-
RAID 5: False Security?
I didn’t want to post this in a manufacturer-specific forum, because this is an issue for just about every serious editor: What’s so great about RAID5?
I had a very lucky escape Friday night: I finished a project just minutes BEFORE my RAID5 broke down. (In my case, this was not due to a failed drive. I don’t know my exact problem; unfortunately there is no formal tech support on the weekend.)
But I learned from esteemed COWherd David Weiss that if my problem had been a failed drive, it would have taken my RAID many hours (at 44 minutes/100GB = 7.3 hours/1TB?) to “rebuild” itself. That equals a blown deadline.
If I really want to be prepared for a disaster during a deadline, I should know the following:
– What does “rebuilding time” mean anyway?
– Is rebuilding necessary, or could I finish a show (without RAID5 protection) without rebuilding after having lost a drive?
– Having “rebuilt,” would you have RAID5 protection on the remaining drives?
– If you pop in a new drive to replace the bad one, does that add to the time cost?
– If I have a spare drive on hand, is there any way to shorten that rebuilding time?The big question: for deadline-oriented work with zero tolerance for failure, is RAID5 enough? It sorta seems like a gun with one bullet.
Bob C
-
8 Replies
-
Bob,
I think you’re misunderstanding the “rebuilding” process. When a SINGLE drive goes bad in a raid 5 config, you may continue to work accessing that raid set. You may experience slightly lower performance, but you still can use the raid… that’s the whole point! The “rebuild” process is typically a ‘background” activity, so while it may take hours to complete, it really doesn’t impact workflow that significantly.
Our entire SAN is based on Raid 5 level security. We’ve been up and running for over five years and haven’t missed a broadcast deadline yet. During that time we’ve probably replaced and “rebuilt” over a dozen individual drives. At around the three year mark, it seemed that drives were dropping like flies, and we were popping in new drives just about every week. Since then, it’s stabilized and we haven’t had a drive go bad in awhile. But, make no mistake, EVERY drive will die if you run it long enough. That’s why they invented the RAID schemes.
Having said all of the above, when we’re facing a critical deadline, I make sure that essential media lives somewhere OTHER than our primary SAN, like on a the internal drives of one of our computers. You’re absolutely correct that “time to restore” media if the whole SAN goes done should be considered. I always have the original media on disk or tape, but by the time I recreated what I need, my deadline would have passed. So, if you’re down to the wire, make sure you have a plan “B”.
Mark
-
I know in the case of our older Apple Xserve Raids, rebuilding takes place in the background with a minor performance hit. We have an Xsan with 6 Xserve Raids, each raid array is seven drives, half of an Xserve raid. When I pop in a new drive, there’s more activity on that particular array, but none of the up to 12 users who are connected even notices a performance hit.
Raid 5 always seems to be the best combination of security and throughput. Compared to the risks of no redundancy, I’ll take raid 5 any day. Data is always at risk in this business, the best we can do is minimize that risk.
Cheers
John Christie
-
So Bob, what’s your redundancy plan if a motherboard or graphics card fails? Nothing is impervious to failure anywhere in the system. RAID 5 is a good compromise between a single drive failing and speed. There are other RAID configurations that will give slightly more security but may slow your drives as well.
As with anything, buy good quality, have redundancy strategies and if you have the budget for a spare MacPro in case one dies, then you might afford a second RAID5 setup that mirrors your main drive system and constantly backs up the work drive. i have three different RAID devices here and often back up current media and project files to a second RAID.
-
Mark, John, Michael: Thank you. Very reassuring, and great points all.
I infer that “rebuilding” means bringing RAID 5 status back. I assume that while I’m editing on the RAID which is being rebuilt, I’m in RAID 0 with no data security? During the rebuilding process, do I have read/write access to the RAID or just read?
I have an LTO (tape cartridge) backup device on the network. My initial thought is to add the RAID to the list of hard drives which the LTO backs up every night. (Already have the backup computer — a laptop.)
Thanks again.
Bob C
-
Yes, whilst rebuilding, a second drive failure is usually catastrophic. If the new drive you inserted fails, then only the rebuild fails. Tape back up is still a good idea although large projects with tons of media can make it slow and expensive.
-
How does the laptop access the RAID 5 enclosure?
Tape backup is super important but it is no means a zero downtime solution. The only true solution would be a totally redundant setup, Another Raid and system to go to at a moments notice.
With redundancy you are reducing your chances for disaster not eliminating it.
Glad it happened at a “good” time.
-
[Michael Sacci] “How does the laptop access the RAID 5 enclosure?”
CalDigit makes an Express Card. I haven’t bought it yet, and I’m thinking I may rather put the money toward a second MacPro anyway. I bought the laptop for location editing, using firewire drives for media.
Yes: no way to prevent any possible calamity — just trying to minimize it. Despite my questions, I’ve actually become rather philosophical about such things… as long as I know I’ve done everything reasonable to prevent disaster in the first place.
Thanks!
Bob C
-
[Bob Cole] “as long as I know I’ve done everything reasonable to prevent disaster in the first place. “ That is all we can do.
Reply to this Discussion! Login or Sign Up