TR Forums

As the subject reads, I'm considering replacing my WHS2011 installation with W8Pro. At this point I am going to be reinstalling my server OS regardless - either with WHS2011 (again), or Windows 8(.1) Pro. I'm trying to decide what will be best (and require the least maintenance) for the next few years to come. I'm leaning toward W8Pro.

I have been running WHS since v1 in 2008 (so about 6 years in total). Lately I've been getting seriously frustrated with WHS2011 as the connector software is acting up (which makes it a real pain trying to figure out which drives/folders have sufficient free space for large copies) - and a lot of other stuff has been sluggish and annoying too (I won't bore you).

Anyway, I use my Home Server primarily for file serving and media serving (I DO NOT use it to backup client systems, and I no longer use it for remote access). With WHS2011 I'm currently not using any Drive Extender add-in functionality (When the connector software actually worked correctly, I actually preferred not having DE in many ways, because it allowed me to choose which drive I put certain files on - Documents, Backups, and DVD Iso's on 5k3 drives, high bit rate media files on faster 7k2 drives). And I could also spread around items that I know may be accessed simultaneously to be sure they were on different drives.

However, the idea of these "Storage Spaces" on W8 is intriguing to me. If I understand it correctly, I could make a pool of 7k2 drives, and put a single folder - for example "HD Media" - on that drive, and then it'd save me the trouble of having "Recorded TV" 1-5 folders all with different series in them. I'm guessing there's no way to define what drive a specific file (or group of files) is sent to and/or to easily see which files are where (as far as specific drive)?

Also, right now I have a "Server" link on my desktop, that links to all of the root of my Server's network shared folders. If I had different Storage Spaces/Pools in W8, could I still make one single "Shared Folders" type directory to easily see all the network shares on my Windows 8 "Server"? (like a single view that had the network shared folders for different pools like: Documents, Business, Recorded TV, Movies, Pictures, Etc) - Maybe using libraries?

The last thing is, I know If I go with Windows 8 I will lose the proper user account controls (as far as assigning folder access permissions, and remote access stuff) - but I think I can live with using the built in sharing and security settings to give users specific rights to root folders in a drive pool, and giving guests read and/or no access. Am I correct?

Aside from my questions, does anyone have any other opinions about why I should go one way or the other? Thanks!

Are you married to having a windows server? Seems like everything that you want to do (and more) could be done with FreeNAS.

http://www.freenas.org/

Deanjo wrote:

Are you married to having a windows server? Seems like everything that you want to do (and more) could be done with FreeNAS.

http://www.freenas.org/

Good point. Back when I originally chose WHS I had considered other options, FreeNAS, unraid, and a few others - but none seemed as appealing for what I was doing back then.

I'll fire FreeNAS up in a VM and see how I like it (now). Thanks for the idea.

Well, I checked out FreeNAS - it has definitely come a long way since I last considered it. However, I have to say, reading the manuals, as well as a few forums over at their site, has me pretty reluctant to give it a real try with my data. My main holdup is that they seem to strongly warn against using the ZFS based file system without ECC RAM (which I don't have, and don't plan on buying new hardware). I've also found a lot of issues regarding ZFS file system corruption when not paired with ECC, and it's virtually impossible to mount a drive and recover data from ZFS, so that's pretty scary to me. Then they also seem to discourage using older UFS file system - and suggest that if you do use UFS, that you do not create volumes greater than 5TB (of which I would have at least a few).

I have unlimited backup with Zoolz ($2/mo) for all of my non-replaceable personal data and work files - but I don't want to have to use it. And I really don't want to face the loss of entire drive pools worth of media. With NTFS I've had drives go bad, and files corrupt - but nothing where I couldn't pull drives and get most of the data back off of them individually...

So, still undecided.

Using storage spaces, in particular with the "Simple" type, is going to be the same if not worse than the situations that you described. Since you came from WHS2011, you already knew Microsoft started down the path of block-based spanning/striping instead of file-based spanning of the v1 DE days. Storage spaces can be considered the next evolution of the tech. For "simple" type, your files will be striped and spanned across the drives making up the pool, similar to RAID0. For "mirror" type, you get 2 (or 3, depending on 2-way or 3-way mirror) copies of the same stripe stored on 2 drives (you don't know for sure which drive). For "parity", you are looking at striping and parity similar to RAID5+ schemes. This all means if you need to reconstruct the array, you need to stick with the tools provided in Windows. Unlike WHSv1 you just merge the folders together from the different drives. NTFS does not magically help you with restoring blocks as storage space functions at a lower level I believe. What I am trying to say is that for truly irreplaceable data, you do need a backup, either on an external drive or online.

For shares, once you have the storage spaces created, when you create a volume from the spaces, you have the option of assigning a drive letter or mounting to a folder. So you can create a folder on C:, and then share that out while mounting all your storage space volumes onto that folder. I think that should work.

@Flying Fox, Thank you for the detailed explanation - I wasn't quite sure how the storage pools were working. Now that I know, and I went and read a few more reviews about them - I don't think Windows 8 will make a lot of sense either.

I was really liking the idea of FreeNAS until I read a lot of negative stuff about ZFS, and the limitations of UFS - so I've ruled that out. However, I think that may have led me to Openfiler. It seems like Openfiler would easily meet my needs, and I like the greater reliability of its options for journaled EXT3 or XFS file systems. So, right now I'm leaning towards that. I plan to test it on a VM tonight.

As I mentioned my absolutely-cannot-lose data is backed up in multiple places (including by offsite service) - but for the rest of it (mostly lots of media) I would like to assume no greater risk than I already do. And from what I've read about FreeNAS/ZFS, it just seems like that would be riskier, as would using Windows 8 Storage Spaces (and it seems storage spaces take a pretty big performance hit too), and WHS is already showing signs of its age for me (so I prefer getting off the WHS platform sooner, rather than closer to its EOL in 2020 when it will be just plain dead).

So Openfiler is my current leaning, any comments/suggestions?

I very recently made the switch from WHS V1 to Windows 8.1 + storage spaces. I was using my old WHS box as a NAS + backup + downloader. The new box does all those things just as well.

But for the rage inducing "thumbs.db" issues, it works quite well. I don't do a lot of management on it. So I am not familiar with how easy/hard it is to move files around share. I just have one big 60TB(I believe this is the biggest it will do) storage space. I keep adding drives as the pool fills up.

FlexRAID is an option too:

http://www.flexraid.com/2013/10/04/tabl ... le-system/

While FreeNAS with ZFS may not be the best option for your current configuration you may still want to read this article Bitrot and atomic COWs: Inside “next-gen” filesystems. Gives one a different perspective on backups.

HorseIicious wrote:

Well, I checked out FreeNAS - it has definitely come a long way since I last considered it. However, I have to say, reading the manuals, as well as a few forums over at their site, has me pretty reluctant to give it a real try with my data. My main holdup is that they seem to strongly warn against using the ZFS based file system without ECC RAM (which I don't have, and don't plan on buying new hardware). I've also found a lot of issues regarding ZFS file system corruption when not paired with ECC, and it's virtually impossible to mount a drive and recover data from ZFS, so that's pretty scary to me. .

Don't scare me now. My Synology DS712+ NAS uses ZFS in raid1

I didn't like the limitations on storage spaces - the fact that the total space is limited by the size of the smallest drive (times multiplier) and you have to destroy the space in order to remove a disk. Plus your data is no longer accessible directly on disk as NTFS partitions.

Have you tried DrivePool or DriveBender? They are essentially what DE would have been - and they have versions designed for Win 8 as well.

HorseIicious wrote:

However, I think that may have led me to Openfiler. It seems like Openfiler would easily meet my needs, and I like the greater reliability of its options for journaled EXT3 or XFS file systems.

I don't recommend Openfiler. It might be a dead project, and it's rather odd since it's based on rPath. I replaced the Openfiler boxes with CentOS boxes, and I've been much happier.

FreeNAS (http://www.freenas.org/), NAS4Free (http://www.nas4free.org/), or OpenMediaVault (http://www.openmediavault.org/) are the three I would recommend.

anotherengineer wrote:

HorseIicious wrote:

Well, I checked out FreeNAS - it has definitely come a long way since I last considered it. However, I have to say, reading the manuals, as well as a few forums over at their site, has me pretty reluctant to give it a real try with my data. My main holdup is that they seem to strongly warn against using the ZFS based file system without ECC RAM (which I don't have, and don't plan on buying new hardware). I've also found a lot of issues regarding ZFS file system corruption when not paired with ECC, and it's virtually impossible to mount a drive and recover data from ZFS, so that's pretty scary to me. .

Don't scare me now. My Synology DS712+ NAS uses ZFS in raid1

In general you shouldn't use ZFS without ECC. It heavily relies on one thing to be error-free: memory (CPU, too, but that's beside the point). Because so many structures, indexes, etc are kept in memory it's definitely best practice to run ECC.

Sure, you could never have an issue, but a bit or two flipped in the wrong places can cause catastrophic data loss with any next-gen filesystem (ZFS included).

Igor_Kavinski wrote:

FlexRAID is an option too:

http://www.flexraid.com/2013/10/04/tabl ... le-system/

That's my current storage server solution. Server 2008R2 (might as well be 7 for all I ended up using), with FlexRAID on everything but the OS disk. Works pretty well, I'm a fan.

Might be going to Debian with the Linux release of FlexRAID soon. I just hope it (FlexRAID) works as well there.

Waco wrote:

In general you shouldn't use ZFS without ECC. It heavily relies on one thing to be error-free: memory (CPU, too, but that's beside the point). Because so many structures, indexes, etc are kept in memory it's definitely best practice to run ECC.

In all fairness, that concern lies with every filesystem out there. Bad ram is going to be a concern no matter what the filesystem. ZFS however offers its best protection when utilized with ECC.

anotherengineer wrote:

Don't scare me now. My Synology DS712+ NAS uses ZFS in raid1

Everything I read was related specifically to ZFS+FreeNAS, so it may not apply to your situation. Also a straight Raid1 mirror is a lot less risky than what I was considering with the zpool, with data striped across many drives with parity (I just don't have enough space for a 1:1 mirror though)

bitcat70 wrote:

While FreeNAS with ZFS may not be the best option for your current configuration you may still want to read this article Bitrot and atomic COWs: Inside “next-gen” filesystems. Gives one a different perspective on backups.

Yeah, the bit rot issue was one reason I was originally very interested in FreeNAS+ZFS once Deanjo mentioned it. Maybe in a few years it will be mature enough that I can justify using it. In the meantime I just don't feel like the risks outweigh the benefits.

Igor_Kavinski wrote:

FlexRAID is an option too:

http://www.flexraid.com/2013/10/04/tabl ... le-system/

Yeah, that is an interesting option, but I'm crazy about the price, and again the compromises in Raid-F vs tRaid.

Still have some thinking to do now...

Evaders99 wrote:

the fact that the total space is limited by the size of the smallest drive (times multiplier)

I don't think this is the case, any links?

Evaders99 wrote:

you have to destroy the space in order to remove a disk.

In my current tests, only simple spaces have that issue. For parity and mirror spaces, I can easily retire a disk and repair kicks in to rebuild the array. Sort of make sense as simple is really RAID0 and you can't just yank the drive out.

Evaders99 wrote:

Plus your data is no longer accessible directly on disk as NTFS partitions.

WHS 2011 already does that AFAIK, so? Most other solutions, like software/hardware RAID, are also block-based and have this same issue. If accessing disk directly as regular file system partitions is your thing, you may as well stick with WHSv1 with DE (which is what I am running, but I do want the ability to use 4TB disks so I am looking also).

Deanjo wrote:

In all fairness, that concern lies with every filesystem out there. Bad ram is going to be a concern no matter what the filesystem. ZFS however offers its best protection when utilized with ECC.

This. If you do want better reliability, you should build a system with ECC as the chance for us hitting bit errors are higher now that we are dealing much more data than before.

Regarding the performance problems with Storage Space, other RAID5 solutions are not fast either (of course, hardware RAID5 is better than SS by quite a bit). Based on people's testing, mirror (RAID10 like approach) seems to be ok performance wise. For Parity spaces, if you want decent performance, dedicated journal disks with write back caches (8.1 or 2012 R2) addresses it quite a bit. You do have to wonder the big media files will be read more than write, so how much write performance do you actually need?

I am partial to Windows-based solutions since I can run other apps on the server, but not everyone has this particular requirement.

I don't know how dated this blog is, maybe they've updated the features to Storage Spaces in 8.1?
http://helgeklein.com/blog/2012/03/wind ... ign-flaws/

Deanjo wrote:

In all fairness, that concern lies with every filesystem out there. Bad ram is going to be a concern no matter what the filesystem. ZFS however offers its best protection when utilized with ECC.

Not so much with "traditional" filesystems though. Even if the filesystem gets destroyed it's reasonably easy to recover files from busted NTFS, FAT, EXT[2/3/4] file systems. ZFS? Not so much due to how it allocates blocks and whatnot.

Unless I'm totally mistaken or something - in which case please set me straight!

Flying Fox wrote:

Evaders99 wrote:

Plus your data is no longer accessible directly on disk as NTFS partitions.

WHS 2011 already does that AFAIK, so? Most other solutions, like software/hardware RAID, are also block-based and have this same issue. If accessing disk directly as regular file system partitions is your thing, you may as well stick with WHSv1 with DE (which is what I am running, but I do want the ability to use 4TB disks so I am looking also).

In this respect, FlexRAID may be right up your alley. It's RAID-over-filesystem, so file-based, it has no per-disk limits beyond what the OS itself has, and does give a lot of flexibility.

I personally chose it because I was an enthusiastic user of WHS v1 and specifically DE, and FlexRAID reproduces that functionality perfectly, with some added bonuses. I don't have it configured at the moment, but I can add a disk of the largest unit size, and enable parity and rebuild for my array. That's pretty high on my "would be appreciated" list.

Flying Fox wrote:

Evaders99 wrote:

the fact that the total space is limited by the size of the smallest drive (times multiplier)

I don't think this is the case, any links?

I've read what Evader99 is talking about in a few places too - regarding total space being hindered (not necessarily limited) by the smallest drive size - also big issues with adding a drive to a pool (and being able to use all of the new space) after the others are mostly full.

Flying Fox wrote:

Evaders99 wrote:

Plus your data is no longer accessible directly on disk as NTFS partitions.

WHS 2011 already does that AFAIK, so?

No, WHS2011 drives (at least how I have mine set up) are each simple NTFS logical/physical drives. I can pull any one of the 10+ drives in my server, throw them in my main workstation, and read all of the data that's on them, and put them back in the server, no problems.

Waco wrote:

Deanjo wrote:

In all fairness, that concern lies with every filesystem out there. Bad ram is going to be a concern no matter what the filesystem. ZFS however offers its best protection when utilized with ECC.

Not so much with "traditional" filesystems though. Even if the filesystem gets destroyed it's reasonably easy to recover files from busted NTFS, FAT, EXT[2/3/4] file systems. ZFS? Not so much due to how it allocates blocks and whatnot.

Unless I'm totally mistaken or something - in which case please set me straight!

I understand this the same way as Waco... ZFS seems to be very very vulnerable (when not using properly configured enterprise class hardware) to a few bad bits destroying an entire pool of data.

As it stands now I think I'm between just reinstalling WHS2011 or going with Openfiler (which I just set up in a VBox, and was impressed with how easy and simple it was to get going). Unfortunately it's looking like one of these two options is probably going to be my best bet (unfortunate because they're both dated). I'm just unwilling to use any file system that is vulnerable to losing an entire disk worth of data due to a few bad bits. And quite honestly, for my uses, I don't even like the idea of RAID anything anymore - (for me as a home user). The disks are just too large - I mean how long would it take to rebuild a lost 4TB disk anyway? And then we're talking about 12TB+ pools - what could go wrong?haha

HorseIicious wrote:

As it stands now I think I'm between just reinstalling WHS2011 or going with Openfiler (which I just set up in a VBox, and was impressed with how easy and simple it was to get going). Unfortunately it's looking like one of these two options is probably going to be my best bet (unfortunate because they're both dated).

Flatland_Spider wrote:

OpenMediaVault (http://www.openmediavault.org/)

It's based on Debian, and it's actively developed. Plus, it has more features then Openfiler ever did.

Waco wrote:

Deanjo wrote:

In all fairness, that concern lies with every filesystem out there. Bad ram is going to be a concern no matter what the filesystem. ZFS however offers its best protection when utilized with ECC.

Not so much with "traditional" filesystems though. Even if the filesystem gets destroyed it's reasonably easy to recover files from busted NTFS, FAT, EXT[2/3/4] file systems. ZFS? Not so much due to how it allocates blocks and whatnot.

Unless I'm totally mistaken or something - in which case please set me straight!

Yes with bad bits you could potentially lose a pools of data with ZFS, but then again, ZFS is more likely to pick up on bitrot like failures while another filesystem such as NTFS (and btw NTFS since Server 2008/Win 7 has had "self healing" features that are just as susceptible to data corruption due to memory errors and gets worse if using something like a raid configuration instead of a spanned volume) is not likely to and merrily continue writing away corrupted data. Also keep in mind that no matter what storage solution you choose, SISO kicks in. So if that non-ECC workstation that you are copying data from has bad data to begin with, no filesystem is going to correct that data on the server but you will have a perfect copy of the corrupted data.

Flatland_Spider wrote:

OpenMediaVault (http://www.openmediavault.org/)

It's based on Debian, and it's actively developed. Plus, it has more features then Openfiler ever did.

Thanks for that, Sorry I somehow overlooked that link in your post - this thread is getting longer than I ever expected. Looks very good. I am going to fire that up in a VM and see how I like it.

Evaders99 wrote:

Have you tried DrivePool or DriveBender? They are essentially what DE would have been - and they have versions designed for Win 8 as well.

Also somehow overlooked this post. I hadn't thought about using one of those with Win8. Might have to consider that too.

Thanks to everyone for all of the great input and suggestions.

HorseIicious wrote:

Flying Fox wrote:

Evaders99 wrote:

the fact that the total space is limited by the size of the smallest drive (times multiplier)

I don't think this is the case, any links?

I've read what Evader99 is talking about in a few places too - regarding total space being hindered (not necessarily limited) by the smallest drive size - also big issues with adding a drive to a pool (and being able to use all of the new space) after the others are mostly full.

Most complaints were during beta time before 8/2012 RTM. For 8.1/2012 R2:
- No automatic rebalancing is still true (you can force a rebuild/repair by adding a new drive and try yanking an old one)
- The smallest drive limitation has been fixed in 8 RTM (they were talking in the context of the beta I believe).
- File-based is simply not going to happen (the blog author is a DE lover, so do I, but have to get over that).

HorseIicious wrote:

No, WHS2011 drives (at least how I have mine set up) are each simple NTFS logical/physical drives. I can pull any one of the 10+ drives in my server, throw them in my main workstation, and read all of the data that's on them, and put them back in the server, no problems.

Right, I forgot the original story. They tried doing the block-based "new DE" but could not get it to work properly, so they dropped "DE" completely. So yes, you are still on a file-based approach. I have to admit I don't run into the flaws of DEv1 myself, so I am happy to remain on WHSv1 if not for the capacity problem. It saved my behind once when a drive failed. Other than the non-duplicated files, rebuilding the array resulted in only 1 small corrupted file. I just needed to merge the folders and my array is more or less back to normal.

I may have to look at FlexRaid myself, though not sure if I want to pay extra on top of the OS. I am already thinking Journal+WBC on SSD, plus a single parity pool. The rule seems to be "only replace/upgrade with larger disks" and life should be good. Looking at C2xx motherboards now paired with a cheap ass i3/Pentium and ECC.

Flying Fox wrote:

I may have to look at FlexRaid myself, though not sure if I want to pay extra on top of the OS. I am already thinking Journal+WBC on SSD, plus a single parity pool. The rule seems to be "only replace/upgrade with larger disks" and life should be good. Looking at C2xx motherboards now paired with a cheap ass i3/Pentium and ECC.

Your other approach is good, too, but as a deeply cheap mofo when it comes to OS/software, I will freely volunteer that I paid the full asking price for the full-fat version of FlexRAID (something around 80$ IIRC) and have been happy with the expense. Having the freedom to take it with me to a different server OS in the future and having lots of upgrade options moving forward appeals to me.

Deanjo wrote:

Yes with bad bits you could potentially lose a pools of data with ZFS, but then again, ZFS is more likely to pick up on bitrot like failures while another filesystem such as NTFS (and btw NTFS since Server 2008/Win 7 has had "self healing" features that are just as susceptible to data corruption due to memory errors and gets worse if using something like a raid configuration instead of a spanned volume) is not likely to and merrily continue writing away corrupted data. Also keep in mind that no matter what storage solution you choose, SISO kicks in. So if that non-ECC workstation that you are copying data from has bad data to begin with, no filesystem is going to correct that data on the server but you will have a perfect copy of the corrupted data.

Oh I'm not arguing against it at all, my personal NAS runs ZFS and I use it daily at work. You just have to have good hardware (read: ECC) to start with or it's a bit of a lost cause IMO.