These days, hardware rates are quite inexpensive. So everybody is moving towards them. However, all mainstream operating systems still support software RAIDs, maybe even for a good reason: an os-controlled software raid may be a bit better to optimize under some circumstances. Anyhow. Microsoft seems to move away from that feature set:
As you probably know, Adiscon provides premier Windows event log processing solutions. Some of our customers use the products for example to monitor if their RAIDs break. And some of them use software RAIDs. So we wrote a nice article on how to monitor RAID health using the Windows Event Log.
Since the days of Windows NT 3.1 (or was it 3.5), the Windows logged an error message if the RAID failed. Actually, I’d consider this a necessary functionality for any working RAID solution. Why? Well, if the RAID solution works, you will not notice that a disk has died. So if nobody tells you, you’ll continue to use the system as usual, not suspecting anything bad. So guess what – at some time the next disk fails and then (assuming the usual setup) you’ll be “notified” by the disk system, with those nice unnercoverable i/o errors. So without any health alerts, a RAID system is virtually useless.
We learned, that Windows Server 2008’s RAID system does no longer issue these alerts! (aka “is useless” ;)). So a long while ago, we reported this to Microsoft. The bug went through several stages of escalation. A few minutes ago, my co-worker got a call from the frontline Microsoft tech. He told him that, regrettably, Microsoft won’t fix this issue. According to his words, Micorosoft has confirmed this to be a bug, and the group responsible for ftdisk has confirmed that it should be fixed but someone more powerful up in the hierarchy has opted not to do that. Boom. The tech tried to persuade us to switch to a hardware RAID, but actually that was not the point of the support call ;)
What does that mean? To me, it looks like Microsoft is actually moving away from providing software RAID. How other can one explain that there is no interest in providing any error message at all if something goes wrong with the RAID. Given the wide availability of hardware RAIDs (which, btw, provide proper alerting), this step does not look illogical. But do they really want to leave Linux with being the only widely deployed mainstream operating system that provides software RAID? Or do they intend to keep it on the feature sheet, but provide a dysfunctional solution like in Windows Server 2008?
Let’s stay tuned and listen what the future brings…