MooseSentryTM and MicrosoftTM System Center Data Protection Manager
The MooseSentry appliance is built around Microsoft Windows® Server 2008 R2 and System Center
Data Protection Manager 2010 (SCDPM). It is therefore a Windows Server, and, ideally, should be joined to the Active
Directory Domain that contains the servers and workstations to be protected. The next step is to
install agent software into the servers and/or workstations to be protected. These agents
are licensed separately, and are referred to as "Data Protection Management Licenses" (DPMLs). As
shown in the drawing at right, there are three kinds of DPMLs:
- Client DPML - used to backup Windows XP, Vista, or Windows 7 workstations.
- Standard DPML - used to backup file shares and directories, and Active Directory System State on Windows 2003 and later systems.
- Enterprise DPML - used to back up Microsoft applications, i.e., SQL, Exchange, SharePoint, and Hyper-V, plus the ability to capture images of entire systems for "bare metal" recovery.
"Express Full" Copies
Create Backup Sets
Periodically, (typically once per day), these block changes will be synchronized to the DPM server – this is known as an Express Full operation. The DPM agent on the SQL Server will leverage Volume Shadow Services ("VSS") to create a “snapshot” of the disk volume prior to synchronization, so that file I/O may continue while synchronization takes place. Once synchronization is complete, the VSS snapshot is released, and the bit mask reset to show only those blocks that were changed since the snapshot was taken. This synchronization may happen as frequently as every 30 minutes, must happen at least weekly, and typically happens nightly. The important thing to remember, though, is that only the blocks that have been changed are sent to the DPM server.
The DPM server itself also leverages VSS snapshots to track the differences between the original baseline copy and each Express Full copy. So as of the first Express Full operation, there are two possible restore points: the newly synchronized copy, and the original baseline copy, which can be reconstructed from the snapshot that was retained. This process is repeated with every Express Full operation, and (storage space permitting) the DPM server can retain up to 512 of these snapshots for each backup set!
Backing Up Transactional Applications Using Transaction Logs
But it gets even
better. Let's continue with the SQL Server example. SQL is a transactional application -
by which we mean it maintains transaction logs that track the changes that are made to the database. On
Day "0," after we have made our initial baseline copy, the agent will send just the changes to
the transaction logs every 15 minutes. (Click on graphics to view full-size.)
That means we can roll back to any 15-minute point during that day by restoring the baseline copy,
and then re-playing the transaction logs up to the desired restore point.
Now, assume that at, say, midnight, we perform another Express Full operation as we move into Day "1." As stated earlier, the DPM server takes a VSS snapshot of the baseline copy before it is updated by the Express Full operation. We also retain the incremental transaction log changes that occurred during Day 0. As we move through Day 1, we continue to send the transaction log changes every 15 minutes. As before, we can roll back to any 15-minute point during Day 1. However, since we can use the VSS snapshot to reconstruct the original baseline copy, we can also roll back to any 15-minute point during Day 0.
Midnight rolls around, we perform another Express Full operation, and move into Day 2. Once again, the DPM server takes a VSS snapshot of the Day 1 image before it's updated, and the process continues. As stated, given enough capacity in its storage pool, the DPM server can retain up to 512 snapshots with their transaction log updates. If you do the math, you'll see that means that if we continue to take one Express Full per day, we could roll our database back to any 15-minute point in time in the last 16 months!
Microsoft Exchange is also a transactional application, so Exchange Server backups are handled in exactly the same way.
Express Fulls and Recovery Points
So far, we have assumed that we're taking one Express Full each night at midnight. It may have
occurred to you, however, that if we want to roll back to 11:45 pm, it might take a long time to play back a
full day's transaction logs for a heavily used database. The solution would be to take Express Full copies
more frequently. For example, if we take an Express Full every 4 hours, as shown in the graphic at
right, we would have far fewer transactions to play back. If we wanted to roll back to
6:45 pm, we could just restore the Express Full from 4:00 pm, and then play back only the transactions that
occurred between 4:00 and 6:45.
Files, Directories, and Non-Transactional Applications
File and directory backups are handled a little differently. As with other
kinds of data, you start by creating a replica on the DPM server of the
protected data. You then decide how often you want to synchronize the
replica with changes that have been made in the protected data - which, again,
can be as frequently as every 15 minutes. We must then decide how often we want
to establish additional recovery points (which are essentially snapshots of the
replica), and how many of those additional recovery points we want to retain. In
the example illustrated at left, a recovery point is established at 6:00 am, before the work day
begins. Additional recovery points are established every two hours until 6:00 pm. Then a final recovery
point is established at midnight to capture changes that may have taken place during the evening.
Meanwhile, changes are being synchronized every 15 minutes. A file
can be restored using the most recently synchronized copy, or it may be restored to any prior recovery
point. (If we didn't expect the data to change very often, we might choose
to perform a synchronization just prior to establishing each recovery point, in which case our only restoration options
would be the recovery points.)
For non-transactional applications, such as SharePoint, we would again decide how frequently we want to establish recovery points, and schedule our Express Full operations accordingly, as illustrated at right.
It's also important to understand that you don't have to know anything about Express Full operations, transaction log changes, or recovery points in order to restore your data. You simply bring up a restore "wizard" that will show you what recovery points are available to you. You choose the point in time to which you want to roll back, and Data Protection Manager does the rest for you.
For non-transactional applications, such as SharePoint, we would again decide how frequently we want to establish recovery points, and schedule our Express Full operations accordingly, as illustrated at right.
It's also important to understand that you don't have to know anything about Express Full operations, transaction log changes, or recovery points in order to restore your data. You simply bring up a restore "wizard" that will show you what recovery points are available to you. You choose the point in time to which you want to roll back, and Data Protection Manager does the rest for you.
Long-Term Archives and Off-Site Replication
FireStreamer will emulate a tape library. For example, you can use it to make one of the MooseSentry's 1.5 Tb hot-plug hard disk drives look (to the DPM software) like a tape library consisting of multiple 20 Gb "virtual tapes." In fact, FireStreamer can do this with any disk volume that the appliance can access: a USB- or network-attached external hard drive, a volume mapped to a file server, even a SAN volume. So when you take a closer look at the drawing at left, remember that the "tape based backup" can be either physical tapes or virtual tapes. (We much prefer the latter.) With proper preparation, you can even swap out hot plug drives or external drives to maintain the same kind of "tape rotation" that you would perform with physical tapes. You might, for example, want to maintain a monthly, quarterly, or yearly archive in addition to the restore points you are maintaining on the primary hard disk storage pool.
Finally, the MooseSentry can solve that nagging problem of how to get a copy of your most critical data out of the building. Simply set up a second MooseSentry in an off-site location, and connect it via a VPN tunnel to your primary network. The primary appliance will then stream data to the secondary appliance. Because, once again, we are only sending changes, bandwidth usage is very efficient. Moreover, you can "throttle" the bandwidth: for example, you might limit bandwidth consumption to a relatively small percentage of the available bandwidth during business hours, and remove the bandwidth limits after business hours.
For more information on the MooseSentry appliance, see our Pricing and Options page. For more information on System Center Data Protection Manager, see http://www.microsoft.com/dpm.
NOTE: The graphics on this page are from a TechNet Edge video by Jason Buffington of Microsoft Corporation.
