux system has FCP tape
drives. The z/OS system has FICON tape drives. The disks containing
the usual Linux file systems are backed up to FICON tape by the z/OS
system, using FDR.

The DR plan for the on campus data center uses a SunGard hot site. The
Linux system is recreated with FDR restores running under a z/OS floor
system provided by SunGard. Once this is done, additional file systems
are created and populated with empty database and recovery log volumes.
The TSM database is restored to these volumes. The disk storage pools
are not recreated.

The first collocation facility has a TSM 6.6.2.0 server running under
SLES 11 Linux in a mainframe LPAR. There is no z/OS system. All disk
is FCP attached. The only tape drives currently available to Linux
are the on campus FCP drives. FCP is layered on top of DWDM for the
connection between data centers. The second collocation center, with
a similar configuration, is expected to be available in a couple of
months. We are not currently using z/VM, but we could do so if
necessary; we have an unused license that was in effect bundled with
the mainframe hardware.

Originally, the interim DR plan for the first collocation facility
(while we were waiting for the second collocation facility) was to
avoid using the facility for systems that needed fast restores after
a disaster. The long term plan was to recreate the destroyed TSM
server as a second instance on the Linux system at the surviving
collocation center. However, we have recently been asked to look
into the possibility of recreating a Version 6 TSM server at the
SunGard hot site after a disaster at one of the collocation centers.

We have looked into the possibility of using a variant of the on
campus data center DR plan, with CMS and DDR taking over the roles
of z/OS and FDR. This appears to be economically unattractive, with
a sizable capital expenditure and significant increases in network
charges to provide for sending the DDR backups over FICON layered
on top of DWDM. Thus far, potential DR plans based on the use of a
Linux rescue system have been dismissed as too slow, too unreliable,
and too hard to test.

Does anyone know of a proven bare metal recovery process for
mainframe Linux systems without FICON tape drives?