Reverse Engineering the Windows Home Server Backup Database

I think it was when ColinWH replied to my request for help I had decided that I'm going to attempt a crude reverse engineering attempt of the Windows Home Server backup database. I didn't want to know how the whole thing worked, just enough to get that darned Control.4096.dat reconstructed.

Unfortunately, Control.4096.dat was not an XML file at all. It was not a very big file, only 4KB, and it did contain something in the middle that looked like XML. I'll save you the trouble of reading through the excruciating description of how I deciphered the binary format that the Home Server uses. Surprisingly, it didn't take long. I had a good part of the basic infrastructure figured out on the next day after I posted the initial request for help. It was somewhat crude and incomplete but it gave me enough information to begin to think about reconstructing the missing file.

In fact, on the next day, I was able to load my so called lost-cause backup database and pull files off. It worked so well that I pulled off 2 entire partitions and since have imaged them to DVDs.

In the process of reverse engineering the DB format I decided to document everything, because it seemed to me that this knowledge can be valuable to other's who might get into similar circumstances. Also, I had the thought of developers making applications that work directly with the WHS Backup DB using my unofficial spec. as the starting point.

One of the reasons that this blog exists is so that I can post my original reverse engineered spec. to a more permanent place. It was originally available in this forum post.

It's now been just over 2 weeks since my catastrophe. I've figured out most of the database format. I've made my first tool based on my spec. (this is coming up shortly). So what follows will be the original spec. edited and updated for correctness.