Richard,
You don’t seem to understand my math. For the cost of a Drobo S I can buy a competing product AND a full set of 2TB drives, totally eliminating all the need to start with second hand drives and work my way up vias painfully slow (and arguably “dangerous”) array expansions. You argue that starting with scrounged drives and working your way up is “the entirety of their value proposition” yet the reason for doing that is that after spending $700 or $800 for a Drobo S we have no money left to buy any drives :-). That makes it a specious argument. What exactly do you not understand about that math? Be precise with your question.
The only way to get a guaranteed 5 years of some sort of support is to:
- Move to Europe
- Give up advance exchange of defective units the first year - return unit and wait how many weeks for a replacement?
- Give up “tech support”, which is specifically denied after 90 days, per the terms of the warranty.
We’ve talked about ways to try to end-run that, but the fact is that DRI’s policies “highly encourage” a 3 year, 3 month lifetime. That’s to get advanced replacement and uninterrupted tech support.
You question my assertion that it takes 64 hours to upgrade a Drobo that has only 3TB of data on it. That is exactly how long it took me to upgrade each of two drives last year, and how long it has taken everyone else who has reported upgrading 3TB or so of data. This is a fact embedded in the archives here. And, after going through that (130 hours of relay) just to upgrade two drives to get more capacity, I had to do it a 3rd time because Drobo did not like one of my otherwise perfectly good new drives.
It is fair to assume that going from 2TB to 3TB (then having about 5TB of stored data) will take about twice as long, or about 120 hours PER drive, or 480 hours for the array. For a 4 disk machine that is 480 hours when the machine is barely usable (throughput is necessarily terrible since DRI chose to give the relay priority) and, more importantly, the drives are banging away 24/7 for TWENTY DAYS STRAIGHT and subject to a single drive failure. The chances of a drive failure under those circumstances is likely quite high. Anyone that does a disk upgrade of a substantial array without a full backup is nuts. They are nuts not to have a backup at all times, but especially in this case.
It is my firm belief that the “expandability” of the Drobo is just not going to scale up well. It works great in the Cali Lewis demos when the Drobo has a couple of gigabytes of data. But upgrading past a set of 2TB drives is going to be extremely problematic because the Drobo is software heavy and processor light. The numbers just do not work. When and if DRI ever gives Drobo V1 and V2 the promised 2TB upgrade you will see this reflected in forum threads about never-ending relays of 5TB of data.
The idea of a Drobo V2 supporting a 16TB array is a fantasy. Going from 3 to 4TB drives will take a WEEK per drive, at least. Just reloading a failed array, assuming one was smart enough to have a backup, would take 250 hours at 10MB/s, which is my estimate on the actual real world reload time- and that could be optimistic depending on the nature/size of the files.
While I spent 190 hours to get my last disk upgrade completed and then stable and working, I came to the conclusion that the next time I do this, I’m probably just going to reformat and start over. It should go much faster (but maybe not!). And as a side benefit there has been much speculation here that that may result in much better performance, at least in some cases. My only concern is that I might never get the thing reloaded, given the dismal throughput performance of the V2 on USB. That is not much of a vote of confidence in the BeyondRaid concept but that is what the numbers tell me- I don’t have any good way to keep my Drobo array expanding.
If you question those relay times, find me a thread reporting less time for upgrading with 3TB or more of data. Try finding a thread reporting relay times for an FS with over 3TB of data- I couldn’t find one. Maybe instead of questioning my numbers you should try it yourself and report the results.
My point here is to try to reinforce that very arguably the current technology does not support drives much beyond 2TB fairly well. It is not going to scale up as easily as people think. The capacity of the drives are exceeding the ability of the hardware, especially in the case of Drobos with relatively wimpy performance. That is why I’m not excited about the promises of expandability. I can, here and now, buy a competing unit and 4 or 5x2TB drives for the price of an S. By the time I need substantially more space I will need substantially more hardware to support that capacity and I will need or severely want a new, even faster box anyway. And the S is about as good as it gets in the consumer line. I will predict that by the time you actually have 5-6TB of data on your FS and you are ready to seriously upgrade that, you won’t be happy with the results. I want to see how long it takes an FS to upgrade to 3TB drives (or 4TB drives next year or so).
Other storage devices have similar scaling problems but they are more disposable because they cost half as much. My issue here is relying on this idea of “it will pay for itself with it’s long lifetime of expansion”. That assumes too many things- that it will have a long lifetime without needing serious tech support, and that you are willing to live with the scaling problems. I’m becoming less willing to buy into that, especially since DRI is intent on limiting the supported lifetime.
Now, why do I say Drobo needs to be kept on Drobocare?
Conventional Raid is fairly stupid. As long as the hardware works the Raid will probably work. Drobo is different. It is very “smart” or, you might say, extremely complicated in terms of how it stores the data. There is a reason that the logs contain almost a megabyte of text. But we are not allowed to see those logs- by intent. Some suggest it’s to hide proprietary information, but that is just an assumption. Most of the proprietary concepts they are supposedly hiding is in the public patent document, which I’ve read.
It could also be argued that the logs are encrypted because DRI doesn’t want you to know what’s going on- they want you to call support. They want you reliant on support. They want to “force” or “highly encourage” you to buy Drobocare. The problem with that strategy (for us, the customers) is that if they do not guarantee a long life of Drobocare they are effectively placing a far more certain (and rather short) life span on the Drobo.
Last year, after replacing a couple of drives, my Drobo performance slowed to an unusable crawl of 1-2MB/s. The result of my support ticket was that a drive that works perfectly fine in any other application was “struggling” (in the words of the DRI tech) in the Drobo.
If I did not have Drobocare it would have been impossible to diagnose the problem because the logs are encrypted and in any event are not “user readable” to that extent. The “bad” drive tested perfectly using WD tools and HDTune, and performs identically to a same model drive when connected as a JBOD disk on eSata.
People that crack the logs only get temperature and a little overall performance data from it. I know; I used to be able to crack my logs but I don’t have Php so I can’t use the algorithm. And I know it’s a waste of time because the logs were designed only to be understood by the engineers, not the end user. The logs, as they are structured, even decrypted, are not much use to end users.
My experience made it very clear to me that a Drobo without a support contract is a time bomb- and effectively it was designed that way from the ground up.
You don’t understand this because you’ve been lucky and have not run across a similar problem.
Regarding the quick discontinuation of support, you said “To me it says that they want to have a fast moving product lineup… I wouldn’t mind that they actually made use of that “fast moving product lineup” potential.”
You are saying that you are happy to see your product discontinued and you don’t mind being “forced” into replacements to keep a Drobo under support, for the good of the cause, I guess? I don’t understand that. And surely, there is no law saying they must discontinue support when a product is discontinued. They can offer support as long as they choose. But apparently they choose to end support as soon as possible.
You try to compare to other products and question their lifespan. Fair enough, but at $300 that is more “disposable” than an $800 product. I don’t consider an $800 box “disposable”. We will have to agree to disagree there.
I am saying that for the ultra extreme premium price (relative to throughput) of the Drobo, and given the way they have eliminated any possibility of user diagnostics, there should be a far longer assured useful and SUPPORTED life. There is no law that says that a discontinued product cannot be supported. That is a policy choice. And for the cost of support ($69-$100/year) they can afford to continue supporting discontinued products. But they apparently choose not to.. I emphasize this because this one paragraph is the entire point of my previous response, which you completely missed.
You said “Not necessarily. Was the Drobo v2 more expensive than the Drobo v1 when it was launched? Prices for embedded ARM hardware drops just as much as desktop hardware. Given that the top of line today is 1.2 GHZ ARM processors, a DroboFS v2 could easily have a dual-core 1 GHz ARM CPU and 1GB total of RAM (512 MB of RAM just for the Linux OS). I don’t think this would increase the production costs of the FS in a significant way.”
What the incremental price of an ARM costs has nothing to do with DRI’s pricing. They are mainly selling SOFTWARE AND SUPPORT, not hardware. The Drobo V2 was a very minor update to Drobo V1, adding only Firewire. And in fact, after making great hay over the “improvement”, DRI’s advice to us Windows users was “don’t use it”. That is a fact embedded in this forum archive. It may work for Mac’s but it DOES NOT WORK for most or all windows users, and is frequently discussed here.
The Drobo S was that incremental improvement… at over 2x the cost! How much more do you think they paid for the upgraded processor in the S? That is the only difference, except the eSata interface (something you see in $30 external docks now) ajnd the 5th bay. You are right that they could speed up your FS for very little cost- but any suggestion they will is pure speculation and likely fantasy, unless accompanied by a very steep price increase (like 2x). As proof, I offer you the DroboPro FS - at $1800!
As far as “You seem to make the same mistake as others when comparing the costs of maintaining a Drobo. With a Drobo there is no need to buy fresh drives…” and so on, I won’t get into all the math beyond what I already have. I’m just suggesting that if “That is the entirety of their value proposition”, as you suggest, then maybe we actually agree that it makes no sense whatsoever to buy a Drobo UNLESS the whole purpose is to store terabytes of precious data with scrounged drives. By my way of thinking, at this point in time and with 2 years of Drobo V2 under my belt, I would rather spend that money on a cheaper box and full blown set of fresh 2 or 3TB drives for the same money :-). Your logic totally escapes me.
I’ve thought a lot about this over the past two years and looked hard at economical alternatives. Given the practical implications, and the growing disparity in pricing between Drobo and competing units, and the need to maintain SERIOUS backups- more than one copy on the residue of the upgrades - that the expansion ability has very little practical economic value. It is swamped by the premium price of the Drobo, the apparent limited supported lifespan and the need for multiple backups. Growing data arrays and protecting them is far more complex and expensive than what goes into the Drobo.
I will also point out that Synology, for example, now supports expandable mixed capacity volumes, exactly like Drobo. Linux allows for expanding volumes although they need to be done in sets. That’s here and now. I expect that to continue, with more competition for BeyondRaid as time goes on.
Here is a $149 Sans Digital USB2.0/eSata 4 bay box with hardware raid that probably performs better than an FS and maybe a little slower than an S. Here is the updated $180 version with USB 3.0. These are nice “disposable” products that will store up to 6TB of data. They may never handle 3TB drives (I don’t know) but at that price, who cares unless they imminently need to store more than 6TB of data? By the time many people need to go beyond 6TB they will need a new box anyway, no matter what they buy.
I will conclude this already too long post by saying that I think there are very valid reasons for choosing a Drobo. My hypothetical Aunt Sara and Uncle Frank (both totally un-knowledgeable computer users) are more likely to have their data survive on a Drobo than conventional Raid, which would be very inappropriate for them. If Aunt Sara can’t figure out a Drobo she is unlikely to amass more than a single WD Mybook’s worth of data and keep it intact for 5 or more years (although she may need to buy a new Drobo after 3 years or less just to keep DRI holding her hand when the lights start flashing).
What I don’t buy is your reasoning- based on the economics or maybe the mostly illusory convenience of array expansion. The math just doesn’t work, mainly because of the low price/performance but also because DRI seems Hell Bent on obsoleting our units in the shortest time possible. Everything I’ve said is based on a reasonably knowledgeable user looking for an economical way to manage a multi-disk multi-terabyte array.
