plans for home file server - to raid or not to raid?

[poofyhairguy]

Be careful settling on RAID based options HardTech. Everyone around here pushed RAID options when I was looking for a media server, and I found them all to be a poor fit.

Why RAID 5 or 6? Both are designed for business use, and require expensive and overkill RAID cards to get maximum performance.

Really for a media server there are two good options: WHS and Unraid

You go for WHS if you want to do many other things with your media server. With the WHS pluggin community you can quickly have the box not only serving media, but also downloading torrents and steaming them to your PS3.

You go for Unraid if you want maximum storage. Unraid gives you every advantage of JBOD-add any disk you want to the array of any size/speed you want and get full use of it- with the bonus of RAID 5 like single disk parity. Since it doesn't stripe the data its not nearly as dangerous as RAID 5- if three disks die the maximum you can lose is the data on those three disks, unlike RAID 5/6 where you are toast. In my experiance my Unraid box easily maxes my gigabit network- anyone that thinks they need more performance for a home server is kidding themselves.

So since there are the same price, which one to pick?

Pick WHS if you prefer to have your server multitasking, and if you entire environment is Windows anyway (which you said its not). The WHS community's pluggins can't be beat. But don't pick WHS if you are after maximum server size- instead of parity you basically get duplication which uses (wastes IMHO) a lot more disk space than Unraid.

You pick Unraid if you want a NAS and nothing but a NAS. Its single disk parity system (up to 20 drives!) makes way better use of space than WHS. But the flipside to Unraid is that is it based on the extremely primitive Slackware Linux. I am a huge Linux guy myself (used Ubuntu since 2004), but even I don't have patience to mess with Slackware. The community has done a good job expanding Unraid with extra services, but honestly its a PITA compared to WHS to get your Unraid box to multitask.

Personally, my previous JBOD server was multitasking so when I decided to move on WHS seemed like an obvious choice. In the end though I went with Unraid because I preferred all my data to be semi-protected not just the data I chose to be duplicated. My solution is to build one of my HTPC frontends into a monster so it can do the AirVideo/TorrentDownloading/etc. that my server formally did.

If you want free options, there is Flexraid. I don't have much experience with it personally and I don't like that it doesn't have live parity.

By the way, I had the same basic thread a few months ago to learn all this stuff. Here it is for a reference:

http://forums.anandtech.com/showthread.php?t=2071210

Last thing: What about RAID 1?

Sure go for it. If you are rolling in money enough that you can pay for twice as much storage as you expect to use, and you plan to use a Norco 20 bay case so that you have enough drive slots to grow to a decent size, then go for it. It is the safest option.

But personally media isn't THAT important. If you lose a little its not the end of the world. I wanted a little protection (say enough for one disk to die), I wanted to maximize my hardware space, and I wanted to be able to use the disks I wanted to. That solution is called Unraid. You should look into it. And before you ask: I ONLY use OSX and Unraid works great for me.

 

[pjkenned]

Again... single parity + large array = high data loss rate. Using consumer disks + single parity (raid 4 or 5) you have a >10&#37; chance to lose data in <1 year at something like 14 or 15 drives. Raid 6 20 drive <1 year failure rate is about .5% IIRC and Raid 1 is slightly lower. That is ONE year or less failure rate which is abysmal. When you have the ZFS option of FreeNAS/ OpenSolaris (or NAS distribution thereof), unRaid does not have the same error checking capabilities built in. Plus, WHS in theory has about Raid 1 data survival chances (assuming you are up to date on your service packs since it was not so good a few years ago), potentially better since you can get duplicated data without immediately inserting a disk if there is enough free space available.

 

[RebateMonger]

There's no perfect AND inexpensive solution to storing and securing large amounts of data. The "best" solution depends on the individual's needs. Most home users have data that's "Can't possbily afford to lose it" and data that's "If I lost it, I'll be sad, but I can get most of it back and it won't kill me". Fortunately, the "Can't possibly afford to lose it" data is usually fairly small for home users. So the "best" solution may be a combination of offsite (disk or online) backups, local backups, disk redundancy, or, maybe for some data, nothing at all.

Unless uptime is a big issue (which it isn't for most home users), I prefer full independent backups over redundancy (such as RAID or WHS' folder redundancy). It's REALLY hard to lose a significant amount of data when you've got separate copies of the data elsewhere.

RAID 1 and WHS' folder redundancy are probably about equal in safety. I haven't decided whether it's safer to go with RAID 5 or no redundancy at all. A CAREFULLY MONITORED RAID 5 array can be "OK". But if something goes wrong with RAID 5, you lose everything. With non-RAID disks, you lose only the content of the failed disks.

 

[Emulex]

openfiler can't touch esxi(free) plus hp lefthand vsa (trial mode after 60 days is still solid).

use hardware raid please. an old perc/cerc/p400/p410 with fanout

 

[Voo]

Unless uptime is a big issue (which it isn't for most home users), I prefer full independent backups over redundancy (such as RAID or WHS' folder redundancy). It's REALLY hard to lose a significant amount of data when you've got separate copies of the data elsewhere.

Wholeheartly agree with that, the performance advantage of RAID is usually completely unimportant, so you practically spend the same amount of money for less safety, greater uptime and a bit performance you won't notice in practice. And corrupted RAIDs are horrible to fix and rebuilding it after getting new drives also may not be as easy as thought..

Though even if I wanted RAID I don't see why I'd ever wanted to use RAID 5 instead of ZFS (RAID Z) - maybe it's just me, but it sounds like the better deal..

 

[sub.mesa]

There's no perfect AND inexpensive solution to storing and securing large amounts of data. The "best" solution depends on the individual's needs.

Perfect is an absolute, but i'd say ZFS pretty much means you can store and forget, can use cheap hardware and also have good performance.

The big downside of ZFS is accessibility; it's simply not (properly) available on any mainstream OS like Ubuntu, Windows or Mac OSX. You would have to use base FreeBSD, OpenSolaris UNIX operating systems, or derivatives like FreeNAS, Nexenta and my Mesa web-interface.

The hopes i had was that building a proper web-interface by creating something that is very easy, comfortable to use while still giving you the full power of FreeBSD and a stable implementation of ZFS (v14). I still understand it's long step from something like WHS; but to anyone that would want a more reliable and powerful way of storing all his home files, i believe ZFS is certainly an option to consider.

Most home users have data that's "Can't possbily afford to lose it" and data that's "If I lost it, I'll be sad, but I can get most of it back and it won't kill me". Fortunately, the "Can't possibly afford to lose it" data is usually fairly small for home users. So the "best" solution may be a combination of offsite (disk or online) backups, local backups, disk redundancy, or, maybe for some data, nothing at all.

Agree; having a backup is so valuable and especially with varying quality of RAID it makes sense to always invest in a backup. That doesn't mean you can't do both, though. For that reason, i use both a main ZFS fileserver and a backup one; the backup server is even cheaper, has no SSDs for L2ARC cache and uses all my older disks, and usually is offline except at night.

But the ability to correct corruption or bit errors on the fly without any maintenance by the user is a great win, so i would recommend both redundancy and backup if at all possible. For example, you could have a much smaller backup even a single 2TB drive for your most important things, and just rely on redundancy for the lesser important things; assuming that fits within your situation.

 

[RebateMonger]

I still understand it's long step from something like WHS; but to anyone that would want a more reliable and powerful way of storing all his home files, i believe ZFS is certainly an option to consider.

Microsoft's proposed "Vail" release of Windows Home Server (WHS2) has added some error checking to the storage system, much like ZFS does. But, like everything else, there's some negatives, too. Last I looked at it, Vail was basically going to REQUIRE folder redundancy in order to ensure that a single-disk failure wouldn't cause loss of many files across multiple disks. The original WHS basically uses something similar to mirroring for redundancy. But the new WHS file model is very close to disk "striping", with its associated risks.

None of this is finalized yet. Hopefully we'll eventually end up with a good, safe, and easy-to-use way to store "very large" amounts of data. We'll see.

 

[sub.mesa]

To be honest, Windows could have used a great filesystem that revolutionized the way files were stored; replacing the dated directory-structured filesystems. As far as i know, WinFS was supposed to be just that, and it would be THE killer feature in Microsoft Vista initially, before the project was marginalized (WinFS would run on top of NTFS) and later canceled altogether.

And the fact that Mac OSX didn't opt to use ZFS as base filesystem in their OS is also a great miss, but likely has to do with uncertainties after Oracle's acquirement of Sun, and i don't think Steve Jobs wants to be at the mercy of Larry Ellison for something important as your base Filesystem. Apple likes to stand on their own feet.

Likewise, Microsoft has all the resources and opportunities to make a great filesystem for their users. Even if WinFS didn't work out in the past, they could form a team that likes to work on a modern multi-disk filesystem that could be rival advanced filesystems like ZFS and Btrfs.

However, until there is a reliable way of storing files on Windows, anyone who wants additional protection against corruption, wants high performance on cheap hardware and doesn't like buying expensive "RAID edition" harddrives which support TLER/CCTL, then ZFS could be a viable alternative to those windows users, perhaps until a better solution from Microsoft presents itself. In all honesty, i haven't read much about Vail yet.

 

[HardTech]

Be careful settling on RAID based options HardTech. Everyone around here pushed RAID options when I was looking for a media server, and I found them all to be a poor fit.

Why RAID 5 or 6? Both are designed for business use, and require expensive and overkill RAID cards to get maximum performance.

Really for a media server there are two good options: WHS and Unraid

You go for WHS if you want to do many other things with your media server. With the WHS pluggin community you can quickly have the box not only serving media, but also downloading torrents and steaming them to your PS3.

You go for Unraid if you want maximum storage. Unraid gives you every advantage of JBOD-add any disk you want to the array of any size/speed you want and get full use of it- with the bonus of RAID 5 like single disk parity. Since it doesn't stripe the data its not nearly as dangerous as RAID 5- if three disks die the maximum you can lose is the data on those three disks, unlike RAID 5/6 where you are toast. In my experiance my Unraid box easily maxes my gigabit network- anyone that thinks they need more performance for a home server is kidding themselves.

So since there are the same price, which one to pick?

Pick WHS if you prefer to have your server multitasking, and if you entire environment is Windows anyway (which you said its not). The WHS community's pluggins can't be beat. But don't pick WHS if you are after maximum server size- instead of parity you basically get duplication which uses (wastes IMHO) a lot more disk space than Unraid.

You pick Unraid if you want a NAS and nothing but a NAS. Its single disk parity system (up to 20 drives!) makes way better use of space than WHS. But the flipside to Unraid is that is it based on the extremely primitive Slackware Linux. I am a huge Linux guy myself (used Ubuntu since 2004), but even I don't have patience to mess with Slackware. The community has done a good job expanding Unraid with extra services, but honestly its a PITA compared to WHS to get your Unraid box to multitask.

Personally, my previous JBOD server was multitasking so when I decided to move on WHS seemed like an obvious choice. In the end though I went with Unraid because I preferred all my data to be semi-protected not just the data I chose to be duplicated. My solution is to build one of my HTPC frontends into a monster so it can do the AirVideo/TorrentDownloading/etc. that my server formally did.

If you want free options, there is Flexraid. I don't have much experience with it personally and I don't like that it doesn't have live parity.

By the way, I had the same basic thread a few months ago to learn all this stuff. Here it is for a reference:

http://forums.anandtech.com/showthread.php?t=2071210

Last thing: What about RAID 1?

Sure go for it. If you are rolling in money enough that you can pay for twice as much storage as you expect to use, and you plan to use a Norco 20 bay case so that you have enough drive slots to grow to a decent size, then go for it. It is the safest option.

But personally media isn't THAT important. If you lose a little its not the end of the world. I wanted a little protection (say enough for one disk to die), I wanted to maximize my hardware space, and I wanted to be able to use the disks I wanted to. That solution is called Unraid. You should look into it. And before you ask: I ONLY use OSX and Unraid works great for me.

awesome post.. thanks a lot for your insights

What about something like FreeNAS? I'm going to be devoting most of this weekend to reading up on it and understanding exactly how it works, but at first glance it looks like a good replacement for WHS while using open source protocols.

 

[poofyhairguy]

What about something like FreeNAS? I'm going to be devoting most of this weekend to reading up on it and understanding exactly how it works, but at first glance it looks like a good replacement for WHS while using open source protocols.

FreeNAS is a FreeBSD distro designed for NAS purposes, and it does a good job of it. Next business NAS I make is based on FreeNAS.

The problem is that with FreeNAS you are mostly limited to the same RAID 1/5/6 options, which gets you back to the same issues. The "extra" thing FreeNAS offers is RAIDZ, which is basically a superior (technologically speaking) RAID 5/6 (it fixes many issues of the two). The problem with RAIDZ is that unlike WHS, Unraid, and RAID 5/6, you can't add any drives to a RAIDZ array.

Now if you plan to buy all your drives at once, and they all will be the same drive, RAIDZ is an incredible option that should be considered (just better than RAID 5/6 in everyway). As I said, its my next business server (a little 2TB thing).

The only problem with RAIDZ that I see outside its stated requirements is the fact that because all the drives are stripped, they are constantly in use (RAID 5/6 have the same problem). One thing I really like about WHS and Unraid is that they power down drives that aren't being used. I feel like running consumer level drives all the time kill them faster.

For this lack of striping you take a performance hit (I bet reads on a RAIDZ server are at least twice as fast as the WHS/Unraid), but as I said for a non-corporate gigabit network Unraid/WHS can still max out the connection. Also not having striping means each disk can be put into a different computer and read, something RAID 5/6/Z can't offer.

I kinda get where you are coming at looking at free options- I personally really hate to pay for software. But you have to consider that most massive storage solutions are business-centric as they are who mostly needs storage servers. Most of the rest of the home server market is taken by the NAS market (like drobo). This leaves only a small niche of people who are willing to build home servers, and for that niche the only players expect payment for their specialized solutions.

Hope I helped, if not please keep posting.

 

[poofyhairguy]

Again... single parity + large array = high data loss rate. Using consumer disks + single parity (raid 4 or 5) you have a >10&#37; chance to lose data in <1 year at something like 14 or 15 drives. Raid 6 20 drive <1 year failure rate is about .5% IIRC and Raid 1 is slightly lower. That is ONE year or less failure rate which is abysmal.

Which is how I learned the hard way that JBOD servers are really not acceptable.

When you have the ZFS option of FreeNAS/ OpenSolaris (or NAS distribution thereof), unRaid does not have the same error checking capabilities built in.

Yeah, but I can't grow a RAIDZ array so in order to use RAIDZ on my 16 drive server I am building I would have to buy all the drives at one (can't happen). And with RAIDZ even if I have two parity drives, if I lose three drives (because consumer drives suck as you stated above) then all my data is gone. Since Unraid does not stripe the data if I lose three drives the maximum I can lose is the data on those three drives.

I figure if I am going to have bad enough luck for two drives to die during a rebuild, misery loves company and it could easily be three. So instead of trying to hot swap myself to a point where I am basically running a RAID 1, I instead chose to go with a solution that gives me a certain degree of protection (one drive failure), but won't flush all my data down the tubes in a worst case scenario. Plus I like only having one parity drive as it leaves more sata ports for actual storage.

In fact, I would argue that my Unraid server would have a WAY lower chance of massive failure than a regular RAID Z server. Why? Because with RAIDZ you have to buy all the drives at one- all the same kind of drive. That means you are buying maybe all out of one "batch," so if there is a manufacture problem you can easily lose three drives or more (bye bye data).

With my Unraid server, I can buy the disks one at a time FROM MANY DIFFERENT VENDORS. My current build has a WD Green drive, a Samsung drive, a Hitachi drive, and a Seagate drive. It is literally impossible for me to get bit by a bad run of production at a single factory. Of course, WHS has the same benefit which is why I consider the two to be the only acceptable options for a home server.

 

[sub.mesa]

Now if you plan to buy all your drives at once, and they all will be the same drive, RAIDZ is an incredible option

Just to clear up on the 'expansion' thing: ZFS does allow you to expand, but in a different way that is both safe and has unique advantages, but doesn't work like many people are accustomed to.

The traditional way of expansion (adding one or more disks to an existing RAID-array) is to create a 'transitional' phase where you have two arrays:
- the original array with all your data on it
- a new 'virtual' array with the new disks and old disks included; to be the new array

Now data is copied from the original array to the new array; overwriting data on your good disks as well. But as the new array is larger than the original array this would work and at the end the new array has more space and all data has now been re-distributed across all drives.

What many people perhaps don't realise, is that expanding this way is quite dangerous. What if a disk gets a bad sector, or some other event disrupts the expansion process; in many cases that means you lose ALL your data, as the array is degraded and in a complex state during the expansion process; many RAID engines do not provide a good way to recover from a situation where during the expansion something goes wrong.

Even if it worked without problems, there are other negative sides about expansion. Due to you adding data disks without adding additional redundancy, that means the level of redundancy drops; and so does the reliability of the array. A 4-disk RAID5 is nice, but a 20-disk RAID5 is not.

Another problem is the capacity of the disks. When you buy your first fileserver, you may use 1TB disks for example, but right now the disks with best price-per-GB would be 1.5 or 2.0TB disks instead. With traditional expansion, you're forced to buy 1.0TB disks instead.

ZFS addresses these three issues in the following ways:
- it does not support the expansion of existing RAID-Z or RAID-Z2/3 arrays (RAID5/6/7)
- it does allow you to add a second array like a RAID-Z to the same pool as your first, sharing the free space and making it available under one unit. In essence, ZFS allows expanding by RAID0-ing multiple arrays together.
- best yet: you can use larger capacity disks for your second array. So you started with 8x 1TB disks and add 4x 2TB disks later if you like; or 6x 1.5TB disks; it should not matter. Though keeping the same disk count would be best for performance.
- expanding this way is totally safe as your original data remains untouched and will not be relocated; you're simply adding a second array to your existing pool.

So for ZFS, expansion works different but i would not say it is not supported; it just works in a different way. And i like it a lot better than traditional expansion, really. The only downside is that it wouldn't be wise to add single disks without redundancy (though possible) so that you must add multiple disks at once. i.e. 4 at a time creating a RAID-Z. This does mean the storage capacity will be slightly lower, due to you adding parity disks to the whole setup. But i think that makes a lot more sense than running a 16+ drive RAID5 or RAID6.

Another neat feature of ZFS: multiple copies. By setting copies=2 on a directory like Documents, you can instruct ZFS to create two copies of each new file created or modified, ZFS would then guarantee that each of those copies are stored on different physical disks, as long as you use ZFS' own RAID and not hardware RAID. With two arrays, one copy would be distributed across arrays instead. This copy would also gain the redundancy of the underlying array. Thus, using copies=2 for your most precious (and often not that large) data is really a nice feature.

And i didn't talk about Snapshots yet, which work beautifully in ZFS. But this isn't a ZFS topic i guess; but if any of you want more info on ZFS, just create a nice topic and i'll find it.

 

[Barnaby W. Füi]

Anyone know how well ZFS on FUSE works? I'd love to be able to use ZFS but really don't want FreeBSD.

 

[HardTech]

Which is how I learned the hard way that JBOD servers are really not acceptable.



Yeah, but I can't grow a RAIDZ array so in order to use RAIDZ on my 16 drive server I am building I would have to buy all the drives at one (can't happen). And with RAIDZ even if I have two parity drives, if I lose three drives (because consumer drives suck as you stated above) then all my data is gone. Since Unraid does not stripe the data if I lose three drives the maximum I can lose is the data on those three drives.

I figure if I am going to have bad enough luck for two drives to die during a rebuild, misery loves company and it could easily be three. So instead of trying to hot swap myself to a point where I am basically running a RAID 1, I instead chose to go with a solution that gives me a certain degree of protection (one drive failure), but won't flush all my data down the tubes in a worst case scenario. Plus I like only having one parity drive as it leaves more sata ports for actual storage.

In fact, I would argue that my Unraid server would have a WAY lower chance of massive failure than a regular RAID Z server. Why? Because with RAIDZ you have to buy all the drives at one- all the same kind of drive. That means you are buying maybe all out of one "batch," so if there is a manufacture problem you can easily lose three drives or more (bye bye data).

With my Unraid server, I can buy the disks one at a time FROM MANY DIFFERENT VENDORS. My current build has a WD Green drive, a Samsung drive, a Hitachi drive, and a Seagate drive. It is literally impossible for me to get bit by a bad run of production at a single factory. Of course, WHS has the same benefit which is why I consider the two to be the only acceptable options for a home server.

This is a good reason... in a nutshell, I really like what UnRaid has to offer:

- Lose the parity drive, rebuild array with no data loss
- Lose data drive, rebuild array with no data loss
- Lose parity and data drive at the same time, partial data loss
- Plug and play different drives to add to storage

If only there were a free option to do these things, in addition to serving files over a network and doing other things like ripping CD/DVD's and handling torrents

 

[sub.mesa]

Anyone know how well ZFS on FUSE works? I'd love to be able to use ZFS but really don't want FreeBSD.

I've not heard too many good things about FUSE, both performance is crap and it seems to crash a lot. And some corruption issues as well. But perhaps i'm doing injustice to this.. unusual project.

If you want to use ZFS for real, you're limited to latest FreeNAS release (finally with stable ZFS v13), FreeBSD v14, OpenSolaris 2009.06 v14 and derivatives. Anything else, unless i overlooked something, is not labeled stable; and often displays a warning message at boot time indicating as such.

Any particularly reason you don't like FreeBSD? It should have better hardware support than OpenSolaris and is known for incorporating advanced technology concepts, sometimes even pioneering them. And both FreeNAS and Mesa web-storage derivatives would be able to provide a nice web-GUI for users to interact with. I'm very interested in what holds you back on FreeBSD; as that's sort of the challenge with my new project, i guess.

 

[Barnaby W. Füi]

Any particularly reason you don't like FreeBSD? It should have better hardware support than OpenSolaris and is known for incorporating advanced technology concepts, sometimes even pioneering them. And both FreeNAS and Mesa web-storage derivatives would be able to provide a nice web-GUI for users to interact with. I'm very interested in what holds you back on FreeBSD; as that's sort of the challenge with my new project, i guess.

I don't dislike FreeBSD; I just don't care to tackle its learning curve. I don't really enjoy tinkering with OS stuff that much anymore. I get my kicks from programming. Messing with OSes just makes me feel neutered because I'm stuck trying to figure out something that someone else programmed. It just ends up feeling like a tedious chore. There is little satisfaction -- merely relief when I've got it all working. I'm much more comfortable with Debian derivatives and can get things done without too much hassle.

I also don't like GUI "control panel" type things that much. They just end up frustrating me and getting in the way. Doing things on the command line is simpler (if you understand it), more powerful and automatable, and I can know everything that's going on in the system and add/change whatever I want. Maybe FreeNAS is totally awesome and doesn't actually get in your way -- but even then, I don't think I would benefit that much from it.

Sorry for the long diatribe.

 

[sub.mesa]

I don't dislike FreeBSD; I just don't care to tackle its learning curve. I don't really enjoy tinkering with OS stuff that much anymore. I get my kicks from programming. Messing with OSes just makes me feel neutered because I'm stuck trying to figure out something that someone else programmed. It just ends up feeling like a tedious chore. There is little satisfaction -- merely relief when I've got it all working.

Well felt like deja vu to me reading the above quote. I've felt that all my life when dealing with computers, and hoped they worked better. That means: that the programmer/designer has done its job properly and the underlying design is solid and not prone to errors or malfunction.

Basically, we've not yet evolved computers to the point where they 'just work'. Filesystems like NTFS and RAID are all flawed in some way in that they could expose corrupt data to the application, which could have unforeseeable or even disastrous effects. It's not that easy to store files reliably; one of the basic foundations of computers.

That's why i like ZFS so much, because it takes so much worry away from the user. Some examples of how ZFS can make your life easier:

• No offline filesystem checks ever; ZFS checks and corrects filesystems and corrupt data on the fly when accessed;
• ZFS never exposes corrupt data to underlying applications; i.e. it ensures end-to-end data integrity. If the data written to ZFS pool was corrupt to begin with, this wouldn't help of course;
• If in any event ZFS cannot correct corruption, you would still know which files are corrupted; which is a lot better than not knowing whether corruption occurred at all like most people using RAID today;
• Snapshots can easily create 'restore points' to where you can return to, much like time machine in Apple OSX.

ZFS has been touted as the filesystem where you can 'store and forget'; meaning you wouldn't have to worry about it or have to do regular maintenance like rebuilds or filesystem checks. It also would maintain itself pretty well; with hot spares and self-healing as core features.

The idea of a ZFS NAS would therefore be quite like an initial setup, testing, setting up email notification to notify you of failed disks and you basically won't have to touch it again as it would maintain itself.

I also don't like GUI "control panel" type things that much. They just end up frustrating me and getting in the way. Doing things on the command line is simpler (if you understand it), more powerful and automatable, and I can know everything that's going on in the system and add/change whatever I want.

Hmm yes, i get the sense the underlying issue is that you still want control over what happens; i.e. you want to know and understand what happens when you press a button.

I designed my web-interface to be at least invasive as possible. That means that you could use both command line and web-interface at the same time. One doesn't exclude the other. All 'dangerous' commands like creating or expanding an array would allow you to show a confirmation page where the actual command to be executed is displayed. This means you can inspect and confirm the important commands for yourself.

One other benefit i found was that i like seeing information about my ZFS filesystems more on a webpage, with more structured and visually pleasing formatting. If you have 10 filesystems the terminal is still nice, but having 50 a web-page could be alot easier. It would also allow quick frontends to commonly used options like compression and copies=n.

The major difference between my project and FreeNAS, is that FreeNAS is a stripped down version of FreeBSD, whereas my project aims at being a full fledged (standard install) portable FreeBSD image, which could be used as normal FreeBSD multi-purpose server. Nothing is removed from the base FreeBSD install, and it allows you to fine-tune to your own personal setup. Some want game-servers to run (FreeBSD is still Linux binary compatible, able to run many game servers), some want Torrents, others want a real webserver/database; you name it. I also use a newer FreeBSD version (8.1 versus 7.3) and newer ZFS version (v14 versus v13).

My aim is at providing a portable package suitable to be placed on small HDDs or USB sticks, able to setup a ZFS pool and filesystem in less than 5 minutes via the web-interface, while still allowing all the flexibility and power of the underlying FreeBSD operating system. That means your initial learning curve will be small as it's easy to get something working quick; but it still allows you to dig deeper and discover a rich world of advanced technologies, which do require a steeper learning curve. So in essence, perhaps i'm trying to combine the best of both worlds (FreeNAS and FreeBSD).

Sorry for the long diatribe.

Not at all, i appreciate your feedback!

 

[pjkenned]

This is a good reason... in a nutshell, I really like what UnRaid has to offer:

- Lose the parity drive, rebuild array with no data loss
- Lose data drive, rebuild array with no data loss
- Lose parity and data drive at the same time, partial data loss
- Plug and play different drives to add to storage

If only there were a free option to do these things, in addition to serving files over a network and doing other things like ripping CD/DVD's and handling torrents

WHS + My Movies (or whatever you want) + whatever torrent client you like (I don't have any installed so I have no clue on these things). Plus, you can offload mail server, VPN, video encoding and etc duties to the WHS also. If you get ambitious, you can even buy hardware raid and run it below WHS. Most, if not all, of the ZFS/WHS solutions at this point support deduplication which saves a lot of space. I set up a test VM and saw eight systems with five snapshot days of 100GB of standard OS/ Application installs use about 260GB. It is pretty wild and that is why it is a big feature in enterprise storage.

Then again, it is your data and you may be a lucky person. I personally find relying upon luck to be an uneasy position.

 

[taltamir]

And BTW taltamir: HW raid 6 on adaptec/ areca IOP348 cards really don't see that poor of performance with Raid 6, especially for home server uses. It isn't like you are going to have 5,000-10,000 users banging the same server 24x7 in a home environment. Unless your goal is to saturate, at most one GigE link, Raid 1 alone with spindle disks is not going to get you performance. Raid 6, Raid-Z2, Raid-DP and etc are all pretty much standard if you want redundancy and speed... but you do need a lot of spindles. Then again, if you start mixing RAID 5/6 with poor hardware, you will probably see issues arise.

1. sure, with 300+$ controller cards you get acceptable PERFORMANCE.
2. its not about performance, it is about reliability. RAID6 has the same write hole, the same problem rebuilding with a non recoverable read error, and other issues that plague RAID5.

Anyways, RAID6 is NOT better than RAID1, not by a long shot. Use RAID1 or RAID1+0 if needed.

 

[Emulex]

freebsd rocks. it is by far the easiest o/s to virtualize and P2V and V2P. i can take freebsd 4 to 8 and build a VM, copy it to any boxen (physical), then virtualize it back with probably less than 5 setting changes (like ethernet interface). so that 10 year old freebsd box running on a 12 year old server can wham be running on a esx server with 6/12/24 cores in no time at all.

 

[taltamir]

freebsd rocks. it is by far the easiest o/s to virtualize and P2V and V2P. i can take freebsd 4 to 8 and build a VM, copy it to any boxen (physical), then virtualize it back with probably less than 5 setting changes (like ethernet interface). so that 10 year old freebsd box running on a 12 year old server can wham be running on a esx server with 6/12/24 cores in no time at all.

its also one of the few OS that can run ZFS... the only acceptable FS at the moment. (when btrfs is ready for prime time, we will talk again... until then ZFS is the only FS you should use)

 

[pitz]

Just do RAID-1, with 4 2Tb disks (2 arrays of 2 disks). Sure, it costs a little more than RAID-5, but there is a nice performance boost as well that you won't get with RAID-5. And never underestimate the ability to simply pull a disk out and install it in another machine if ever your RAID blows up.

If you use Linux, you can even do things like RAID-1'ing 2 drives (ie: 2 x 2Tb), to a single 4Tb drive in the future, which keeps your investment relatively intact and allows you to grow while you keep adding drives. So, for instance, you buy 2 x 2Tb drives today and RAID-1 them together for 2Tb of RAID-1 storage. When 4Tb drives get cheap, you just need to buy a single 4Tb drive, and then mirror that single 4Tb drive to your pair of 2Tb drives. When the 8Tb drives come out, you could RAID the 4Tb drive, and the pair of 2Tb drives up with the 8Tb drive, etc., etc., to the limits of the number of SATA ports available. The more spindles, the faster the overall system can be.

 

[Night201]

Unraid

Unraid is awesome.

I have put together 3 systems, all working great (from 2.5tb to 10tb storage)