RAID levels: Why and which (depending on situation, of course)?

[+Fahim S. MVC]

  On 08/10/2013 at 23:06, riahc3 said:

 

This thread was guided toward standard RAID levels because most controllers and/or NAS boxes, only support standard RAID levels.

Example: Synology doesnt support (AFAIK) RAID-Z on its NAS boxes because it isnt standard. I wanted a apples-to-apples comparision where you dont have to for example buy/make your own FreeNAS custom box to support RAID-Z. Don't get me wrong: RAID-Z is great, its just not possible with a lot of hardware.

 

 

Right .... but you do realise that the drive pooling mechanisms are not supported on any NAS boxes either and that has to be done at operating system level don't you?

 

The point is, the NAS vendors are taking you for a ride, by saying that RAID is important.  It's not.  It's easy to do for them, which is why it is touted as such an important feature.

 

The only way that you get this supported by a 'file server', which is essentially what a NAS box is, is to build your own.

[riahc3]

Hello,

 

  On 09/10/2013 at 08:18, .fahim said:

Right .... but you do realise that the drive pooling mechanisms are not supported on any NAS boxes either and that has to be done at operating system level don't you?

 

The point is, the NAS vendors are taking you for a ride, by saying that RAID is important.  It's not.  It's easy to do for them, which is why it is touted as such an important feature.

 

The only way that you get this supported by a 'file server', which is essentially what a NAS box is, is to build your own.

 

The thread was oriented towards standard RAID levels, where NAS boxes and custom boxes, can be compared equally.

 

A NAS box is simply plugin your drives, select your RAID type, and make a share (or shares): 1 hardware decision, and simple software configuration.

With custom NAS boxes (this is from what I understand, dont get me wrong) such as the HP ProLiant N40L Microserver, there are literally hundreds of things you have to do. Install the OS you want (RAID-Z isnt supported in Windows, right?), choose a decent (which is at a 200+ price point) RAID controller card, configure everything from the OS and RAID almost from scratch. Its is problably less expensive, but also more time consuming (they say time is money) as you have so many questions if you have never even attempted something like this. Its one of the reasons Im not too intrested in the custom route (Id love to but I dont see the time).

 

 

Thank you for your post, .fahim

[+Fahim S. MVC]

  On 09/10/2013 at 08:38, riahc3 said:

Hello,

 

 

The thread was oriented towards standard RAID levels, where NAS boxes and custom boxes, can be compared equally.

 

A NAS box is simply plugin your drives, select your RAID type, and make a share (or shares): 1 hardware decision, and simple software configuration.

With custom NAS boxes (this is from what I understand, dont get me wrong) such as the HP ProLiant N40L Microserver, there are literally hundreds of things you have to do. Install the OS you want (RAID-Z isnt supported in Windows, right?), choose a decent (which is at a 200+ price point) RAID controller card, configure everything from the OS and RAID almost from scratch. Its is problably less expensive, but also more time consuming (they say time is money) as you have so many questions if you have never even attempted something like this. Its one of the reasons Im not too intrested in the custom route (Id love to but I dont see the time).

 

 

Thank you for your post, .fahim

 

I am not sure what makes you think you need to install a controller card in a Microserver as it supports 4 drives out of the box, 6 with some minor mods, 14 with a controller card, careful selection of discs and reasonable modding.  You can have a Microserver up and running as a file-server including the OS in about 2 hours - if you are a complete doughnut, less if you have a clue.  You might want to test your assumptions.

 

Agreed that this is probably twice as long as it takes to set up a NAS, as you shove discs in and off you go.  But you get a very basic file server with little in terms of options for storage approaches (and most of these don't really have the benefits that you are sold, such as RAID5), but the NAS vendors are very good at selling them - they conveniently don't tell you that replacing a faulty disc in a RAID5 array can batter the discs so another one dies or can cause parity issues that means the array can't be recovered.  They certainly don't tell you about issues such as silent data corruption and the benefits of scrubbing.

 

I am not saying that ZFS is the be all and end all (I know all my links point to ZFS articles) - what I am trying to say, is that file system design has moved on considerably since the invention (and consumerisation I guess) of traditional RAID.

 

As far as I am aware and I would love to be corrected if untrue, multbay NAS don't support multiple different approaches per set of discs (i.e. if I have a 6 bay NAS, can I have RAID5 on three discs, RAID0 on another 2 and an independent drive for the final disc?)

 

I have a QNAP NAS (old, single bay) and a MicroServer N54L. The Microserver is a much better NAS than the QNAP hands down because of the capabilities that software like FreeNAS (or even Windows if that's your poison) has.

 

It's no skin off my nose what you actually choose to buy - I'm just trying to help you  :D

 

Also, read this:

http://www.zdnet.com/blog/storage/why-raid-6-stops-working-in-2019/805

 

If this doesn't scare you away from RAID, nothing will.

[riahc3]

Hello,

  On 09/10/2013 at 09:37, .fahim said:

I am not sure what makes you think you need to install a controller card in a Microserver as it supports 4 drives out of the box, 6 with some minor mods, 14 with a controller card, careful selection of discs and reasonable modding.  You can have a Microserver up and running as a file-server including the OS in about 2 hours - if you are a complete doughnut, less if you have a clue.  You might want to test your assumptions.

See here is where BudMan might be correct when he stated that I am stuck on ways that things were done years ago: Years ago, I was recommended to get a hardware RAID card, not a software one. But now, from what I understand, yourself and Budman recommend this pooling method, which is completely software based.....Could you please clear the hardware vs software RAID up for me?

Also, you state 2 hours. A NAS box can do standard RAID (excluding intializing the RAID) in 2 minutes. There is a big difference.

 

  On 09/10/2013 at 09:37, .fahim said:

Agreed that this is probably twice as long as it takes to set up a NAS, as you shove discs in and off you go.  But you get a very basic file server with little in terms of options for storage approaches (and most of these don't really have the benefits that you are sold, such as RAID5), but the NAS vendors are very good at selling them - they conveniently don't tell you that replacing a faulty disc in a RAID5 array can batter the discs so another one dies or can cause parity issues that means the array can't be recovered.  They certainly don't tell you about issues such as silent data corruption and the benefits of scrubbing.

If you put your drives in a RAID5, I believe you are a person that knows that if two HDDs fail, your data is lost. I dont think any vendor has ever tried to write RAID as a backup solution (which it isnt).

These issues exists but they dont lie about them either: They simply do not tell you. Thats business and understandable.

 

  On 09/10/2013 at 09:37, .fahim said:

I am not saying that ZFS is the be all and end all (I know all my links point to ZFS articles) - what I am trying to say, is that file system design has moved on considerably since the invention (and consumerisation I guess) of traditional RAID.

 

As far as I am aware and I would love to be corrected if untrue, multbay NAS don't support multiple different approaches per set of discs (i.e. if I have a 6 bay NAS, can I have RAID5 on three discs, RAID0 on another 2 and an independent drive for the final disc?)

Yes, NAS boxes/racks support multiple RAIDs (if thats what you mean). You can have 2TBx2 in RAID0 and then 1TBx2 in RAID1. This is supported on a QNAP I have.

  On 09/10/2013 at 09:37, .fahim said:

I have a QNAP NAS (old, single bay) and a MicroServer N54L. The Microserver is a much better NAS than the QNAP hands down because of the capabilities that software like FreeNAS (or even Windows if that's your poison) has.

 

It's no skin off my nose what you actually choose to buy - I'm just trying to help you  :D

Well I made this thread wanting to know more about RAID so it might influence what I buy in the end. BudMan was making a huge deal about it so I might as well learn some more :).

For a NAS, I personally wouldnt go the way of FreeNAS because I need something "quick and simple". FreeNAS is simple but I just need that plug and play feeling that I should be able to do things without looking up information (Yeah, Im a Windows sort of guy)

  On 09/10/2013 at 09:37, .fahim said:

Also, read this:

http://www.zdnet.com/blog/storage/why-raid-6-stops-working-in-2019/805

 

If this doesn't scare you away from RAID, nothing will.

At the end, I dont care much if the RAID fails; I make my backups. So if it fails, it takes me 2 hours to go to the store and back and replace the drive.

[+BudMan MVC]

"Yes, currently Im around 1-1.5TB of data; Important, there is problably 10GB at max? The rest is either replacable or VMs."

 

You have only 10GB of critical data and your setting up a 4x3TB raid 5???  WTF dude??  Talk about waste of space..  If your set on raid 5 why not 3x2TB or 3X3TB?

 

"I just need that plug and play feeling that I should be able to do things without looking up information"

 

And drivepool does that hands down!!  A freaking monkey could get it going in 30 seconds.

 

"It seems intresting but I need to know how it works and what exactly I need it to work."

 

I would suggest you contact them if your curious to the tech details -- but at a high level its really simple.. The software creates a virtual disk to the OS..  And it then lets the os access the files through the virtual disk while the actual files reside on the real disks.

 

There is a free trial - just grab it ..  Oh ###### I am late for work!!! ;)

[riahc3]

  On 09/10/2013 at 13:53, BudMan said:

"Yes, currently Im around 1-1.5TB of data; Important, there is problably 10GB at max? The rest is either replacable or VMs."

 

You have only 10GB of critical data and your setting up a 4x3TB raid 5???  WTF dude??  Talk about waste of space..  If your set on raid 5 why not 3x2TB or 3X3TB?

The 4x3TB RAID5 is set up a LONG time ago. 2 years at least.

An opinion: Lets say I use a VM with Windows something that I play a game daily on that is only compatible with that version of Windows. It isnt critical (as that is replacable) but the thought of having install the OS again and install the game again, takes up critical time, which I rather spend on play the actual game. Its pure laziness.

  On 09/10/2013 at 13:53, BudMan said:

"I just need that plug and play feeling that I should be able to do things without looking up information"

 

And drivepool does that hands down!!  A freaking monkey could get it going in 30 seconds.

Didnt know things like Drivepool existed. I have a issue with Drivepool though...

 

  On 09/10/2013 at 13:53, BudMan said:

There is a free trial - just grab it ..  Oh **** I am late for work!!! ;)

I thought that these things were free (software implementations of this sort). You get support if you pay for it, I guess, but thats not something intresting. Id like something in the $0.00 area.

Thank you

[+BudMan MVC]

"but the thought of having install the OS again and install the game again"

Well then just setup the folder that VMs disks sit in as duplicated.. Now even if a disk fails VM is on more than 1 disk and don't have to reinstall or restore from backup.

As to it not being free? Dude its $20, you can affort 4x3TB 2 years ago.. But now you don't want to fork out $20? How much does a raid controller cost, how much does a nas cost with raid 5 support? Etc.. With $20 software any windows box can be your access to your pool.

If you want free check out http://www.greyhole.net/ it not windows based but there is a VM you can run that gives you access to the pool.

More of parity solution but still allows creation of pools of different sized disks, etc. http://www.vilett.com/disParity/

There is http://snapraid.sourceforge.net/ that comes to mind as well..

As I said before there are many many ways to skin the cat.. Thinking that raid 5 is still the end all solution to easy access and redundancy of critical data is out dated.. Unraid has free option, limited to 3 disks I believe. Thing with unraid is it doesn't really make a lot of sense unless your doing more than 3 disks.. So the free version isn't getting you much.

[eXtermia]

I am actually quite happy using RaidZ1 even more so then Raid5. It supports hardware that wont work with normal raid solutions and actually is in some case more robust. I am running NAS4FREE on top of a esxi5 hypervisor on a box that didn't support native hardware raid5 running SATA drives. creating a new raidz volume and sharing back out to the host machine and other boxes over iscsi

 

esxi hypervisor free

NAS4FREE free

even I think the virtual server 2012 hypervisor from MS is free

 

add VEEAM backup, you can even move virtual datastores (also free) but the machine does need to be shutdown , removed from the esxi inventory and re-added to the inventory once moved. This gives you similar functionality to vmotion in higher price classes of exsi and vsphere.

[eXtermia]

NAS4FREE runs better for me then FREENAS. couple things I like about my virtualized san. I can back up the host easy even to a iscsi lun. I can use data deduplication (need about 8gbs for zfs and about 1gb memory for each terrabyte unless you use SSD caching of the transaction logs). I can setup multiple boxes to mirror the main san automatically.

 

even if the zfs is on the main file system my iscsi luns can be formatted as GPT NTFS disks. even if the main NAS4FREE fails or the main hardware fails i can slap the drives into a windows box running VMWARE quick and mount the virtual os and have at least access back in literally minutes. (or even from a live cd)

 

my next project is to create a CARP connection back to somewhere offsite and back over the wan. the beauty here is as the NAS is virtualized it doesnt care what hardware runs underneath and I can mirror a whole raidz volume (or even break or in case of failure) it without causing additional stresses, as compared to raid 5, on the renaming good data drives. Once i do replace the existing failed drive I don't to spend the hours rebuilding the volume like I do on hardware raid either.

 

Sure a good controller with memory would give me even better performance, but I am having fun with RaidZ volumes

[riahc3]

Hello,

  On 09/10/2013 at 17:49, BudMan said:

"but the thought of having install the OS again and install the game again"

Well then just setup the folder that VMs disks sit in as duplicated.. Now even if a disk fails VM is on more than 1 disk and don't have to reinstall or restore from backup.

Yes, I understood that, I was just explaining it with my RAID5 example. With your pool setup, this would not happen.

  On 09/10/2013 at 17:49, BudMan said:

As to it not being free? Dude its $20, you can affort 4x3TB 2 years ago.. But now you don't want to fork out $20? How much does a raid controller cost, how much does a nas cost with raid 5 support? Etc.. With $20 software any windows box can be your access to your pool.

Its a personal pet peeve not to spend money on software unless I really need it (such as Windows for example). Im trying to use as much (open source) freeware as I can on my PCs nowadays to stay close to 100% legal and use free on software.

  On 09/10/2013 at 17:49, BudMan said:

If you want free check out http://www.greyhole.net/ it not windows based but there is a VM you can run that gives you access to the pool.

I rather stick to Windows as Im more comfy in that enviorment. I did set up my DD-WRT router pretty OK and havent had any problems....but I perfer not to. Ill try it in a VM anyways.

  On 09/10/2013 at 17:49, BudMan said:

More of parity solution but still allows creation of pools of different sized disks, etc. http://www.vilett.com/disParity/

There is http://snapraid.sourceforge.net/ that comes to mind as well..

As I said before there are many many ways to skin the cat.. Thinking that raid 5 is still the end all solution to easy access and redundancy of critical data is out dated.. Unraid has free option, limited to 3 disks I believe. Thing with unraid is it doesn't really make a lot of sense unless your doing more than 3 disks.. So the free version isn't getting you much.

Yeah, Im pretty convinced about RAID5 being worse. The only pitfall is that I would have to make my own custom NAS box and I would have to take the time to learn the in/out of the OS and later the software technology I would use to put all the disks in the pool. The old saying "if it aint broke, dont fix it" comes to mind, but since we are working with small critical data, I cant hurt to switch from RAID5 to this.....

Thank you

[+BudMan MVC]

"With your pool setup, this would not happen."

Says who - If I deemed those files critical and duplicated them then yes those files would always be available as long as 1 disk in the pool was online. And full speed btw.. And if I replaced the bad disk then all the files that are set for duplication would be copied over automatically.

In your setup if disk fails files are there at reduced performance, and at the size of your array - its quite possible when you go to rebuild it the rebuild fails because you hit a read bit error.

"The only pitfall is that I would have to make my own custom NAS box and I would have to take the time to learn the in/out of the OS and later the software technolog"

You lost me -- you click go on an exe! You add the drives to the pool.. They don't even have to be empty BTW.. The space that is used by files is just not useable in the pool.. And you would have to access them via the actual disk vs the pool disk. This runs on your windows OS, there is nothing to learn.. You don't know what a Disk is? The pool looks like a disk and just has all the space of all the disks you put in it. You click options and pick the folders or files you deem critical and creates copies on the number of disks you want to put copies on in the pool.

With the amount of files you have 1.5TB.. Just pull out a disk from your array, connect it to anything running windows, be it VM or physical.. Copy your files from your now degraded array. Then add the disks you want from your old array to your new pool.. Pick which files you seem critical and duplicate them be it 2 or 3 or even all 4 disks in your pool.

[riahc3]

Hello,

  On 09/10/2013 at 21:52, BudMan said:

"With your pool setup, this would not happen."

Says who - If I deemed those files critical and duplicated them then yes those files would always be available as long as 1 disk in the pool was online. And full speed btw.. And if I replaced the bad disk then all the files that are set for duplication would be copied over automatically.

In your setup if disk fails files are there at reduced performance, and at the size of your array - its quite possible when you go to rebuild it the rebuild fails because you hit a read bit error.

Yes, I know. Thats the reason I stated with YOUR setup :)

  On 09/10/2013 at 21:52, BudMan said:

"The only pitfall is that I would have to make my own custom NAS box and I would have to take the time to learn the in/out of the OS and later the software technolog"

You lost me -- you click go on an exe! You add the drives to the pool.. They don't even have to be empty BTW.. The space that is used by files is just not useable in the pool.. And you would have to access them via the actual disk vs the pool disk. This runs on your windows OS, there is nothing to learn.. You don't know what a Disk is? The pool looks like a disk and just has all the space of all the disks you put in it. You click options and pick the folders or files you deem critical and creates copies on the number of disks you want to put copies on in the pool.

With the amount of files you have 1.5TB.. Just pull out a disk from your array, connect it to anything running windows, be it VM or physical.. Copy your files from your now degraded array. Then add the disks you want from your old array to your new pool.. Pick which files you seem critical and duplicate them be it 2 or 3 or even all 4 disks in your pool.

Question: if ALL files (both critical and noncritical) I select to be on all 4 disks, what is this roughly equivilent to? A RAID1?

I havent had time to download and tinker with the VM, so possibily trying it out would answer my own question :)

Thank you guys for all the information.

[riahc3]

Hello,

Trying out the DrivePool trial now, but I believe you said that the interface didnt matter (including USB). I plugged in two USB drive but DrivePool doesnt see them.

[+BudMan MVC]

Did you add them to the pool?  Are they formatted NTFS?  They have to be NTFS to be part of the pool or RAW partitions that it will format as ntfs.. Is this a VM, does the OS see the drive?  The disks can not be dynamic..  Guessing if they are small usb flash or disks even they are FAT..  Yeah change them to ntfs and you should be good.

 

Let me give mine access to a usb disk and verify it sees it..

 

edit:  Grabbed the trail and ran it on this box since easier than connecting a passthru usb to my vm..

 

There you go - sees USB just fine and can add to pool

 

 

And there you go - pool has the usb disk in it.

 

 

edit2: "Question: if ALL files (both critical and noncritical) I select to be on all 4 disks, what is this roughly equivilent to? A RAID1?"

 

I wouldn't select non critical files to be on all 4 disks..  That is just overkill and waste of space..  But sure you could do that I guess.  Wouldn't it just be more cost effective to do what with 2 or 3 -- your worried that 3 disks out of 4 fail?  Be it raid parity or even mirrored data or one of the newer methods that allow for parity and duplication are not BACKUPS of your critical data..

 

You do understand that data in an array no matter what kind could be corrupted or destroyed be it due to hardware issues or user errror or viruses, etc.  So it is NOT a backup!! Your "critical" data needs to be on different system that is not connected in real time to your existing data.  So that if something goes wrong with your data you have that non connected copy to restore from.  Raid and or duplication in a pool just allow data to still be available even with something like a disk failure -- you still need a valid and robust backup method or no matter how many disks your data sits on in an array or pool it could be lost.

 

edit3:  Is your usb being seen as removable or local disk?  If seen as removable its quite possible those can not be added.  But many newer flashes show up as local - I know my tiny 16GB little cruzer does, I believe there are some tools out there to flip the RMB bit.  I would contact their support about disk not be able to be seen by the software..  Their support is quick and very impressive.

[riahc3]

Hello,

  On 10/10/2013 at 12:12, BudMan said:

Are they formatted NTFS?

That was the problem. I thought you could mix file systems as well...

  On 10/10/2013 at 12:12, BudMan said:

edit2: "Question: if ALL files (both critical and noncritical) I select to be on all 4 disks, what is this roughly equivilent to? A RAID1?"

 

I wouldn't select non critical files to be on all 4 disks..  That is just overkill and waste of space..  But sure you could do that I guess.  Wouldn't it just be more cost effective to do what with 2 or 3 -- your worried that 3 disks out of 4 fail?  Be it raid parity or even mirrored data or one of the newer methods that allow for parity and duplication are not BACKUPS of your critical data..

 

You do understand that data in an array no matter what kind could be corrupted or destroyed be it due to hardware issues or user errror or viruses, etc.  So it is NOT a backup!! Your "critical" data needs to be on different system that is not connected in real time to your existing data.  So that if something goes wrong with your data you have that non connected copy to restore from.  Raid and or duplication in a pool just allow data to still be available even with something like a disk failure -- you still need a valid and robust backup method or no matter how many disks your data sits on in an array or pool it could be lost.

My question was more directed as a learning question then real world usage. This way I know its equivilent to a RAID1.

OK, I played a bit with DrivePool. The GUI seems simple enough...

See but here is the confusing part, at least for me....

I got 2 USB drives 2GB each and put them in the pool. More or less the space is 3.76GB. This is what Windows reports. I copied a ISO and it worked great and now tells me there is 2.83 GB out of 3.76 GB free. I copied a dummy file that weighs in at 2.50 GB and Windows tells me there isnt enough space. I IMAGINE this has something to do with duplication...Im still tinkering so it might be a option that I havent activated.

[+BudMan MVC]

"I thought you could mix file systems as well..."

I would have to go back through my posts, I don't think ever said you could use different file systems with drivepool from stablebit? If I did my bad. I don't see how you could have different file systems all under the same pool with drivepool. Maybe some other software could do that.. Greyhole for example can mount remote file systems.. So in that case I have to assume you could have say riser on one remote system and ntfs or ext4 on other systems and through a remote mount add them to a storage pool. There are lots of projects that work with distributed file systems. GlusterFS, Ceph and windows own distributed file system DFS.

Its possible that drivepool might add that in future releases?

As to your not enough space -- yeah that could be related to not enough space to create the duplication.

Back to your FREE way of doing this - I do believe windows 8 has storage spaces.

[riahc3]

Hello,

  On 10/10/2013 at 14:54, BudMan said:

Back to your FREE way of doing this - I do believe windows 8 has storage spaces.

Sadly I cant test that out as WS2012 requires a 64 bit processor which right now I cannot virtualize.

[+BudMan MVC]

What are you using as your VM software, what hardware exactly? I know for sure that in the past I could create a 64bit vm, even with the host was only 32bit.. I use to have an old P4 32bit, that I ran 64 bit vms on.

Yup just as I thought even virtual box had this feature back in version 2.1 back in 2008

https://www.virtualbox.org/wiki/Changelog-2.1

This version is a major update. The following major new features were added:

Support for hardware virtualization (VT-x and AMD-V) on Mac OS X hosts

Support for 64-bit guests on 32-bit host operating systems (experimental; see user manual, chapter 1.6, 64-bit guests, User Manual page 16)

[riahc3]

Hello,

  On 10/10/2013 at 16:01, BudMan said:

What are you using as your VM software, what hardware exactly? I know for sure that in the past I could create a 64bit vm, even with the host was only 32bit.. I use to have an old P4 32bit, that I ran 64 bit vms on.

Yup just as I thought even virtual box had this feature back in version 2.1 back in 2008

https://www.virtualbox.org/wiki/Changelog-2.1

This version is a major update. The following major new features were added:

Support for hardware virtualization (VT-x and AMD-V) on Mac OS X hosts

Support for 64-bit guests on 32-bit host operating systems (experimental; see user manual, chapter 1.6, 64-bit guests, User Manual page 16)

I knew that but I thought the processor had to support VT-x.....For some reason, I could have sworn my home processor didnt support it but it does!

In general I use VMWare Workstation. Virtualbox is a pain sometimes to work with.

[+BudMan MVC]

just used that as an example...

[+Fahim S. MVC]

I'm confused now.  Surely the reason you wanted a NAS was because the 'network attached' part was important?

[+BudMan MVC]

And what is not "network attached" ?

[+Fahim S. MVC]

You can't have a pooled based technology unless you build your own NAS/Server or leave your desktop on all the time and share the pool from there.

OP doesn't want to do the former, or that's what I thought.