Originally Posted by BlackenedTush
I see that "it's beyond you why anyone would use a hardware RAID..."
I've read your posts and you seem to be intelligent and quite experienced; but to me your strong conclusion about RAID is a bit puzzling. You're seeing it as kinda B&W thing where a lot of other intelligent people would be seeing it as gray. The idea that it's "...beyond me"
would suggest to me that you are either 1) admitting that you are not as expert as I would have surmised from your posts, or 2) that you consider your knowledge as definitive and above the opposing and studied opinions that I have read in my own research on the subject.
It seems to me tho that on a very basic level you're promoting the principle that putting all of your eggs in one basket is not wise. I would heartily agree with you on that point.
The way I see it, when one is building a storage strategy, one must set out to to strike the most appropriate balance between performance and risk.... and that strategy is out of necessity qualified by one's budget. But not only that; but it is also qualified by one's "knowledge" of the choices. From my view, that "knowledge" factor is far more nebulous than we'd ever want to believe. I say that because the stuff that forms this knowledge is largely based on our impressions and a mixture of heresay, 2nd-hand data, and the perceptions, sales propaganda, prideful bravado, and assumptions of others (which we assign credibility to). And for the most of what we learn in this world, we're dealing with a large amount of questionable input.
I looked at the same Marvell company that you did. And my "knowledge" of them has been largely formed from my impression of their company and their products based on forums and reading other internet babble about peoples' Marvell issues. I would imagine that your own impression of them and their products is second hand as well. You see a very positive view of the quality of their products, their longevity as an on going entity, and the immutability of their programming strategy going forward. And it seems you assume that such a foundation ensures a predictable support environment for the future of a current Marvell software RAID configuration. Yet I look and see something different. I've read enough negative things about their chipset being problematic on our particular board that I was wary when configuring my own system. And indeed in my initial testing I did have my own issues with the chipset. Now I had no way of predicting or verifying what their support structure will be in the future toward addressing a RAID failure- I could only assume. I only know that I was only too happy to avoid my troublesome Marvell subsystem by using only the Intel chipset and to employ a hardware RAID controller the my research justified buying. Certainly there were many other factors that I considered. Rendering efficiency for large video composites was one of the primary factors. And after a almost a year in service, I am happy with the performance I've experienced. And so far it has been reliable. And of course I'm hopeful that my controller card is quality enough to never be a problem. But we can say that about each and every piece of HW and SW (and their synergy) we've spec'd for our builds. There are numerous potential failure modes now and in the future including new OS, drivers and incompatible hardware we bring into the mix. And users of this particular complicated MoBo have varied experiences with it. And we haven't even talked about the innumerable enemies of our data coming from software/hardware corruption, virus attack, acts of God, using large/dense consumer HDDs, etc...
In any case, I made sure that my own storage strategy included redundancy beyond the meager amount that any RAID setup offers - it includes rotating two hot-swappable BU disks, and offsite. That is a reasonable approach considering what I think I know, my budget, the relative importance of the data, etc..
I believe that the RAID approach anyone takes, whether hardware or software, is justified by the application. There are pros and cons for each and one's own application determines their particular weight. But risk always needs to be mitigated. Nothing is perfect; but we should do the best we can. In my view, reconstructing a RAID array should not be the primary strategy for dealing with a controller failure (just a last resort). So whichever is adequate or not at reconstruction is really beside the point. My personal recommendation would be to rely on one (or more) backups of our data on at least one quality disk. And that approach is valid whether prime storage is HW RAID, SW RAID or a more simple drive setup.
Thank you for your input on this board. Cheers.