Drobo

Relayout Performance?

I’m curious to hear about any relayouts done with a significant amount of data; the disk config, approximate amount of data in the array, and about how long it took.

I have issues with my DroboV2 relayout times (it cannot scale up forever and is already past the point where some things make sense) so have an interest in the S, which would be my next logical upgrade.

My Drobo S with five 2Tb drives and dual disk redundancy is 80% full, and changing a drive requires a relayout that runs about a week. It seems to be roughly linear in terms of percent space used, so, very roughly, a day’s worth of relayout per 10% space used.

I know that this is Drobo S forum. Just posting this for comparison. I have Drobo v2. The relayout times for me seem to correlate to about 24 hours per TB.
I have 4 2TB disk and only single redundancy, of course. I would guess that the dual-redundsncy is what is making Mr. Protocol’s Drobo S take so long.
How is the Drobo S without dual redundancy?

A week of relayout seems preposterous for Drobo-S which was supposed to be “faster” than Drobo-V2 (and noisier because of that). This makes a full 5 disks (sequential) upgrade more than a month… :frowning: :frowning:
I confirm yoyoma numbers for Drobo-V2 : about 3 days for an 80% used 2TB drive relayout.

I had 2.6TB of data on my V2 when I upgraded it with 2x2TB drives. Each relay took about 63 hours. Consistent with yoyoma’s numbers of 24 hours per TB of data. I had 2x1.5TB + 2x1TB GP drives, replacing a 1TB drive and a 1.5TB drive with 2TB GP drives. (One of the existing 1.5TB drives had issues and needed replacement)

geeji - could you be more specific? You are implying you had 4.8TB of data on the array or maybe I misunderstand.

I agree that a 1 week relayout per drive is not viable. At a sustained transfer rate of 10MB/s I could, in principle, copy 3TB of data in 83 hours. That is a very reasonable number on my V2 for copying large files. I don’t know if it would hold across my mix, which includes at least a half million small files too. At 40MB/s, which should be reasonable on an S with large files, 3TB should copy in about 20 hours, but again there are issues with small files so mileage will vary. But no so much that it would take a month to reload an array. It calls into question the utility of upgrading the array.

It would be interesting to see if the relayout times are different with single drive redundancy. Hopefully we will get some comparable numbers here from other S owners.

No, I only had initially 4x1TB drives, filled up to 85%, thus amounting to about 2.4TB of loaded data [Drobo-V2].
If you accept the 1TB/day figure, I should have been able to upgrade sequentially the 4 disks in less that 3 days total.
Except that it did not work at all that way, and that EACH disk upgrade took more than 2 days and less than 3 (sorry, that was a year ago, I do not remember the exact figure).
And to make everything worse, I had 2 (!) power outages during the unprotected relayout period, and was very frightened I would lose my data; luckily, it did not happen…

Curiously, now that I have 2TB disks, a relayout STILL takes more than 2 days and less than 3.
So I am afraid the relayout times vary widely from one disk and Drobo and firmware (and cables and…) to the next, except that in all cases, a full + week appears way to long, and I do not even want to consider a month, even with dual disk redundancy… :frowning:

Interestingly, a FULL Drobo to Drobo copy of all files, through Firewire with a Mac in between, takes about as much time as a SINGLE disk relayout; meaning it is 4 times faster than a sequential disk upgrade (in addition to being way safer !). But this is probably to be expected since when the Drobo is almost full, any relayout will have to touch ALL other disks.
And there is the “small” drawback you need a second Drobo, with upward/downward disk set compatibility (to return the new copied disks to the original Drobo).
And AFAIK such compatibility exists mostly only between identical (types of) Drobos… :frowning:

geegi,

I interpreted yoyoma’s 24hr/TB as meaning for a single disk relayout, not an entire pack. Please correct me if I’m confused, yoyama :slight_smile:

If I interpreted him correctly then all of us are seeing about the same - about 60-65 hours to do 2.5TB for ONE RELAYOUT. That is also in line with other Drobo V2 (and probably V1) relayout times I’ve seen disclosed here with 2-3TB of data.

I don’t think the relayout has anything to do with the size of the disk because Drobo does not “initialize” new packs or new disks. From what I’ve seen, it may be fairly linear in relation to total capacity used.

I think you are agreeing with what I said previously- I question the utility of the disk expansion feature because at least with my current volumes I can copy the data back faster than I can expand two drives. At that point I’d rather have a fast conventional raid system that might take 12 hours to initialize an arrray and then might copy the data back at 40MB/s or faster.

I am shocked that a Drobo S is no faster. At least it has the option of dual redundancy but a month or more to do a drive upgrade is just not feasible. I’m hoping other S owners will add their experience here in case Mr Protocol’s experience is an out-lier for some reason. This alone would prevent me from ever spending that kind of money on an S, or really, at half the price because it’s not a money issue, it’s a fit for purpose issue.

I didn’t follow your point about the two Drobos. I maintain my backups on eSATA connected NTFS formatted drives so I don’t need two Drobos to copy my data back.

You are correct that I meant 24hr/TB, per each disk relayout.[hr]

[quote=geeji]No, I only had initially 4x1TB drives, filled up to 85%, thus amounting to about 2.4TB of loaded data [Drobo-V2].
If you accept the 1TB/day figure, I should have been able to upgrade sequentially the 4 disks in less that 3 days total.
Except that it did not work at all that way, and that EACH disk upgrade took more than 2 days and less than 3 (sorry, that was a year ago, I do not remember the exact figure).[/quote]

That sounds exactly right, IME, geeji. On my Drobo v2, 2.4 TB would take more than 2, less than 3 days to relayout after I upgrade ONE disk. If you are going from 41TB disk pack to 42TB, then 10 days might be needed to go through your whole upgrade. I have done that, and wound up with a sluggish Drobo. Even a week or month later, I don’t believe fragmentation was optimized. I deleted and reformatted the entire disk pack. Running a straight copy of the original data to the Drobo was peppy and afterwards, ran smoother, IMO.

So we all agree :slight_smile:

Unfortunately, I reached the same conclusion : your actual Drobo performance may heavily depend on Drobo relayout history.
The best performance is only available for a newly formatted Drobo, uploaded in bulk. But the catch is that most of us don’t have a second spare Drobo nor a few empty Terabytes standing around to offload the data while upgrading+reformatting, then reload the data.
Actually, for many of us, the Drobo IS the backup (yes, I know, RAID is not 100% failsafe, but …). Many of us bought a Drobo for the “easy” upgrade path and a reasonnable $/TB ratio.
And unfortunately, Data Robotics never advertise the performance consequence of multiple successive upgrades :(.

[quote=“geeji, post:9, topic:2360”]

I don’t know how anyone hanging around this forum could trust irreplaceable data to one single copy on a Drobo (or any other disk array made by man).

In particular, the 60+ hour relays to replace a failed drive appears to lead to a relatively high probability of a 2nd failure simply because our consumer grade disks are not very good, disks tend to be bought in batches that may share manufacturing faults, quirks or just similar MTBF characteristics. And stress leads to failure on marginal drives; there cannot be anything more stressful for a drive than to bang away non-stop for 60+ hours doing a relay. Or, in the case of an expansion, up to 240 hours non-stop (for 4 drives).

I maintain 3 copies of my Drobo data. I like the redundancy but I do not factor that into my backup plan. Just my personal opinion.

Edit: sorry for preaching but I just see too many horror stories to let your comment slide by.

NeilR,

It’s an excellent point that often gets overlooked by ‘hobbyists’.

RAID is NOT backup. If you invest in a Drobo, you should invest in a backup strategy for your Drobo.

If you are on a Mac, you have even less of an excuse, Time Machine works great for the home/SOHO user attached to a stand-alone USB/FW drive.

Personally, I have a WD 3TB external (USB2) that Time Machines my local iMac and the attached Drobo. I consider that a ‘weak’ solution because I have no off-site rotation schedule. I’m waiting on NewerTech to update the Voyager to work with disks >2TB and then I will be rotating a disk off-site.

Here’s a different solution, and the one I use because the Windows equivalent to TimeMachine is VSS, which is implemented into various applications, including Win7. From what I understand there are problems using VSS with Drobo (likely the thin provisioning).

I have a growing pile of 1TB drives, plus a couple of 2TB drives I’ve bought to ease this process. My Drobo is organized into top level folders that I don’t allow to exceed the “least common denominator” of my raw backup disks.

I have a couple of BlacX docks. I keep enough (non-redundant) storage always online to be able to backup the Drobo with a simple $30 backup scheduler (SecondCopy). That copy is always current and I can run the scheduler as often as I want, daily or even hourly or probably constantly with some apps. The raw drives get popped into the BlacX docks to get synced and then are able to go offsite.

The BlacX docks eliminate the cost of enclosures for my backups. The only downside may be wear and tear on the SATA connectors over time.

I use Weibetech disk cases to store the raw drives and transport them. I use the same boxes when I remove the Drobo drives before transporting it by vehicle, which I need to do from time to time.

My system does not work well with 1.5TB drives :-(. It’s better to stick with even 1TB increments.

If not for the need to take my data on field trips the Drobo would be the backup and the online backup would be primary, simply for performance reasons, they being eSata attached and as fast as internal hard drives (some are internal hard drives on a Win7 desktop file server).

Of course you are perfectly right.
Except that this data is not really “irreplaceable” since it is mostly ripped videos of DVDs I own and kept.
So if worse came to worse, I would “just” have to spend a week or two ripping again all my DVDs. Certainly not something I dare to contemplate, but I am ready to take the risk of that happening to save about $1000, which I very much prefer to use elsewhere.

One could also argue that multiplying insurances will never protect you against ALL kind of catastrophes, just lower the probability of some : the Drobo does not protect you against a 2nd disk failure during relayout, the Drobo-S does not protect you against hardware failure screwing a full disk-set, neither against a buggy Drobo firmware update screwing up some of your data, a single save will not protect you against corrupted data you did not discover as such before making the save, a backup which is not remote will not protect you against a robbery or a fire, you will discover your backup is useless the first time you actually need to restore it, etc…, etc…

As an engineer, I believe too much in Murphy’s law to buy the false dream of total security. The only question is how much are you willing to pay for protection against more and more unlikely events ?

And BTW, even if I agree with you that it is not “best practice”, I remain convinced that a huge majority of Drobo users are relying on the RAID feature of the Drobo to protect their data against disk failure, since it is the main advertised benefit of a Drobo (beyond single disk image). The fact that most of them are not too proud to admit they took this shortcut for economy reasons does not change that reality.
How many computer users do you think have ZERO backup, or an obsolete or useless never tested one ? 20% or 60% ?

You are taking a calculated risk and your risk is the inconvenience of feeding DVD’s back into your machine for a week (or two or three).

You are arguing that since the probability of total (or significant) data loss cannot hit zero then it is pointless to go any further? I don’t agree with that.

I agree that possibly even the majority of Drobo owners rely on that one copy of their data, but that does not make it a good decision. Some are sucked in by the marketing hype. DRI includes this “quote” on their “Customer Success Stories Link”:

“Simply knowing we can’t lose data and that there is always redundant backup on the fly, has given us a great sense of security.”

Personally I think that is an egregious lie, and I’m backed up by the “Best Practices” FAQ put on the site by the engineering side. Obviously marketing and engineering are at odds on this issue and obviously engineering has history on their side.

Some refuse to back up their Drobos because “I spent all that money to guarantee my data was safe” or some such thing (it’s all in the forum archives). They are deluding themselves and refuse to admit it even when confronted by the facts.

Some come here on their hands and knees BEGGING and pleading for someone to resurrect their failed Drobo array. They apparently have more than DVD rips on the Drobo :-(.

The classic case was a guy that loaded a pile of drives into his Drobo, feeding the original disks into the Drobo as he went along. He even added a 3rd or 4th disk to the Drobo unnecessarily and, of course, he lost the array before he was even finished. He was one of the beggers and pleaders and that was sheer stupidity (or blind faith in DRI marketing, depending on your point of view).

I suspect that the probability of losing data over a 10 year period approaches 100% with no backup or redundancy. I suspect the probability of losing a Drobo array over that period is about 10% (for any and all reasons, some beyond the control of the Drobo).

I suspect a second copy takes that down to the 2-4% area, depending on how it’s managed, and each successive copy cuts the probability in half, more or less.

It only costs about $330, max, to make a copy of a Drobo array in a Win or Mac readable format, not a thousand. So it all comes down to what your data is worth. My data is mostly images that are irreplaceable. The most replaceable part is my 600 or so music CD’s ripped into iTunes but even there I don’t ever want to have to do that job again.

I keep 3 additional copies of my Drobo data. all in native OS readable form (NTFS). I know I can read them because I do binary DIFs from time time to time. As disk gets cheaper I make more copies :-). I have a 4th copy of my images - straight out of camera - set aside specifically over concerns about corruption. Yea, I’m anal but I’ve run my probability down to near zero for very reasonable money (relative to the value of the images and what it cost me to acquire them).

I would venture as high as 80% of computer users. 60% of “advanced” computer users, and down to 25% of users who have used computer for 25 or more years.

[quote=“NeilR, post:14, topic:2360”]I have a 4th copy of my images - straight out of camera - set aside specifically over concerns about corruption. Yea, I’m anal but I’ve run my probability down to near zero for very reasonable money (relative to the value of the images and what it cost me to acquire them).
[/quote]
Yup. Trust no one (device, technology, whatever)!

Attempting to find out the relayout time on a DroboS. 5 drives, 23TB+ 21TB+ 1*1.5TB drives. 4.95TB of data on the unit. 86 percent full. Pulled a 1TB drive to simulate failure, lights go red. Waited a minute or two and put the 1TB back in. Not sure if the unit will return to ready state( based on knowing that it is good drive returned to service) or go through relayout. No time estimations so far after 45 minutes. I’ll edit or post an update in the future.

ETA:
It took 1 full hour, then the DroboS is back into full protected state, all lights green. I suppose that it recognized the 1TB drive when put back into service and was able to do minimal reconstruction. I will have to do a real test by pulling the 1TB drive and replacing with a new 2TB drive.[hr]
Again, Attempting to find out the relayout time on a DroboS. 5 drives WAS 23TB+ 21TB+ 11.5TB+12TB drives. and is NOW 5 drives=23TB+ 11TB+ 11.5TB+12TB drives. (I replaced one 1TB with one 2TB drive) There continues to be 4.95TB of data on the unit. 73 percent full now. Rebuild time is estimated at 28 hours.

yoyoma, is the S in single or dual redundancy mode?

Single.
And surprisingly, it is reporting about 2 hours left. Very quick.

It would be good to get more input on this from other S users. In fact, in my own mind, the viability of BeyondRaid as a useful storage strategy hinges on the answer.

NeilR,
It completed. It took just about 10 hours for the DROBO S with single redundancy to complete the rebuild. With 4.95TB of data on the drive. Interesting. I suppose that I will need to try it again using dual redundancy mode