This topic has been marked solved and closed to new posts due to inactivity. We hope you'll join the conversation by posting to an open topic or starting a new one.
I plan to upgrade from my ReadyNAS Duo that served me well for 5+ years running 2x2TB in Raid 1. Consider to buy 2- or 4-bay (or even 6-bay hope not) NAS and it is for home use with small office.
I have questions concerning the advantage of separating data on different disks or more software division in volumes and share provide enough separation.
Typically I need on the NAS for the coming years A : Archive data - photos, media etc - accumulate over time does not change - about 1.5 GB growing B : Backup of 3-4 Macs - need about 1 TB for that - change as backups do C: Shared folders in the LAN - need about 0.1 TB for that - change a lot 😧 Shared folders externally over internet with business clients - need about 0.2 TB for that - change a lot - need encryption and security E: Web-page - 0.01 TB or so - need internet security etc
Today I only have A-C but consider to expand the use with D and E with the new NAS The old Duo likely will be the new backup destination.
Questions (after reading briefly in Readynas Software Manual 6.2):
1) Enabling bit-rot protection is mainly interesting for A and perhaps B Should this be on typically 2 raid1 diska and physically separated from another 1-2 disks (for C-E)? Hear that it should perhaps not be enabled where data is changed all the time since that would lead to disk fragmentation, but I may misunderstand this.
2) Enabling encryption for a part D (for business sharing over internet) is that possible? and if so recommended to have that for a separate disk? or enough with a separate volume on a disk that is otherwise not encrypted?
3) Security against un-friendly internet access. Consider to house a web-page or two on the NAS but worried for that it opens up for attacks from outside. Does it help in anyway to have such stuff on a separate disk? or enough with separate volume?
4) I know the volume can be divided up in shares and each share can be password protected and so I actually work today. Is there any advantage of dividing more physical into separate disks or share separation is enough? So would a 2-bay system with shares for A-E give about the same integrity, security, performance?
5) I read in the manual that enabling checksum and bit-rot protection decrease performance. If i have I have 2 disks raid1 configured and divide into two volume one with checksum and bit-protection enabled and the other without, will they then have different performance? Or do I need to have the volume without checksum and bit-rot-protection on a separate disk to get the performance gain?
b) Same question for disk fragmentation?
6) Last question Is the limitation to 4TB disks on 3-series a hardware or software-constraint, i.e. can a future upgrade of OS 6 open up for larger disks?
Generally speaking I do not think performance is an issue. My focus is mainly on data-integrity and internet security if I open up for internet access option D and E above.
The questions span a large range and would appreciate also partial response.
(1) Bit-rot protection can be enabled on a share by share basis. Snapshots and bit-rot protection shouldn't be enabled on shares where there is a lot of data churn. They can lead to heavy fragmentation.
(2) You might be confusing disk encryption with network encryption. Encrypting the data on the disk will not make business sharing over the internet more secure. Disk encryption will make it less likely that someone stealing the NAS hard drives can get the data. If I were in your shoes I'd be looking at hosted web services to handle your business client sharing - it will probably offer better performance, and eliminates any chance that a business client can reach the other data. Part of the answer here depends on whether you have long-term relationships with your clients (e.g., is it practical to give them user accounts on the NAS or not).
(3) More volumes will not be more secure. The most significant security vulnerabilities allow access to the operating system itself. In-house web pages won't create extra risk unless you allow the NAS to be reached over the public internet.
(4) It sounds like you might be using share security. That is deprecated even on your duo, the current mode is user-security. Each user can have different access, and has his/her own password. More volumes will not be more secure (repeat after me... More volumes will not ... 🙂 )
(5) You can't divide the RAID-1 array into multiple volumes. You can set the checksum/bitrot protection for each share though. If they are turned off on a share, then the extra overhead shouldn't be incurred. The performance loss is only on writes btw.
(5b) Same answer.
(6) The 300 series HCL has 6 TB drives, I am not sure why you are thinking only 4 TB.
Thanks for a good and comprehensive answer! A few follow-up questions.
1) Bit-rot etc. I read in the manual pg 27 that checksum function is enabled on volume level, and then I understand on pg 40 that bit-rot protection is set per share or even per folder. Then I wonder if turning on the checksum function for the whole volume does not slow down it all, even though I may use bit-protection only for a part of the volume?
1b) Also wonder if regular backup-up of a few Macs should be characterized as too much “data churn” - and usually you would not care to do bit-rot protection on?
1c) In my scenario question is whether A (which naturally would have bit-rot protection on) and B (which perhaps should have bit-rot protection off) should be allocated to the same pair of raid1 disks?
2) Good you emphasize the difference between disk encryption and network communication encryption. Those customers I work with over 1-2 years I thought it would be simple to share a folder with common documents with over my own NAS, rather than using a hosted web-service. But maybe not a very good idea.
2b) How much would encryption slow down?
3) Internet web-page. For these years with my old ReadyNAS Duo I have hesitated with sharing photos or even a web-page over internet, since the Duo has been my golden archive. Just a gut feeling. Now with a new NAS and much improved security I wonder if this hesitation is still valid?
4) Good. I will look more into user-security.
6) Good I see that hard ware compliance sheet does allow 6 TB, although the Netgear comparison sheet etc only state 4 TB (while the new 200-series boasts with 6 TB).
7 new) Looking for good reasons to buy 4-bay rather than 2-bay. Yes, much cheaper to expand later on with 4-bay, but I would like to see something more, and perhaps is among the answers above.
Would be good with a follow-up answer before this forum pages are totally changed, any day 🙂
1) We link enabling Bitrot protection to enabling/disabling CoW (Copy on Write). So the main reason for not using bitrot protection for some shares would be if using CoW is inappropriate for those. If you are going to make a huge number of in place changes to file then CoW is inappropriate. CoW makes a copy on write so with a huge number of writes you will get a lot of fragmentation.
2) Disk encryption is encryption of the disks, whereas network communication encryption is encryption of the data in transit from one location to another e.g. backing up over the web.
2b) It does depend on what you are using it for, but you would need a more powerful NAS than if you are not using encryption.
3) I wouldn't recommend forwarding ports, but using e.g. ReadyCloud would be good.
4) The Duo only had user security.
6) You are probably looking at an old comparison sheet. When the 300 series was first released 4TB disks were the max available.
7) You can have a volume of 3x the capacity as if you were using smaller disks. You get a nice LCD screen on the front. The CPU and RAM are the same as for the 2-bay.
1a) I guess you must enable check sum on the volume in order to later be able to enable bit-rot protection etc for folders, right? Or do I misunderstand the role of the check sum function?
1b) Does this enabling of the check sum of the volume slow down even if I do not user bit-rot protection at all, or only for smaller part of the volume, and then I mean the performance of the not bit-rot protected part?
7) I try to find arguments for when 2 separate volumes may be a good idea. Then a 4-bays system makes additional value to me. Or you would say the main argument is raid5 and less redundancy and you more “effectively” use your disks.
1a) I guess you must enable check sum on the volume in order to later be able to enable bit-rot protection etc for folders, right? Or do I misunderstand the role of the check sum function?
The bit rot protection is a differentiating feature of Netgear's, so the available details are a bit sketchy. But I think you are correct that the checksum is needed for it to work.
janpeter wrote:
1b) Does this enabling of the check sum of the volume slow down even if I do not user bit-rot protection at all, or only for smaller part of the volume, and then I mean the performance of the not bit-rot protected part?
Of course it is an extra step, and it needs to be computed/saved for every disk write. I have it enabled on my RN202, and have not seen any obvious slowdown. Of course if you want best performance you could get a RN516. That also provides error correcting RAM (ECC).
janpeter wrote:
7) I try to find arguments for when 2 separate volumes may be a good idea. Then a 4-bays system makes additional value to me. Or you would say the main argument is raid5 and less redundancy and you more “effectively” use your disks.
First - it is always cheaper to expand the NAS later on if you have an empty slot. For instance, if you have an RN312 with 2x4TB disks, then increasing the capacity to 6 TB requires you to buy two new replacement drives( $500 US at the moment), and stop using the old ones. If you have an RN314 (or an RN316) with empty slots, then you can get grow from 4 TB to 8 TB by adding a single 4 TB drive to the array ($160 US at the moment).
So in this example the RN312 upgrade costs $250 per added TB, while the RN314 upgrade (with an empty slot) would only cost $40 per added TB.
Second - having 4 or more slots available lets you use some of the more advanced RAID modes. These modes don't increase storage efficiency, but they have other advantages. RAID-6 provides dual-redundancy (survives all combinations of 2 disk failures) and RAID-10 (better performance than RAID-6; survives some combinations of two disk failures but not all).
But to get to your question - Using two RAID-1 volumes lowers capacity, but recovery of data if the array fails is generally easier, because the mirroring allows each drive to be independently mounted. Writing to RAID-1 is also somewhat faster than writing to RAID-5.
Also if you are considering RAID-0, having one volume per disk is much more reliable than creating a single RAID-0 volume that spans multiple disks. If any disk in the RAID-0 array fails, the entire volume is lost. So if you want maximum capacity in an RN312, I recommend having 2 volumes (and would recommend 4 in the RN314).
Keep in mind that RAID is not a form of backup. It will keep your data available through "routine" disk failures and it allows you to expand your storage while keeping your data on-line - both of which are great. But RAID arrays can fail, and there are other system failures (and of course disasters) that result in data loss. So you should invest in a backup solution as part of the purchase.
Thanks again! If I buy 314 then I consider 2 disks raid1 and a 1 disk separate volume and 1 empty slot for future.
When I google on “btrfs raid5” I get the feeling that this in some implementations is not that well working and some fundamental development are going on. I see that the Linux Kernel version plays a role and some key improvement came with Kernel 3.19 but hard for me to judge this information, even though I searched our forum on kernel too. What Kernel does OS 6.2.4 have?
I guess raid1 is more “safe” sort of right now, but would be good to hear something that lessen my worries.
Closely related here is whether improvement for raid5 requires some hardware upgrade, or it is rather about firmware updates and fundamental ones like in the Linux kernel?
From the conversation I somewhat realize that I may be able to continue live with a 2-bay system 312 for a few years and then invest in a multi bay system when raid56 and necessary hardware is more mature.
When I google on “btrfs raid5” I get the feeling that this in some implementations is not that well working and some fundamental development are going on. I see that the Linux Kernel version plays a role and some key improvement came with Kernel 3.19 but hard for me to judge this information, even though I searched our forum on kernel too.
Those articles are talking about the vanilla kernel found on kernel.org. Moving to a new kernel has its advantages and disadvantages.
janpeter wrote:
What Kernel does OS 6.2.4 have?
6.2.4 is using a 3.0.x kernel but it has a large number of BTRFS patches backported amongst other things.
The 6.3.x firmware (currently early access for the 300 series) uses kernel 3.12 again with a large number of BTRFS patches backported amongst other things.
A new kernel brings new features and other enhancements. There was an annoying bug present in early 3.19.x kernels. By steering clear of kernels while they are bleeding edge we can hopefully avoid most of the unexpected issues that a new kernel series brings and backport important fixes that are stable.
We are not using the in-built BTRFS RAID. We are using md raid. We have a much more limited exposure to the RAID-5 write hole, since we use a transactional filesystem and we respect barriers. So we can get very close to doing atomic writes. We can also repair the data if just one chunk is out of sync. Our competitors are much, much, much more exposed.
janpeter wrote:
I guess raid1 is more “safe” sort of right now, but would be good to hear something that lessen my worries.
I'm using RAID-6 with my 516 and I'm comfortable with using that. We have users who have been using RAID-5 and RAID-6 since we first released our OS6 devices a few years ago.
janpeter wrote:
Closely related here is whether improvement for raid5 requires some hardware upgrade, or it is rather about firmware updates and fundamental ones like in the Linux kernel?
The RAID we use md raid is software RAID, this is very mature and we have been using it for a number of years. The in-built BTRFS RAID is also software based.
So a hardware upgrade will not be required to get improvements to the md raid we use. There may however be some hardware limitations e.g. for things like disk capacities supported. We can only guess as to whether higher capacities not available yet will work or not.
7) I try to find arguments for when 2 separate volumes may be a good idea. Then a 4-bays system makes additional value to me. Or you would say the main argument is raid5 and less redundancy and you more “effectively” use your disks.
I'm FAR from an expert when it comes to the ReadyNAS systems, but thought I'd share that I found a place that has the RN316 for not much more than the RN314 (and they are an authorized retailer, so the warranty is good). I chose that route myself to be able to expand to six disks in RAID6 later on. Don't know what the policy of "advertising" retailers in public is, or I'd just tell you the retailer here.
Don't know what the policy of "advertising" retailers in public is
No issue here with quoting prices, though if a seller were to abuse that I'd at least warn them, or possibly ban. Not sure what other mods would do. Overall, policies here are quite permissive.
Amazon US pricing today has a gap of only $40 between the RN314 and the RN316 (both diskless). That makes going with the 316 a no-brainer.
The gap between the RN312 and the RN316 is $140, so budget for home users might start to kick in there. Though if you look at the analysis I posted earlier, you recover that extra money (and more) the first time you need to expand your storage.
Mdgm replied, but the more succinct answer is "no".
BTRFS has some experimental features that integrate RAID into the file system itself. They are interesting for the future, but not ready yet - and Netgear isn't using them.
Don't know what the policy of "advertising" retailers in public is
No issue here with quoting prices, though if a seller were to abuse that I'd at least warn them, or possibly ban. Not sure what other mods would do. Overall, policies here are quite permissive.
Amazon US pricing today has a gap of only $40 between the RN314 and the RN316 (both diskless). That makes going with the 316 a no-brainer.
The gap between the RN312 and the RN316 is $140, so budget for home users might start to kick in there. Though if you look at the analysis I posted earlier, you recover that extra money (and more) the first time you need to expand your storage.
Mdgm replied, but the more succinct answer is "no".
BTRFS has some experimental features that integrate RAID into the file system itself. They are interesting for the future, but not ready yet - and Netgear isn't using them.
OK, good to know. Yes, seemed like a no-brainer to me as well, since I was already looking at the RN314. Reason I went for the RN316.
Cheaper, and where I bought it from was here: http://www.nextwarehouse.com/item/?1378932 That one is currently $572.47 with free shipping, but no tax. You can also pay for faster shipping, which still ends up being cheaper than from Amazon after having to pay for tax. Get a "credit" if you go for the paid shipping also. I went that route, to later buy another hard disk from them to put in the RN316.
Edit: Oh, if you live in CA or MA, you have to pay for tax when ordering through NextWarehouse also... in that case, Amazon ends up being a better deal.
Thanks a lot again! I want to round of this topic with a couple of questions concerning performance and stretching upwards.
1) I wonder how big role has the ECC RAM-memory for data integrity, which comes with upgrade to 5-series. I understand it improves somehow the bit-rot protection and more. Would be good with some brief explanation of reference to where it can be found.
2) I consider more use of PLEX for media handling which I today mainly do through a Mac mini in our network. I have seen in PLEX documentation that 3-series work well although here are some limitations around video-streaming. My idea is that I should continue to run PLEX from Mac mini but “integrate” ReadyNAS better with it and use ISCSI-connection or similar. Is this a fruitful path for readynas 3-series or much better to climb to 5-series?
3) I work regularly with VirtualBox on my Mac and seen that it can also be run on ReadyNAS at least the ones with stronger processor. Just wonder if 3-series atom processer can work with VirtualBox and appreciate some links. Or also here if you really need to climb to 5-series.
1) I wonder how big role has the ECC RAM-memory for data integrity, which comes with upgrade to 5-series. I understand it improves somehow the bit-rot protection and more. Would be good with some brief explanation of reference to where it can be found...
ECC RAM is a must-have in data center servers.
The main thing to keep in mind is that bit-rot protection in ZFS and BTRFS protects the data on the disk. It doesn't protect the data while its in RAM (either on the way to the disk on a write, or on the way to the user after a read). Clear evidence of true "bit rot" is pretty hard to find, but everyone agrees that memory errors in data centers occur (the ECC memory reports them!).
Memory failures are relatively rare, but when one does occur the odds of it resulting in data corruption is about 10%. That is based on this study: http://research.cs.wisc.edu/adsl/Public ... fast10.pdf). This particular study focuses on ZFS, but potentially is applicable to BTRFS as well (since AFAIK neither implementation guards against memory failure). In a file server application, about 3.6% of the time a single bit flip resulted in corrupted writing of the data. Odds of a corrupted read was about 7%.
How much this matters depends on how frequent memory failures actually are. There are a couple of studies on these failures, but I find it difficult to draw any simple conclusions from them - other than that memory failures do occur, and that ECC fixes almost all of them. One of those studies is here: http://www.cs.toronto.edu/~bianca/paper ... rics09.pdf
I don't have a NAS with ECC memory myself. FWIW bit rot has never been reported in my RN102 or my RN202, and I don't have any clear evidence of data corruption in any of my NAS from any cause.
Me either. Though truthfully I don't use iSCSI at all.
Keeping your media in ordinary shares gives you more options. For instance you can also play your media directly from the NAS from another PC or media player (iSCSI is designed for only one client at a time). You can even run plex on both the NAS and the Mac Mini (sharing the same media library).
1) So little incentive for small office home to invest in ECC memory I guess? I work in the life science sector and people working with databases for genetic material etc are usually very keen to ensure ECC memory and often use ZFS systems. But I have not much of this kind of data on my desk.
2) iSCSI - can you sketch roughly what the main problem is with letting an aged Mac mini stand for the CPU-power and ReadyNAS 3-series for the storage of media using Plex? Much better to let ReadyNAS be the Plex-server in a stand alone way?
3) Virtual box - so 32-bit VMs would work I understand your comment as. Good enough for me. Just want to be able to learn the technique and see if can be useful from a central NAS in some way. For heavier work I naturally run VirtualBox locally on my laptop or so.
1) So little incentive for small office home to invest in ECC memory I guess? I work in the life science sector and people working with databases for genetic material etc are usually very keen to ensure ECC memory and often use ZFS systems. But I have not much of this kind of data on my desk.
Some high-end workstations support ECC memory, though generally it is a server feature. It does cost more, so Netgear choose to use it only on the high-end NAS. So you'd need an RN516 to get it (RN700 has it as well).
janpeter wrote:
2) iSCSI - can you sketch roughly what the main problem is with letting an aged Mac mini stand for the CPU-power and ReadyNAS 3-series for the storage of media using Plex? Much better to let ReadyNAS be the Plex-server in a stand alone way?
You can do this. We are just saying that iSCSI isn't the way we would do it. You can set up Plex to access the media even if it is in a normal share. Just mount the share on the mini.
FWIW Plex on my RN200 works quite well, and can transcode 720p acceptably. I suspect the 300 would be at least that good, though the 500 would be better still (giving you a second reason to go for it).
I work in the life science sector and people working with databases for genetic material etc are usually very keen to ensure ECC memory and often use ZFS systems.
Our ReadyDATA devices use both ECC memory and ZFS. However you can't run Plex Media Server on a ReadyDATA. I think a ReadyDATA would probably be overkill for your media collection.
The ReadyNAS 516 and 716X both have ECC memory. It is nice to have.
janpeter wrote:
3) Virtual box - so 32-bit VMs would work I understand your comment as.
Yes, however with just 2GB RAM in the 300 series system you are going to be quite limited in what you can run.
I tend to relate ECC memory to how OS6 deals with data integrity when backup of the NAS is done to an external HD or another NAS. I read around page 220 and forwards in the OS 6 manual but I cannot find the information I want.
1) Does OS6 provide possibility to format an external HD in BTRFS and in this way ensure that data integrity is kept?
2) Does the backup system perhaps need to be under OS6 also in order to get the whole system NAS and its regular backup procedure to be safe against bit-rot?
3) Is ECC memory a key component to make the system NAS and regular backup safe against bit-rot?
4) Is there other reasons than to avoid bit-rot that motivates investment in ECC memory?
I think its all been said. Memory errors can result in server crashes and unpredictable behavior. Sometimes that unpredictable behavior corrupts the data (and bitrot protection will almost never prevent it). ECC memory basically eliminates the threat of memory errors having those bad consequences. Whether its worth the money depends on how big you think the risk is, and how big the economic consequences of the unpredictable behavior are.
ECC isn't coupled with bitrot protection. Servers all over the world use it, and most of them aren't running btrfs or zfs.
janpeter wrote:
1) Does OS6 provide possibility to format an external HD in BTRFS and in this way ensure that data integrity is kept?
No it doesn't. And Netgear's bitrot protection depends on RAID, which isn't on the external disk.
janpeter wrote:
2) Does the backup system perhaps need to be under OS6 also in order to get the whole system NAS and its regular backup procedure to be safe against bit-rot?
There are maybe two questions here. The backup system does not need to be OS6. And the NAS does not protect the backups from bitrot. I suggest multiple backups myself.
janpeter wrote:
3) Is ECC memory a key component to make the system NAS and regular backup safe against bit-rot? 4) Is there other reasons than to avoid bit-rot that motivates investment in ECC memory?
I would like to change focus on my last question from ECC memory to integrity of backup which is the key question. And sure I admit I do not understand ECC-memory and BTRFS well enough and may questions maybe obscure.
I do understand that BTRFS is very useful if you interact with the NAS to some extent, but too much then you get disk fragmentation. Right now I am more interested in the low frequency of interaction that you have for an archive, i.e. mainly reading files and occasionally adding files in accumulative way. In this context bit-rot protection (BTRFS and raid-configuration) gives added value especially if we consider storage over long time, say 5-10 years and more.
Central to long term archive is of course to have an external backup procedure that maintain the data integrity you have in the primary storage, the NAS with BTRFS. I see some alternatives with: external HD, another NAS (for me older ReadyNAS with “OS4” and EXT3/4), or even a newer NAS under OS6. These different backup solutions have different levels of long time security of data integrity. But right now I think just of the short term backup-procedure.
1) When a file is backed up externally, what ensure that the backup is correct and that a recovery will give back the correct file?
a) I think of incremental backup. I thought that the use of checksum that can be enabled in OS6 and BTRFS also could be used somehow for external backup. I guess that if you do backup to another ReadyNAS under OS6 you might be able to exploit some usage of these checksums to ensure that the (incremental) backup/recovery is correct. Is that true?
b) Use of checksums (or something else) perhaps can be used even if backup is made to a non-BTRFS media?
c) Is here some difference in this aspect using the standards built in backup procedure in OS6 compared with for instance rsync or some other backup-software that can be used?
2) In this context of making backup/recovery I guess that ECC-memory provides an extra level to ensure data integrity when doing backup/recovery - or do I misunderstand this?
3) Would you say that ECC-memory is mainly interesting for intensive use of NAS as server rather than the low intensity use of NAS as archive?
Thus, as a user I wold like to be able to say that once the file is in the archive NAS then data integrity will be kept both at the primary NAS, during backup/recover to external site, and at the backup site.
---
See that in the new FORUM it is my responsibility to close the discussion and I think we are soon there.
I do understand that BTRFS is very useful if you interact with the NAS to some extent, but too much then you get disk fragmentation. Right now I am more interested in the low frequency of interaction that you have for an archive, i.e. mainly reading files and occasionally adding files in accumulative way. In this context bit-rot protection (BTRFS and raid-configuration) gives added value especially if we consider storage over long time, say 5-10 years and more.
If this is your main use, then you won't see much (if any) disk fragmentation with btrfs.
1) When a file is backed up externally, what ensure that the backup is correct and that a recovery will give back the correct file?
I've never seen a case where a backup / restore failed silently - usually there's an error reported (either by the disk or by the backup utility). That is not to say that "silent bitrot" is impossible. Just that if it has happened to me I didn't discover it later.
This is obviously very important to you. So I'd suggest that you put in place your own methods of validating data integrity, and not rely exclusively on the file system or RAID.
There are tools that verify when you copy (teracopy is one example). Ensuring that the restore done years later gives back precisely the original file likely requires saving an independent checksum to verify. I do this manually with some media files (creating an SFV file that checksums all the files in the folder).
a) I think of incremental backup. I thought that the use of checksum that can be enabled in OS6 and BTRFS also could be used somehow for external backup. I guess that if you do backup to another ReadyNAS under OS6 you might be able to exploit some usage of these checksums to ensure that the (incremental) backup/recovery is correct. Is that true?
Well, the checksum is verified when you copy to the backup. But the checksum is purely local, so another ReadyNAS doesn't see it, and cannot use it. BTRFS isn't supported right now on the external USB drives, so the checksum isn't computed on the USB drive either.
b) Use of checksums (or something else) perhaps can be used even if backup is made to a non-BTRFS media?
As I mentioned above, there are tools available that can do this. But the ReadyNAS doesn't (and no other NAS does either).
c) Is here some difference in this aspect using the standards built in backup procedure in OS6 compared with for instance rsync or some other backup-software that can be used?
I think I addressed this above.
2) In this context of making backup/recovery I guess that ECC-memory provides an extra level to ensure data integrity when doing backup/recovery - or do I misunderstand this?
3) Would you say that ECC-memory is mainly interesting for intensive use of NAS as server rather than the low intensity use of NAS as archive?
I wouldn't say that ECC is related to intensive use. ECC-memory increases reliability for both applications. Clearly that's a particularly important aspect for you, so you should probably consider a NAS that supports it.
Thus, as a user I wold like to be able to say that once the file is in the archive NAS then data integrity will be kept both at the primary NAS, during backup/recover to external site, and at the backup site.
My main advice is that there is always some risk, and at some point you need to accept it. I think you will need to put some of your own measures in place in order to help ensure that your own data integrity threshold is met.
My own approach is multiple copies (I maintain 2 local independent backups of all data, and a fourth copy on a cloud backup server). That doesn't alert me if one copy is somehow corrupted, but it does give me some recovery options if I discover that down the road. I'm not saying that will work for you. But you'll need to sort out what measures are enough.
From you answer it seems that NAS external backup is the weak link for ensuring data integrity, although at this forum key people iterate the need for external backups.
When I download larger software packages (eg python etc) I have encountered the advice
to calculate checksum of the downloaded file and compare with what it should be.
And I have encountered the standard MD5 (and easy to check on a Mac in the terminal).
Just thought that this use of checksums likely was automated in more serious backup software. Therefore I wondered if Netgears OS6 use that or similar.
For OS6 I also expected some emphasis on ensuring data integrity when say backup to another NAS with OS6 at least, but maybe not there yet.
Could also be that all backup software do some kind of checksum checks etc, I just do not know.
Of course check of checksums is not perfect either.
---
Part of my archive is a collection of 50 000+ photos. At least two photos are found corrupted after 4 years of raid1, but I am not sure the corruption happened as bit-rot. At least for one of the photos am pretty sure it happened during file transfer at an early stage. For the other photo I found the backup also corrupted, but backup was made at some later stage, and no backup of original, only of NAS. My concern it to minimize this kind of irritations.
From you answer it seems that NAS external backup is the weak link for ensuring data integrity, although at this forum key people iterate the need for external backups.
I wouldn't call it the "weak link", as there are other steps that are equally vulnerable. For instance, the original copy of the data from your PC to the NAS.
And yes, we do repeat the need for backing up the NAS. RAID arrays and servers do eventually fail, and sometimes they fail in ways that lose data.
If you simply want to verify your photos, you can use something like quicksfv on each main folder. If you want some ability to repair lost data independently from the NAS or backup software you could use quickpar.(also on each folder, not on each photo).
