There’s safe and then there’s safe – Let’s talk RAID
// March 1st, 2010 // 1 Comment » // Rambling, Represent, tech
There’s a lot of information out there about using RAID arrays for drive redundancy or speed or both. And one thing I hear a lot is, “Yeah, I read it. But what does it mean to me?” If you’re a video professional, an animator, a photographer or any number of professions where you generate and use a lot of data, you’ve looked into storage options and, unless you’re looking for something like a ReadyNAS or a Drobo, RAID keeps coming up as ‘the way to go’.
What you probably want more than anything else from this hypothetical storage system is safety. Redundancy. Speed may be important, to varying degrees, but not losing your data along the way is paramount. So what do you need? And what the heck are all those levels and types and why do you need to care?
If you’re someone who just needs to get stuff done and doesn’t want to spend a lot of time sorting all that out, I’m here to help. Assuming you trust me… *grin* So here’s the short course. Skip straight to the graphic if you want. But hopefully you’ll at least skim the following.
CONTROLLER TYPE
Hardware RAID on a controller card is better than software/firmware RAID on a controller card. Software RAID on a dedicated controller is [usually] better than software RAID done by the operating system. The latter depends on how well the OS knows its local hardware. For example, software RAID on a Mac Pro is likely better than most software RAID cards. But that’s the exception, not the rule. A dedicated controller doing hardware RAID is always better. And always more expensive.
LEVELS
This is the bit that’s most confusing to many. What RAID levels signify is the method used to make the data on multiple disks redundant. And, in general, the higher the number the more ’safe’ your data is and the better the methodology. That’s not 100% true, actually, I’ll take RAID1 over RAID3 any day, but for our purposes here, let’s forget about RAID levels 2 through 4. And 0 is right out… mostly. Sort of.
Hang on. We’ll sort all of that out in a sec. Trust me.
RAID 1 is a method for mirroring drives. The most common example is two drives, one mirroring (an exact duplicate) of the other. If one drive fails, the other can soldier on. Yield Example: Two 1TB disks yield 1TB of space.
RAID 5 is three or more drives with the data spread out in such a way that if any one drive fails, the array can continue on until the drive is replaced. Several popular proprietary methods (Netgear’s X-RAID, Drobo’s BeyondRAID, etc.) are similar to RAID 5 but may offer some added benefits, depending on your need. Yield Example: Three 1TB discs yields 2TB of space.
RAID 6 is four or more drives which works similarly to RAID 5, but can tolerate up to two drive failures simultaneously. Yield Example: Four 1TB drives yields 2TB of space. Five 1TB drives yields 3TB of space.
RAID 1+0 (or RAID 10) combines the best of RAID 1 and RAID 0 (a level which provides speed, but no safety whatsoever) along with parity (striping) like RAID 5 or 6. Multiple mirrored sets can be striped together for very high redundancy without sacrificing speed. Requires an even number of drives (4 or more). Yield Example: Four 1TB drives yields 2TB of space (but spectacularly redundant data!).
BACKUP
And all of these types of RAID control are better when paired with some sort of backup. That’s right… as the saying goes, RAID is not backup. If you build a RAID array, back it up. If you don’t, while your chances of data loss might be decreased, they’re not alleviated. If you have 1 copy of something, it’s in peril. 2 copies, it’s backed up. 3 copies, you’re serious about your data.
ONE MORE THING
Port multiplication. If you’re looking at SATA enclosures, you’ve likely seen this mentioned (or heard someone try to scare you off from it). It’s always better to have every SATA drive connected to it’s own line (channel), but sometimes… sometimes that’s just not in the budget. Port multiplying enclosures (and controllers that support PM) are a necessary evil when you can’t buy your way to where you’d like to be. Just to where you hope you’ll get by. I freely admit there’s a PM box hanging off my Mac Pro. It’s used to back up my internal RAID. I see that as a justifiable risk-to-expense decision.
The primary cons for port multiplication are:
1) The maximum speed attainable will be that of the channel, not the box. So don’t expect to get more than 300 MB/s (and plan for less, PM has additional overhead which translates into speed penalties) from any PM box. Think of it a s splitter, with all those drives sharing a pipe meant for a single drive.
2) Some controllers do PM better than others. If you need to scrimp on the enclosure, try not to scrimp on the card. In port multiplication, only one drive can be talk/be talked to at a time and there’s a lot of talking to be done to handle that big flow coming into that small pipe.
Like I said. If you gotta do it, do it. But go into it knowing you’re on the razor’s edge. So plan accordingly (Backups! Unless, of course, this IS the backup device for a more robust array) and don’t set your expectations levels the same as you would for a multi-channel enclosure.
LET’S GET TO IT
So what do you need? What can you afford? What are you willing to give up or not give up? Well, here’s that graphic I promised. Hope it helps.
That middle line is what I like to think of as the ‘make or break’ line. Once you’ve passed that line, then you can start sleeping better at night. But, really, it all depends on your needs and means. Do the best you can. I hope this post helps you sort out what that is.
Legend: H/W = Hardware RAID – S/W = Software RAID – OS S/W = Operating System S/W RAID – BU = Backup
Maggie 3/1/10
