drobo alternative

[randomlinh]

I'm looking into getting a Drobo/WHS or alternative, but trying to keep the costs down as much as possible. So we'll say the cap is ~$330 (going price of drobo v2 on amazon, not drobo S) not including any drives.

Ideally, it would have easy upgrade/adding of storage. Drobo is hotswap I think, and that is not necessary for my needs. Network sharing is the plus on WHS, but not a requirement.

Since this will be my main storage, it has to be easy to backup. I'd ideally like to just drop an external drive via USB or whatever, and have the critical data backed up, and then disconnect. but since I want incremental and history sorta like time machine, I suspect I might have to just do the backup through a machine.

Finally, it needs to be compact.

The Acer WHS is nice, but $400. And a slow atom processor. I know it's not a big deal, but I can build a cheap single core 2.2ghz amd machine for less. I lose on size though.

There are a lot of ReadyNAS things that might fit, but look to be over budget.

Last time I used FreeNAS it was pretty slow, and annoying to set up. It didn't have the ease of "just pop a drive in." Openfiler was even more painful.

Everything seems to point to getting a drobo, but I want to make sure I didn't miss anything. And it also loses out on the big network share option.

So, is there anything else out there I've missed, or is it likely a drobo for me?

 

[aceO07]

An alternative to the ReadyNAS is the QNAP. From what I've read it's as good and cheaper.

 

[billyb0b]

I say go with a QNAP device.

 

[Rifter]

Build a box and run whatever os you want on it, WHS if all MS network or some linux based for mixed networks. Should be able to find a mobo with 6 or more SATA pretty easy and run OS drive and optical off IDE/USB to save sata for your RAID. I built mine out of mostly old parts i had around, didnt need to buy much.

 

[RebateMonger]

The Acer WHS box is $390 with free shipping on Amazon, so it DOES meet your $330 (diskless) cost requirement since the included 1 TB hard drive is worth $70-$90. The Atom processor works fine for WHS. And WHS meets all your other requirements.

You can't really build your own WHS box for less. I've looked at trying to build cheap WHS boxes many times and I've never come up with anything nearly as nice as the pre-built boxes (like the Acer and HP) for even the same price. The components just aren't available.

 

[nobb]

The Acer might seem overpriced initially since you can build a better performing machine for the same price, but keep in mind that over the long term, its lower power consumption will pay itself off.

 

[randomlinh]

The Acer WHS box is $390 with free shipping on Amazon, so it DOES meet your $330 (diskless) cost requirement since the included 1 TB hard drive is worth $70-$90. The Atom processor works fine for WHS. And WHS meets all your other requirements.

You can't really build your own WHS box for less. I've looked at trying to build cheap WHS boxes many times and I've never come up with anything nearly as nice as the pre-built boxes (like the Acer and HP) for even the same price. The components just aren't available.

That's a good point, I hadn't realized it was a 1TB drive (thought it was a smaller 250GB).

I know the atom will work fine, it's always that nagging "you can get more for your money performance wise.. just in case" feeling =)

I guess I'll go with that acer, probably wait a bit and see if CES doles out some new boxes to discount the old one =)

 

[RebateMonger]

That's a good point, I hadn't realized it was a 1TB drive (thought it was a smaller 250GB).

Yeah, a 1 TB hard drive AND 2 GB of memory. While that 2 GB of memory was worth ten dollars a year ago, it'll now cost you at least fifty dollars. Yikes!

That's probably what's been driving up the cost of HP and Acer WHS servers the past several months. When memory was cheap, HP upped their standard WHS memory from the original 512 MB (which worked fine). And then memory prices went out-of-sight.

 

[pjkenned]

Last time I used FreeNAS it was pretty slow, and annoying to set up. It didn't have the ease of "just pop a drive in." Openfiler was even more painful.

I'm building a 30+ drive box right now. Tried FreeNAS and Openfiler during burn-in. Decided to stick with WHS (this time in Hyper-V on Win Server 2008 R2 since it has 12GB ram and an i7 920). They are priced right, but WHS really makes it easy.

Truthfully, an Atom based WHS is going to be fine for 99% of people. One thing that really doesn't get mentioned enough is that while company XYZ's NAS may have a ton of add-ins, WHS just has more. It has fewer dedicated add-ins than something like a Drobo... but it is built upon Windows, and Windows basically has software to do whatever you want available.

If I was going to get a pre-made WHS, I would get a HP hands down. HP has taken a leadership position in WHS software. If HP made things like their transcoding software available for $50-75, I would buy it.

I have a few friends that have drobos, and love them, but the WHS boxes are hard to beat right now if you know how to use remote desktop, and spend 30 minutes adding things like MyMovies, P80, Whiist and etc.

 

[DivideBYZero]

I love my WHS. Self built unit with a cheap mATX mobo, cheap ASUS case w/PSU, 2x 1Tb Hitachi drives, a dual core Celly 2.2Ghz and 2Gb RAM. Nice little box that lives in a closet, backs up all my other systems in the house and serves up movies, music and pictures a treat. Sharing from home with secure login for family is great, too, and as it's IIS and W2003 underneath, pretty much anything is possible. For example, I use mine as a caching Name Server to speed up DNS queries in my home network.

I also have the Storage and notification add ins so I can see graphical storage views with drive temps and get error reports to my BlackBerry should anything fail.

 

[latch]

I agree that you should just build your own and put whatever OS you want.

I recently built an Unraid server after using WHS since its first availability (I bought an HP EX470 when they first became available).

I much prefer Unraid over WHS for the simple fact that it scales much better. WHS uses replication as its redundancy strategy. So if I want 9TB of storage, I'll need 12 1.5TB drives - which will cost ~$1200. With a parity system, which unraid uses, I'll need 7 drives, which saves me $500.

The other problem with my WHS is the 4-bays...which forces you to buy poorly priced 2TB hard drives to scale...you can't just add another 1.5TB drive.

Of course, unraid is a straight-up NAS with network shares...so if you want incremental backups and stuff, you'll need to find (maybe even buy) software for your desktop - whereas WHS comes with all that stuff (for windows desktops).

And if you don't plan on storing > 4TB of data, then replication in a slim form factor is fine (4 bays is fine).

 

[pjkenned]

I think unRaid is very cool, as are their pre-configured systems.

After testing it a bit, the things I noticed were that the RAID 4 is single parity, like Raid 5. Also, I couldn't find good de-dupe.

For large arrays raid 4/5 makes me a bit nervous because it is pretty easy to have a second drive fail or temporarily drop from the raid. On larger arrays, having two parity drives plus a hotspare makes a lot of sense.

Dedupe is kinda cool since it lowers the amount of storage you need to keep things like backups by only storing a file once. If you have a lot of PC's with the same programs and files, this saves a bunch of space.

I do agree though that WHS scaling with multiple drives could use an improvement. It would also be great if you didn't need MBR volumes!

 

[RebateMonger]

I do agree though that WHS scaling with multiple drives could use an improvement. It would also be great if you didn't need MBR volumes!

I'm afraid I don't understand your and the O.P.'s comments about scaling. You can add as many disks as you want of any size you want. That certainly seems like scaling.

If you are referring to the WHS Folder Redundancy, which requires double the space for the redundancy, the, yeah, it'd be nice if redundancy took less space. RAID 5 obviously isn't the answer because of reliability issues. unRAID may be a better answer.

RAID 4 did the something similar, using a single disk to contain parity data, but unRAID does away with the striping that puts the entire array content at risk in a failure. Despite higher reliability, RAID 4 lost out to RAID 5, probably because of 4's slower speed.

http://en.wikipedia.org/wiki/Non-standard_RAID_levels#LimeTech_UnRAID

Wikipedia on "unRAID". It seems it's not perfectly "scalable" either:

"Disadvantages include being significantly slower than any single disk in both read and write, being extremely slow during drive rebuild, filesystem overhead(additional checksums are required if the parity-drive is to avoid querying the other disks to check the data disks in use), harsh scaling problems in computation overhead with many drives, and the main problem, that the parity drive has a much larger IO burden than the other drives (reducing reliability) & may bottleneck operations when multiple drives are used concurrently. The parity drive must also be at least as large as any single drive in order to provide protection. UNRAID is implemented as an add-on to the Linux MD layer."

Folder Redundancy in WHS is a nice added feature. I tend not to use it because I prefer backups over redundancy. But it's not horribly risky (certainly much better than no backup at all, which is what most people have), and it can be enabled with a single mouse-click in WHS. And it IS "folder redundancy", so the owner can choose which data needs to be redundant, unlike a RAID or unRAID system where redundancy is all or nothing.

As far as MBR, it's not, of course, an issue with WHS yet since disks larger than 2 TB aren't available. Certainly the next iteration of WHS will need a way to handle disks larger than 2 TB. The next generation of bigger hard drives is going to create a lot of issues with operating systems, disk utilities, backup programs, and data recovery programs.

 

[Emulex]

i wonder if you could take 2 WHS and use DFS-R to create a striped/load balanced pair of storage servers? would be cool to have one upstairs and one in the basement. then for magic they could team together using a virtual IP like lefthand VSA

 

[pjkenned]

As far as MBR, it's not, of course, an issue with WHS yet since disks larger than 2 TB aren't available. Certainly the next iteration of WHS will need a way to handle disks larger than 2 TB. The next generation of bigger hard drives is going to create a lot of issues with operating systems, disk utilities, backup programs, and data recovery programs.

MBR Is limited to 4 partitions per drive/ device though. So effectively it means that any disk in a raid array can only have 4 partitions, making maximum space 2TB * 4 Partitions = 8TB. With large drives, 1.5-2TB drives, it means you run (excluding hotspare) 8x 1.5TB Raid 6 arrays maximum. If you didn't have this limitation, you could, in theory, run a larger array keeping 2 drives for parity + 1 hotspare and reducing your overhead (with 1.5TB drives you get 2/3 of installed capacity as usable in raid 6 + 1 hotspare). That's what makes MBR an issue in large WHS systems.

If WHS could do software Raid 6, that would be awesome! In large arrays it just isn't practical to lose half of your capacity, and still be suceptable to two drive failures (since you could lose a drive with the primary data and the second drive with the copy. It's actually why I prefer raid 6 for four drives over two raid 1 arrays at this point.

Raid 4 is still fairly popular... NetApp uses it with WAFL, but that is a different category of hardware/ software than unRaid.

 

[RebateMonger]

MBR Is limited to 4 partitions per drive/ device though. So effectively it means that any disk in a raid array can only have 4 partitions, making maximum space 2TB * 4 Partitions = 8TB. With large drives, 1.5-2TB drives, it means you run (excluding hotspare) 8x 1.5TB Raid 6 arrays maximum. If you didn't have this limitation, you could, in theory, run a larger array keeping 2 drives for parity + 1 hotspare and reducing your overhead (with 1.5TB drives you get 2/3 of installed capacity as usable in raid 6 + 1 hotspare). That's what makes MBR an issue in large WHS systems.

Folks have published articles on WHS systems larger than 20 TB, so that, by itself, isn't an issue.

If you are referring to using RAID 5 or RAID 6 underneath WHS, it gives me a headache every time I think of it. Somebody here built a WHS server with RAID 5 underneath and asked about enlarging the array. It made something very simple (just adding another disk and mouse-clicking to add it to the WHS storage pool) into something very complicated.

My general observation is THERE'S NO INEXPENSIVE WAY to securely store "large" amounts of data. The safest is 1:1 copies using backups to disks, tape, online, etc. Anything less reduces the safety. There's no "wrong" choice. It's just a question of how much risk is acceptable to the data owner.

For most people, storage and backups shouldn't be a big issue anymore. A WHS server with a pair of 1 TB disk will make automated backups of all their PCs and all those songs and family photos and will allow plenty of redundant storage for some movies. The "problem" comes when wanting to securely store 10 TB of data. It's leading-edge stuff and cheap solutions are a problem.

 

[pjkenned]

My old/current WHS has over 25TiB worth of drives and is migrating to 44TiB+. The issue is clearly not the amount of storage, it is with the MBR requirement over being able to use GPT disks.

You can easily install 12x 2TiB drives on SATA ports and have a 12TiB usable array. Switching to raid 6 you can have two raid 6 4x 2TiB MBR partition arrays for 16TiB usable with either array being able to lose one third of its drives, but because of the MBR limitation, you are stuck with 16TiB usable and four of the twelve disks being used for parity. On the other hand, if you could use GPT disks, you could, in theory make one array with two parity disks and ten 2TiB drives for 20TiB available. While this doesn't seem like a lot, that extra 8TiB usable is 16TiB required using Raid 1/WHS and therefore and extra $1200 (8x 2TB drives * $150 each).

I'm using Raid 6 + hotspare on every volume. Adding a new 2TB drive to a 6TB raid 6 array is online capacity expansion isn't too difficult at all. Then it's just a matter of adding the 2TB raid 6 logical device to WHS where it is treated like a normal 2TB disk. Remove = remove from WHS, then remove partition, then remove array. There are 1-2 more steps involved on either end, but it isn't bad at all.

Just to make your head spin, the new environment does involve running WHS in Hyper-V, so it will require two more steps of setting the drive to offline, then adding it to the Hyper-V WHS to manage directly (in the test setup it is taking me about 2 minutes including logging into RDC on the 2008 R2 box.) There are going to, of course, be multiple controllers running the arrays so there is a small interim step of choosing the correct controller now in the Adaptec Storage Manager before doing all of this.

Now, is it cheap? No way. Although I will probably have a FAS2020 in a month or two that will take over backup duties (and that can only scale to about 65TB right now), but Raid-DP is a software raid 6 more or less supporting up to 24ish drives per array IIRC (not that you would want that many drives in a single array).

 

[randomlinh]

wow, this blew up... heh. I doubt I'd build my own at this point because my focus is on size. Though, it does eat up my ability to expand.

However, it seems you all have pointed out a small factor I misunderstood, I hadn't realized WHS just used folder replication. This will likely push me to a drobo now.

 

[DivideBYZero]

It uses folder replication to separate disks, so you have redundancy. It's not replicating on the same disk.

 

[pjkenned]

However, it seems you all have pointed out a small factor I misunderstood, I hadn't realized WHS just used folder replication. This will likely push me to a drobo now.

The replication isn't a bad thing as it stores copies on separate disks. Also, if you want you can replicate some folders but not others. This is super useful in the event you have a bunch of stuff collected that you don't really care about. One example is all of the random video car driver, firefox version, and etc installers that make life easier to have locally, but there's really no need to have redundancy for that stuff since new versions can be easily downloaded from nvidia.com.

The other thing that WHS has is dedupe, which means that you aren't saving the same big file three times if you have three PC's. I haven't tested it to get an exact number, but my ballpark guess is this feature is saving me in the low 100's of GB's right now, mostly because I run a pretty standard configuration on each client. I also replicate some media on mobile devices.

Overall, the WHS system is probably better than Raid 1 from a storage efficiency standpoint because of dedupe and the ability to change if a share has duplication (that sounds funny but it is what they are called.)

 

[RebateMonger]

The other thing that WHS has is dedupe, which means that you aren't saving the same big file three times if you have three PC's. I haven't tested it to get an exact number, but my ballpark guess is this feature is saving me in the low 100's of GB's right now, mostly because I run a pretty standard configuration on each client.

It was almost embarrassing when I put in my first office Windows Home Server. With ten business PCs, I'd advised the client to add a 1 TB disk to the 500 GB disk that came with the (early) HP Windows Home Server.

After the first backups were finished, I was amazed to find that the "used" disk space on the server was so tiny it barely showed up on the Storage pie chart. Those PCs all had Windows XP Professional and MS Office 2003, so only a single copy of most of the files was needed.

 

[latch]

I'm afraid I don't understand your and the O.P.'s comments about scaling. You can add as many disks as you want of any size you want. That certainly seems like scaling.

Sorry, when I said "doesn't scale" I meant "doesn't scale as well as other solutions with respect to cost" and then went on to explain the hard drive cost of duplication and how that proved to be a limiting factor for me (to the point where building a replacement my machine was actually cheaper than buying the necessary hard drives).

 

[pjkenned]

After the first backups were finished, I was amazed to find that the "used" disk space on the server was so tiny it barely showed up on the Storage pie chart. Those PCs all had Windows XP Professional and MS Office 2003, so only a single copy of most of the files was needed.

That, of course, is why all of the big storage vendors offer this in some fashion. If you have 1 client it isn't a big deal. By the time you have 4-5 clients, it is saving a ton of space. The reason I use WHS backup software versus built-in Windows 7 or what have you backup is that you get both differential backups, and dedupe. Cheap backup programs won't do that, and it is a big WHS differentiator.

 

[trexv]

i don't think anybody mentioned flexRaid: http://www.openegg.org/FlexRAID.curi it basically uses Raid4 parity to ""protect data, not drives"". so you don't create an array, you just tell flexraid what data you want protected by parity (whole drives or just folders) and where to store the parity data. it's not real time RAID. it's useful for static data. they also have something called flexraid-view, which creates a unified view of one or more data sources (drives, folders, network shares, etc). for example, if you have 3 ntfs drives, you can put them under a new one, for example z. so you don't need to ""initialize"" them like WHS wants, just create the view and you have access to all the data form in one place. you can also create virtual folders as well for grouping different folders from one or more data locations. this one is available here (reg required): http://www.openegg.org/forums/posts/list/156.page folks have been using it by itself or with WHS, hardware RAID, etc. flexraid basic + flexraid-view is a pretty powerful, feature rich system. in the past i've used freenas, moved to openfiler for my 6TB server (both on hardware RAID 5 +hotspare), but i think i'll be moving to flexraid. oh, did i mention it's free?

 

[imagoon]

That's what makes MBR an issue in large WHS systems.

Meh just enable and use GPT in WHS:

http://social.microsoft.com/forums/en-US/whssoftware/thread/e9f9d27a-f8b1-4fc0-872b-bc8e3b2609f0/

I personally am confused what RAID 4 is talked about so much here also. RAID 4 is a basically RAID 5 where the parity is not rotated. Without special configuration tricks this tends to limit the speed of the array to the speed of the parity disk. I know that NetAPP has a lot built in to RAID 4 tech but they use a ton of tricks to increase the performance of the machine like doing a mirror of the parity disk. RAID 6 is also a cool tech but it is more like RAID 5 with dual parity (not all that much unlike Netapp's RAID 4 approach.) However I agree with rebate monger, when you have a ton of data you need to keep alive, RAID 5 and even RAID 6 get far to risky. The odds of failure weaken the system to much in my opinion. This is where you start heading in to things like RAID 10, RAID 50 and RAID 60. I start worrying about "enclosure loss" after a certain point.