What's The Catch?

Well, it's not really a catch. You do benefit from the performance benefits of a RAID 0 array as well as the data redundancy of RAID RAID 1. But you need to understand what the Matrix RAID really does, as well as its limitations.

When I first heard of the Matrix RAID, I had the impression that it allows me to create a RAID 0+1 array using only two hard disks. This didn't quite make sense as the idea behind combining RAID 1 into the RAID 0 array was to create a redundant mirror on two separate physical hard disks to increase the MTBF (mean time before failure) of the array.

I was left wondering what technological magic Intel had conjured up that enabled them to do the same with just two hard disks, until I read their documents on this technology.

The reason why I initially misunderstood the concept was because when I heard about Matrix RAID, I read that it provides the capabilities of both RAID 0 and RAID 1. I assumed it would work as a combination array, whereas Matrix RAID actually has separate RAID 0 and RAID 1 volumes!

So, the technology really works as Intel advertises - RAID 0 and RAID 1. What you won't get with the Matrix RAID is the capability to combine both RAID 0 and RAID 1 for a RAID 1+0 or 0+1 array. This is simply because it does not make sense to have the RAID 0 array mirrored on the same hard disk.

Remember, such a RAID 0+1 array will need to use the RAID 0 and RAID 1 volumes (both on each hard disk) created by the Matrix RAID. So, if one hard disk fails, both RAID volumes will be affected and you won't be able to recover all the data. That's why Intel prevents you from doing so with Matrix RAID.

What About Performance?

It is important to remember that, the hard disk's read-write heads can only be at one particular location at any one time. Now, let's recap what we know of the Matrix RAID array setup. We know that :

1. Each hard disk will be shared by both RAID volumes, and
2. A single hard disk can read from or write to only one location at any one time.

So what does all this mean?

Since the RAID 0 and RAID 1 arrays are sharing both hard disks, there will be performance hits when both volumes are accessed at the same time. This is very different from normal RAID arrays, where each RAID array has their dedicated set of hard disks, and can therefore take full advantage of parallel hard disk accesses.

You may be wondering if this statement is even valid, as we have seen some pretty good benchmark scores on some other hardware sites which have reviewed the Matrix RAID technology. Well, one must remember that most benchmarks are synthetic and therefore, only tests a single volume at a time. In such cases, the performance impact of the RAID volumes sharing the same hard disks would be very minimal.

Let's cook up a quick scenario of how Matrix RAID can impact performance. Let's say we create the first volume in RAID 0 and store the operating system and program files in it. Then we create a secondary RAID 1 volume to hold all our important documents, photos, home videos etc.

Now, imagine what happens when you decide to edit some videos stored in the RAID 1 volume (for better data redundancy). Meanwhile, the operating system, swapfile, video editing software and scratch disk are all residing in the RAID 0 volume for better performance. Editing those videos would result in heavy disk accesses from both volumes as both the paging file or scratch disk as well as the video files will be read from and written to quite extensively.

Each physical hard disk can only either read from or write to one location at any one time. Therefore, when the hard disk is asked to read the video file, it needs to move its head to the physical location of the file and read from it. But before it finishes reading the file, the operating system may request it to write some data into the paging file. The hard disk then has to switch back and forth between the paging file and the video file until one or both operations are complete.

As you can guess, this will cause the hard disk to "trash" as it attempts to perform those operations simultaneously on different physical locations. This reduces its performance as frequent seeking reduces the hard disk's transfer rate.