RE: Windows Integration and Longhorn

Hi John,

Yes, and that's exactly the problem.

Please read the link i've posted. You will get an USN rollback.

Active Directory uses multimaster replication, and it uses USN to keep track of
replicated and received changes, restoring a disk image will give you a DC in
an inconsistent state.

There are special procedures to be followed when restoring a domain controller
- using a disk image ist _not_ a proper way to do this.

If you don't believe me, or the Microsoft KB, please talk to one of your Active
Directory guys.




-----Original Message-----
From: midrange-l-bounces@xxxxxxxxxxxx on behalf of Jones, John (US)
Sent: Mon 02.04.2007 22:28
To: Midrange Systems Technical Discussion
Subject: RE: Windows Integration and Longhorn

No, with NSW storage spaces you are not 'restoring' the DC. You are
replicating it's disk image. From an x86 perspective, the disk image is
the same as it was before; there was no change, no restore, no copy, no
changes to the boot sector, no changes to the registry, no changes to
detected hardware, no changes to anything at all. Windows, AD, etc.
thinks it was shut down and later restarted. That's all.

Our IXS servers were set up once. They don't require any day-to-day
admin that isn't covered by normal iSeries operations. They are
manually touched for issues like Windows Update, but since they're on a
private hardened subnet that's really optional/irregular.

We didn't size up the hardware to support the IXS. More the reverse; we
have excess DASD capacity as we bought enough arms to ensure good
performance on iSeries workloads. That resulted in spare capacity which
we've used for the IXS. I do agree, though, that buying iSeries DASD
expressly for supporting IXS servers isn't logical unless the other
benefits justify the expense.

Tower-wise, you can get 1.4TB after RAID5 in a 5U 0595. I've got 3 of
'em. Or get the new 24 disk tower and double that; 2.8TB of RAIDed 15K
RPM DASD in 5U.

For the processing power you get, a 2GHz PentiumM, the IXS is not a good
deal. However, factor in the reduced admin, configuration flexibility,
eliminated footprint, etc. and for some workloads it makes a great deal
of sense. Not for all workloads, but for some.

Remote KVM: never needed it. And even if I thought I'd use it, from a
cost-effective architecture standpoint it makes way more sense to buy
that feature once (in the KVM switch) v. once for every server in the
rack.