There are clear signs that the Home Server team “gets it” when it comes to reliability – I really like the idea of a “divorce-able offense”. That’s the right way to look at protecting this data; now, let’s start talking about how it’s going to happen.

We all (us propeller-heads) understand RAID. I don’t yet understand how Home Server is going to achieve reliability, nor what the yield for new space will be (how much of each new drive will be available for storage), nor what the limitations of this scheme will be.

For example, it stands to reason that if I have a 500 GB disk and a 100 GB disk, I’ll have something less than 600 GB available for storage, but how much less. If the 100 GB disk dies, it’s reasonable to believe that I won’t lose any data, but if the 500 GB disk dies, I will. What, then, are the guidelines for configuring disks such that I have a reliable platform? We understand all these things for RAID, but we don’t yet understand them for Home Server. We need to.

One of the reasons I pointed out Infrant is that they do a very credible job of explaining what X-RAID is, how it differs from RAID, what the benefits are, and how it works. I get X-RAID. I like it. I’d store my photos there because I know what makes it tick.

I’m not suggesting that Microsoft avoid innovating, I’m suggesting that if they’re going to ask me to store my wedding pictures on this thing, I’d like to understand why I should believe that they’re safe.

Cheers, Steve

From the Paul Thurrott article:

“RAID is an insect spray,” he cracked. “With RAID, you must understand the technology, add disks in sets, and its hard to remove drives.” With WHS, storage is hot-swappable. You can plug in an 80 GB hard drive, for example, and configure it quickly with the WHS Add Drive wizard. When you want to remove it and replace it with a 500 GB drive, there’s a simple wizard for that as well.

I don’t know a ton about RAID, but it seems pretty clear to me that modifying existing arrays (either adding or removing drives) is a problem with RAID–pretty much the only scenario that doesn’t require you to rebuild the array is replacing a failed drive with an identical one (at least with RAID 0 or 5), unless you are just talking about adding RAID 1 pairs. From this description, WHS allows you to use whatever heterogeneous collection of drives you have, and add and remove arbitrary drives without much trouble.

Parts of what WHS offers don’t seem new (I’ve used LVM on Fedora, which worked well although it was far from friendly and doesn’t provide redundancy by itself), but for the target market, it seems pretty accurate to me that this has it all over RAID–if it works.