Slashdot Mirror
Microsoft Releases Public Beta of Data Protection
Torrey Clark writes "Microsoft has released the public beta of its disk-to-disk backup product, Data Protection Manager. The product is designed to make backups easier than simply backing up to tape. Disk-to-disk backup completes images in significantly less time, meaning much less downtime for systems during backups."
We've been doing disk-to-disk for a year or so now using rsync's --link-dest feature to create apparently complete mirrors each night, but with only those changed files actually occupying disk space (beyond that of a symlink). Makes restoration an absolute breeze compared to tape, but I'm not sure if this M/S effort does the same? *runs off to look*
What's wrong with:
o
m /e valuation/faq.mspx
dd if=/dev/hdb1 of=/mnt/hdh1/path/to/desired/backup/image/here.is
Oh, not available for Windows, so you'll have to buy a product instead. But isn't dd much easier than using a program that expires after 270 days.
http://www.microsoft.com/windowsserversystem/dp
Q. When does the DPM beta expire?
A. The Data Protection Manager software expires 270 days after installation.
You don't with Windows either, but you have to make sure there are no handles to critical files when you do. After that you can just use dd or whatever, I use dd, because I came from a unix background and found it the most simple solution.
If you're not in the know and still reboots, why not just g4u?
Disk to disk backup would require the system to be shutdown, drive added, removed and reboot, configure etc. How this is better than Tape Backup? In fact Tape Backups do not require downtime at all.
Unless they are talking about removable media like CD/DVD/USB devices, this does not make sense. But in that case, this method would be useless for data larger than maximum allowed space on these devices.
Someone criticized the "downtime" thing. Frankly, in order to get a good backup, any other processes running on data should not be in flux or the backup itself could be corrupt. So even in most conventional backup schemes, there is a period of time in which backups run and nothing else does.
Another point is that I do not see where it will support operating systems other that Windows. This is to be expected, but a mature solution should be capable of backing up multiple operating systems as many sites I have seen have a heterogenus computing environment. At my site there are Windows servers but there are also Novell, Linux and SunOS. Is there a solution for those too?
On the other hand, if we're talking about what essentially amounts to "dd" I am sure there could be a handy Knoppix CD created to suit the task in some automated way. It could actually be quite simple in that at a certain time of day (night?) power to a bootable external CD drive is enabled, the system is scheduled to reboot at the same time, it boots from CD, runs "dd" per the scripting in the custom Knoppix where it finishes the job by writing out information to a log file about success or failure and then reboots the computer again. That's just off the top of my head but I am sure that even more elegant schemes could be cooked up. This solution would be effective at creating viable images at a good speed and could even utilize compression along the way.
If Microsoft wants to make a "ghost" backup, then maybe they should just license the technology from Symantec.
Well, on linux and unix, it's trivial, as most enterprise-grade systems are configured with an LVM,which allows snapshotting of a consistent filesystem state when combined with a filesystem such as XFS. On linux you just use xfs_freeze to quiesce the filesystem, lvcreate a snapshot volume (large enough to hold changes that will happen to the main system during the backup - the paranoid use a snapshot volume the size of the main volume, but I usually get away with 20% or so with a wide margin), xfs_unfreeze, and backup the snapshot volume. No application interruption, just minor performance impact.
I think what they are getting at here is the time it takes to restore the data. Rather than taking the time to restore from a tape which is very time consuming, it images to a hard drive. That drive can simply just be plugged in when the machine is down for the drive replacement. As opposed to taking down the system, replacing the drive, reloading the OS, installing the backup software, restoring the data, verifying it's integrity...
Disk-to-disk backup? In fact, I (ie. my computer!) do it every night. Simple copy command? I think that does not cut it. I'm in a tight development cycle and each day write a lot of code, documents and receieve/generate lots of data files. I need to back up all important data but surely I don't need to make backup of the executable files, temp files, OS system files and such. The solution that I use is simple: I have two hard disks in my computer. The files that I need to have back-up from, are scattered on these two drives. Now, I have made a BackUP directory on each one of these drives and put a copy of all important files in them. So, I have 3 copies of every important files: the original, and two back-ups. In case a hard-disk goes banana, I always have a copy of all important files on the other one. I run the back-up every night. Just need to copy the files that has been changed or the whole new directories made during the day. So the problem is: I need two desinations for each source. I need to be able to select which directories or even which files to back-up (or not to backup) and I need to check which files have been changed or which new files (or directories) have been created. I need to be able to schedule the back ups for midnight and I need to forget about all these details in practice as I have to focus on my work :)
How I did it? Well, I tried a script in the beginning but found it difficult to manage over the time and it was very tedius. Now I use SyncBack which is a freeware program with all these features that I need (and more! like FTP and compression to Zip, etc.). QED.
Will it back up locked files? Hmm?
Will it copy the entire filesystem so you can just drop in a replacement when your primary drive fails? Hmm?
I've already switched to samba and rsync. Microsoft's backup was outdated by at least a decade, and even failed to complete at random when I've used it for disk to disk backups. And Windows' mandatory file locking policy makes safe, reliable backups entirely impossible. An xcopy backup is even dangerous, because it temporarily locks files as it opens them for reading, potentially causing other server processes to fail if they attempt to write to the files.
From the FAQ:
a customer has to purchase a server license for every DPM server that is deployed and a Data Protection Management License (DPML) for every server they protect.
Now they have incentive to never upgrade the poor quality backup software already included in Windows. Admins will have to buy their backup software seperately or look elsewhere. Server operating systems are expected to come with _good_ backup software, so from a strictly technical sense Microsoft is being an ass.
Don't know about the rest of the world but we don't have to take systems down to backup them here.
Yeah, it's pretty pathetic. Contrast their approach with our simple "poor man's RAID" backup solution which has worked on Sun systems, *BSD systems, and GNU/Linux systems for over 10 years:
(install two identical hard drives)
dd if=/dev/hda of=/dev/hdb bs=1048576
Run as frequently as you need a backup image. This has worked, as I said, for over a decade, and has allowed quick and easy recovery of every machine we've ever had a hard drive crash on. There are theoretical limitations, e.g. a gradual failure of the source drive that causes corrupt data to overwrite the backup, so "poor man's RAID" should only be one component of one's backup strategy, not the entire strategy, but it has worked wonderfully. Why not use RAID instead? Well, in some limited cases we do (including dual RAIDS that act as "poor man's RAID" to each other), but the problem with RAID alone is that you can't necessarily snatch the drive out of one machine and put it in another (hard drives aren't the only things in a computer that break) and because a RAID isn't a backup, it's a highly available drive, which means when you need to recover older data after, for example, human error damages the current copy, RAID does you no good while "poor man's RAID" gives you an instantly accessible copy of last night's backup.
That, coupled with more traditional backup to tape, or, as we've been doing lately, a tar archive of the entire system's filesystem to a central RAID repository across the net, is pretty damn bullet proof, easy, and trivial to recover with.
The Future of Human Evolution: Autonomy
Unison is better for the same purpose. It also works over SSH.
Fascism trolls keeping me up every night. When I starts a preachin', he HITS ME WITH HIS REICH!