Drobo

Two common misconceptions about Drobo

It seems lately there’s been an increase in posts by people disappointed because of two general misconceptions:

  1. Drobo will definitely increase performance compared to a single drive on the same interface (“It’s RAID so it should be faster”)
    and
  2. Drobo will “save” my data from catastrophe (“If Drobo keeps my data safe, I don’t need backup”)

So I think it’s worth addressing these.

Misconception #1: Drobo will increase performance compared to a single drive on the same interface (“It’s RAID so it should be faster”)

Well, that all depends on what you’re comparing its performance to.

While increased performance is a feature of striped RAID strategies, when fault-tolerance is introduced, there is an initial performance penalty incurred to get that extra reliability.

When you write data on a non-fault-tolerant drive, you just write the data. However, when you write data on a fault-tolerant system, you must write the data, and write the extra data needed for the fault-tolerance.

This much like having two people sign your holiday cards instead of one. It adds a small bit of extra work.

Another give-and-take in performance is the fact that Drobo lets you mix drives of different makes and sizes. This is a great feature for upgrading and maximizing usable storage, but it hurts performance because different drives have different performance specs.

Back to the holiday card-signing example, this is like having multiple signers, and the different signers take a different amount of time to write their signatures. If you think of the data transfer as the speed of the “flow” of cards being signed, it can be bottlenecked by a slow signer, just like Drobo can be bottlenecked by a slow drive.

Also, because Drobo is an embedded system, its processing power is limited by the speed of its internal components. The newer Drobo S has a faster CPU, so transfers on the same interfaces compared to Drobo (non-S) are improved.

Right about now you’re probably thinking “Well, don’t enterprises use RAID for high-speed storage like with video editing and database work?” and the answer is Yes.

RAID striping does help, but it requires a large number of drives. Given the same drive parameters, an 8-drive striped array will generally outperform a 4-drive striped array, and so on.

Storage mechanisms optimized for pure speed generally sacrifice fault-tolerance, decrease storage density (less usable storage per-TB of required storage), or they involve a much larger number of striped disks in the array which then sacrifices some degree of fault tolerance, as the more drives you have, the greater the statistical likelihood of having more drives fail simultaneously.

I understand that Drobo isn’t a speed demon - if I need high-speed storage for video editing or other work, I’ll use internal drives and stripe them if necessary.

Granted, Drobo isn’t slow, especially compared to drives of a few years back. It’ll likely be faster than older drives/technology, but just keep in mind that Drobo’s secondary focus is on speed, but its primary focus is on fault-tolerance. Which leads to…

Misconception #2: Drobo will “save” my data from catastrophe (“If Drobo keeps my data safe, I don’t need backup”)

Drobo will keep your data safe from a single drive failure in standard single-disk redundancy (SDR) mode, or from two simultaneous drive failures in dual-disk redundancy (DDR) mode (Drobo-S, DroboPro, DroboElite). Unfortunately, disk failure is only one kind of catastrophe.

If lightning strikes your power line or Drobo, Drobo and its data will very likely be lost.
If the building your Drobo is in goes up in flames, Drobo and its data will very likely be lost.
If someone steals your Drobo, Drobo and its data will very likely be lost.
…and so on.

Beyond just physical catastrophes, there are a number of data catastrophes that can occur as well, that are independent of Drobo and therefore cannot be prevented by Drobo.

If your system gets a virus that deletes or corrupts files, data on your Drobo will very likely be lost.
If your operating system upgrade goes awry and it deletes the attached disks, data on your Drobo will very likely be lost.
If your cousin who’s visiting from out-of-state plugs in their USB stick and mistakenly formats the Drobo volume instead of the stick, data on your Drobo will very likely be lost.
If your nephew overwrites the business proposal you’ve been working on for two months with his online chat session log, data on your Drobo will very likely be lost.
If your dog trips over the power or data cable to your Drobo while it’s in the middle of writing something, data on your Drobo will very likely be lost.
…and so on.

How much data loss can occur, really depends on the nature of the event.

I don’t mean to panic people here, but it’s human nature to get lazy in our thoughts and actions. Drobo’s fault-tolerance is not a substitute for backup! Drobo can be part of your backup strategy, but fault-tolerance is not backup.

Another way to think of this is…
Fault-tolerance is “being careful” - Drobo is part of you being careful.
Backup, on the other hand, is insurance - insurance for those times when being careful is simply not enough.

You wouldn’t cancel your automobile insurance because you think you’re a really good driver.
You wouldn’t cancel your health insurance because you think you exercise and eat well either.

You can be the safest driver in the world, but it’s not going to stop the car that just sped through the red light from hitting you.
You could be the healthiest person of all your friends, and still drop dead at 30.
You know in the real world, there are things that happen out of your control.

And hopefully by now, you know that you need to have a backup of any of your truly-important “can’t lose” or “hate to lose” data.

This is in no way a bash or criticism of Drobo. I love my Drobo. In fact, I just got another one. Guess what for? BACKUP!!

I’m just trying to make sure people have realistic expectations and aren’t buying a Toyota and expecting it to have BMW features.

Granted, you can’t get into this forum if you’re not already a customer, but maybe the marketing folks at DRI will find a clever way to stifle some of the misconceptions in the outside world (like on the website).

Brandon
EDIT 2009/12/29: Clarified point #1 so it doesn’t sound so much that Drobo is slow. Thanks for the feedback Switcher!

I agree with point #2.

Point #1 is murky. You assert an absolute, and in this you are wrong. It depends on what model Drobo you have and what drive you compare it to. My DroboPro is boat loads faster than the internal drives of any of my Macs. My gen2 Drobo was about the same., and my original USB only Drobo was slower.

I agree with your point of #1 being murky. I’m not trying to say it’s slow per-se, but people should have realistic expectations.

I think the more tech-savvy folks are the ones walking away less-impressed, even though they’re the ones that should have better understanding of the benefits and trade-offs. :wink:

The most important aspect of RAID drive performance is dictated by the interface it is used to connect to the host computer, not the striping drive array. The reason why your Mac drives are slower is because Apple gives you what you need to start with. Apple does not arbitrarily provide you with a WD Raptor or a WD Velociraptor drive as a standard on all their Macs either. So yes, it is an absolute in terms of drive performance that sometimes, a single drive can be faster than a RAID array. The downside to a single fast drive is in the cost vs capacity perspective. It does not give you the best bang for the buck if you are a photographer or a videographer, for which those clients I serve. Which is why RAID (Redundant Array of Inexpensive Disks) were conceived back in the late 80s and that was to address performance and cost issues compared to a single expensive high performance single drive.

Sadly though, the marketing aspect of most RAID providers these days emphasized on the fault tolerance aspect of RAID and confused it with drive reliability by showing the audience the “WOW” effect of RAID 1 to 1+0, which is pulling one drive out while the drive array keeps chugging along. This is demonstrated countless of times in many trade shows as well as in Youtube. It’s the single biggest selling point of RAID. Can you do that with a single drive Mac?!? The answer is a definite no. Which is why, most people confused this with the need for no backup, because if RAID eliminates drive failure by providing redundancy and redundancy itself sounds more like a backup, why should the customer provide another redundancy solution?

Besides, there are other more cost effective RAID 5 solution that provides more cost effective but potent performance compared to the Drobo S or Elite that other Mac users use and the fact that, unlike other RAID solutions, the Drobo introduced another fault of its own, which is the drives do not have to be equal in capacity, performance and quality either. This introduces another possible fault failure with mixed drive array performance since the weakest link in the Drobo performance and reliability itself is with the single drive used. The well known but dated Seagate firmware problem had caused many problems with people loosing data. If you bought say 2 Seagates + 2 WD green and the 2 Seagates somehow died at the same time, the Drobo provides no protection nor with a RAID 5. A typical consumer does not know this.

Having said that, a striping RAID array performance when interfaced with the fastest host computer providing enough buffering and an efficient operating system is more likely to be faster than the fastest single drive. However for the majority of the users I know and serve, I would advise a single drive solution like a WD Velociprator as a Photoshop or Final Cut scratch drive and then move the finished product into a RAID 5 double array or a Drobo double array.

@DumbTechDude, so you agree with me that DroboPro is faster than most internal drives. Nice essay.

really?

i would say most internal drives on the market hover around/above 100mb/sec (at least on the outside tracks)

drobopro’s advertised top speed is 90mb/sec from memory - although under ideal conditions this would be consistent across the whole available space

I wouldn’t call being able to mix drives of different sizes a fault, per-se, but that feature does come with a price.

yes, i would say “characteristic” at best

plus its easily “disabled” - just use identical ones

Generally yes if you are working with big files – Photoshop multiple layers and HD content both come to mind. Another benefit to having a RAID 5 or Drobo is the fault tolerant aspect. There is no workflow stoppage if one drive fails or if you have DDR, a dual drive failure. There is with an internal single drive. This is one of the most neglected aspect of any digital workflow. It is a fallacy to assume that you won’t have a dead drive – you will but it is just a matter of when? When that time comes, are you going to tell your client that you can’t do your work because your drive fails? This will not project a very confident image of your company or reputation moving forward!

To answer Docchris; you have pretty much answered yourself. Anyone can play with HDTUNE and show impressive numbers. Real world performance notwithstanding. When you are dealing with terabyte of data, consistency of average data transfer is more important. Granted that due to the ovehead and redundancy(parity) added to the RAID 3 and higher array, it would seem that a single drive is faster. In some situations, I would advise people to opt for a single drive scratch drive setup and backing it up to a standby RAID array and then backing up again to another RAID array somewhere else at night. Some people don’t feel too easy with these arrangements – having everything in one basket. People need to balance between speed, reasonable data security through redundacy as well as minimal downtime. A good RAID array provides them all.

To answer Bhiga;

Consider this. Would you consider having just 2 snow tires mounted either on the front or the rear of the vehicle? Some states mandate that you must have all 4 snow tires, but that will not stop people by having just 2. It’s cheaper. Same with the Drobo… It’s cheaper to mix drives because in the technology world, time is on your side. Stuff gets cheaper when it gets old, so you can slowly upgrade your array. Ofcourse in this case, people who do this usually don’t put a lot of value on their data, until when they loose it. When they do, they get mad, real mad! So what, they made the choice! With a standard RAID array, you must make the choice right away by having 2 and more identical drives and they would choose the most reliable drives, willing to pay the cost to have that security. These people value their data.
I’m not saying that Drobo’s approach isn’t good – they aren’t alone in this field. I’m just saying that if someone thinks that by having 2 snow tires will give you the same performance as having 4, that someone is dreaming.

as for drive failures…

when i did my masters, there were 12 students… bought 12 brand new pcs. in the course of the 1 year course, 3 had total hard disk failures, outright dead. that was 25% in one year (luckily everyone used their network space for all the work!)

Agreed. I think some people are buying 2 snow tires and expecting the same performance as 4 (or 12)…

its interesting…i agree that there are some misconceptions.

i bought my drobo after seeing reviews from Rauol Pop, plus questions/answers and good 2nd hand experience knowledge and advice from Docchris :slight_smile:

but the key factor was that the ‘perception’ i got about drobo, was that it would:
a) help protect my data more than just my c: drive
b) allowing me flexibility to swap out drives per slot and keep expanding the free space with bigger drives as they come out.
c) help me merge all the data from my various smaller ide hard-drives and zips/cd’s etc into a single area (whilst making it all available to me and meeting (a)
&
D) so important i almost forgot… i didnt know much about raid-related stuff :slight_smile: so it had to do it all for me - that’s so obvious for most users i almost forgot to mention it.

(i still bought a 2nd drobo (v2) which i use as a physical local backup, and offsite version/offsite data storage is planned at some stage) but those were the main reasons for me…

(i agree about misconceptions, especially from reading some peoples posts here … its just that i never personally had a notion that the drobo would be a speed demon, just these factors:

ease of use, more security, better feeling about being able to not waste smaller, less used hard drives that you’ve paid for… just bung them in the drobo and grow it with time :slight_smile: