Are NAS and RAID really worth?

[geno888]

Hi, I hope I'm posting in the right section of the forum

I was discussing with a friend, and we both started to have doubts about NAS and RAID being really the best solution for data storage. I hope that someone with direct experience will help us to better understand things

I hope that I'll be able to explain clearly the argument, sorry for any misunderstanding because English is not my primary language.

Let's say that I'll have a 4 bay NAS (it's just an example), and I put inside 4 x 10TB drives. In the best case scenario, I'll use NAS drives (IronWolf series from Seagate or Red series from WD), so let's say that we are in the best conditions a consumer user can get. The system will be configured to make a RAID 5 so the system can support failure of 1 drive.

Let's say that the drives installed in the NAS have a 3 years warranty. So, theoretically these drives should work for at least 3 years. Let's say that after three years one drive fails: installing a new drive will cause no data loss because the NAS will rebuild all missing data from the other drives.

And this is where the argument started. After 3 years, all drives are pretty much teared (even if still working) and rebuilding a so large amount of data (10TB is not a joke) will stress a lot all the old drives. My friend was telling that this will cause one or more of the other drives to fail at the end of the rebuilding or, worst, during the rebuild process, causing the loss of all data.

Even if in theory this reasoning seems making sense, I answered my friend that I'm not sure that this is always true. I'm wondering what happens in large data farms (like google or facebook just as example) that use even larger arrays. If the above is true, i.e. that repairing an array will tear all other disks then every time a disk fails all disks must be replaced as soon as possible to avoid data loss. If this is true, google must spend millions only to replace HDDs in their data farms... It seems pretty improbable to me but of course I'm not an expert in these things.

My argumentation is if data rebuild is so much stressful for the system, then why RAID is still so used around the world? Maybe the drive failure after a rebuild is not that common (even if it can happen).

After much discussion we couldn't find a true answer to this question.

I thought to ask here hoping that someone with direct experience could answer us. If the above reasoning is true, the larger are HDDs used, the more a NAS becomes unreliable. But, on the other side, using a NAS with small HDDs is not cost effective anymore, so again I'm wondering if using a NAS is really worth.

Maybe there is a sort of green zone that makes things more balanced? Just to make some examples:

1) maybe using only 4TB drives is the best compromise to reduce the chance of killing HDDs in case of array rebuild?

2) Or maybe the best option is to use at least 8 HDDs instead of just 4 (so each drive will be stressed lesser because the data needed to rebuild the faulty drive are more evenly distributed among all HDDs)?

3) Maybe the chance of killing drives is directly proportional to the amount of data to rebuild, i.e avoid to fill the NAS more than 50% will reduce the chance of damages? But if this is the case, it is not cost effective at all: what is the point in having 30TB of storage space if I can use only 15TB? Again, NAS seems not worth.

4) Maybe the best options depends on the HDD size, for example for 4TB drives a 4 bay NAS is the best option, but for 10TB dives is better to get a 8 bay NAS to spread data on more disks to avoid too much stress in case of rebuild.

Actually at this point I'm just confused.

There is someone with direct experience of data rebuilding of a large disk (let's say at least 8TB)? Is it really a so destroying procedure to rebuild an array with large HDDs?

The only alternative I can think is using single HDDs with no raid and store these drives in a drawer, but also this is not a 100% guarantee that data will be safe, because even if not used a HDD can become useless after some years in the drawer (maybe the motor will not spin anymore, or whatever problem can arise). I have read about people complaining that a HDD used few times is not working anymore after staying completely still in a drawer for a long time.

I hope that I explained clearly enough.

Thanks anyone who will read all this and give me some answers

 

[Essence_of_War]

Hi, I hope I'm posting in the right section of the forum

I was discussing with a friend, and we both started to have doubts about NAS and RAID being really the best solution for data storage. I hope that someone with direct experience will help us to better understand things

......

My argumentation is if data rebuild is so much stressful for the system, then why RAID is still so used around the world? Maybe the drive failure after a rebuild is not that common (even if it can happen).

......

The only alternative I can think is using single HDDs with no raid and store these drives in a drawer, but also this is not a 100% guarantee that data will be safe, because even if not used a HDD can become useless after some years in the drawer (maybe the motor will not spin anymore, or whatever problem can arise). I have read about people complaining that a HDD used few times is not working anymore after staying completely still in a drawer for a long time.

I hope that I explained clearly enough.

Thanks anyone who will read all this and give me some answers

What do you mean by "best solution"?

Any discussion of "best" solutions has to take into account the intended use case. RAID is fundamentally a tool to increase availability and performance. If you have a RAID-5 composed of 3-drives, when (not if) one of the drives dies, you don't have to restore from a back-up. If you're a home user you may not need availability. Restoring from backups might be fine.

One way to think of NAS (Network Attached Storage) is a tool to share a pool of storage between multiple users, or to separate storage from compute. If you want a workstation, a laptop, a tablet, and a phone on the same LAN to all have access to the same media collection, or to all back themselves up to a central location, it can be really handy to have a NAS. If you don't need or want to do that, a NAS obviously isn't the "best" choice.

I think you're just confusing yourself in the weeds of discussions of drive failures and rebuilds. All drives fail. Make backups. If you can't tolerate going offline to restore from a backup, ADDITIONALLY build your drives into a RAID.

 

[Rifter]

RAID is not a solution to having to make backups, it never has been, dont use it as such. ALWAYS MAKE BACKUPS.

RAID is for maintaining uptime, and for speed, nothing more.

 

[cfenton]

It's definitely not true that drives will always fail during a rebuild. That would make RAID 5 or RAIDZ1 mostly useless. It's certainly possible that another drive will fail during a rebuild, but that's the worst case scenario. If you're worried about that, just use a setup that's more tolerant of failures like RAID 6 or RAIDZ2. RAID isn't supposed to be a backup solution, so it will never be 100% safe, but RAIDZ2/RAID6 would protect you from non-catastrophic failures. Always have important data in as many places and formats as possible.

 

[poofyhairguy]

First of all three years isn't right. I have had drives in my array stay on for six plus years and still work.

Secondly part of the reason I use Unraid is because the drives are readable outside the array so if a second drive dies during a rebuild I just lose the data on those drives and not everything. That has actually saved my ass before.

 

[dave_the_nerd]

NAS and RAID aren't the same thing. If you need NAS, you need NAS. If you need RAID, you need RAID. You can DAS a RAID or NAS a JBOD; you can even PASS a MELON if you try hard enough and have a good reason.

As for your concern about rebuilding - it stresses the drives, but not that much. There's a lot of fearmongering out there. Don't worry so much. Lighten up. Have a drink.

Usually, though, it's faster/easier to destroy the array, assemble it with new drives, and restore the array from a backup. So if you have current backups and can tolerate some downtime, that's what you do.

 

[BonzaiDuck]

Well, I can explain my own dabbler-DIY-home-server use-case. Let me say this all depends on the data you intend to capture, the data you intend to use, the data that is persistent and doesn't change much or at all, what data is volatile and changes more often, and what data is important in both business and personal respects.

For the latter case, I archive movies for my system that provides HTPC function using relatively few clock-cycles, or maybe 5%. what I save in DVR captures ends up on the server after a time with no duplication. If I lose those files -- I just lose them. They're not that important.

Overall, I use StableBit Drivepool and may continue to use drive-pooling for my server. I can duplicate data at the folder level, so I assure that nightly client-system backups are duplicated, as are personal finance, correspondence, business for rental property and association board and my e-mails. For all the household workstation, all e-mail is backed up by archiving PST files. There are PDF document archives for taxes, receipts, book-scans I want to OCR into readable PDF for word and phrase searches. Those items are all in "User" subfolders on the server, and they all get duplicated.

I use four 2TB disks -- Seagate NAS -- for an 8TB pool. It is now 1/3 full, because I DVR'd 54 episodes and all 5 seasons "For Breaking Bad," various movie and news-program captures -- very many unencrypted non-DRM captures, and so on.

I can "trip-li-cate" instead of duplicate, but that means more intensive use of drives or more drives -- it requires slightly greater capacity, or it will fill capacity by adding 50% to duplication.

If any drive fails, all of the duplicated folders are saved, but the unduplicated files on the dead disk are also dead and lost.

What about backups? I back up the server OS on a 320GB hot-swap. I back up the important personal and business data on -- believe it or not -- a 500GB Hitachi IDE drive that may be ten years old with low mileage, converted to SATA with an adapter switched by the drive-bay key-lock - a custom job with the bay-wiring using a 5V regulator. The client-backups are backed up on another similar 500GB drive, but I suspect I'll need to use a 1TB soon. the duplication of those client-backups could actually be "enough," but backing them up to an otherwise off-line hot-swap is done probably quarterly. The important stuff is backed up every week or two.

You should just need to sit down and figure out how much data you want to keep available, how much requires "extra insurance," how much you could lose with little regret.

The Drive-pooling allows me to use AHCI storage-mode, actually improve performance a little bit for the duplicated folder access, and makes it easy to restore data since each disk is written in files and folders independently accessible in any regular SATA connection to another system.

That leaves you to figure out your capacity, your mode and redundancy strategy, and the size of backup disks.

 

[DrMrLordX]

If you're into drive pools, I would think a FreeBSD box with a ZFS zpool would be a nice solution, if you're in the mood to set it all up.

I currently do not use a NAS because my network performance leaves something to be desired. Maybe if I wired it directly to the router, but still my wireless NICs have all generally sucked.

 

[BonzaiDuck]

If you're into drive pools, I would think a FreeBSD box with a ZFS zpool would be a nice solution, if you're in the mood to set it all up.

I currently do not use a NAS because my network performance leaves something to be desired. Maybe if I wired it directly to the router, but still my wireless NICs have all generally sucked.

Others have pointed me to FreeBSD, but there are learning curves, I feel old all the time and StableBit/CoveCube software has been stellar as far as I can see. Further, their tech-support is VERY much willing to help on a lot of things.

I'm setting up a new server-replacement on a z68 gen3 mobo with i5-3470 and 16GB Corsair XMS 4x4 16GB RAM. I intend to initialize a 2-disk pool on top of a 250GB OS-boot with Win 2012 R2 Essentials. Then I'll consider trimming the 4-drive WHS pool down to 3-drive, and move both disks and files. Eventually, the WHS will have less storage and the 2012 will have more, until I can totally decommission the WHS2011 having successfully set up the 2012 to do what the WHS2011 was doing.

Even so, I was thinking to add web-server function to one or the other, and I'm beginning to think it will be the 2012 system.

The Wireless part of a router network can stream television from a Silly-Dust tuner box, and provide access to a server on a wired network. I don't spend a lot of time with folks savvy about tech-change except at this website, and don't know what innovations other than 10Gb that have taken hold. We originally made a cable-drop from attic through second-floor to first floor, tapping into our cable-coax, and if I remember, the IP-tech initially set it up for 1pc and the cable-modem, which I immediately changed by inserting the router and reconfiguring. So I have Gigabit, but the link to upstairs is CAT-5. My retired phone-company-tech brother tells me that CAT-5 is satisfactory. Never had a problem with it, but that's the weak link -- everything else is CAT 6.

For something like a cable-drop through the wall, you only need a face plate and an RJ45 port for the cabling. Then another router if there are more connections on the same floor and room.

Wireless networking is only an extension of my wired router, and the server is definitely wired, as are the other devices and desktops. I use it for my laptop and I-phone. We have printers that are hermaphroditic for USB, RJ-45 Ethernet, and wireless. I wouldn't use my printer as "wireless" if I had the wired infrastructure I have here. I can see where it would be useful. But I don't need to use it unless I want to rid my room of one less cable -- for the printer next to my desk requiring only a 6-footer.