Difference between MS's Drive Extender and JBOD in a NAS?

[stevech]

Difference between MS's Drive Extender and JBOD in a NAS is?

I don't recall that JBOD has any redundancy to preclude data loss on drive failure. Same for MS's Drive Extender?

[gplasky]

The drive extender allows a bunch of drives to appear as a single drive and supports replication (folder and file). Does JBOD do the same thing?

Gerry

[PLUCKYHD]

Originally Posted by gplasky

The drive extender allows a bunch of drives to appear as a single drive and supports replication (folder and file). Does JBOD do the same thing?

Gerry

JBOD is just what the name stands for "just a bunch of disks". There is no redundancy or drive pooling.

I would recommend a raid 5 or raid 6 (if you can afford it) hardware not software configuration before I could ever recomend WHS again. I am currently moving everything off WHS bad drive extender onto raid arrays. Raids give you the same ability (see all drives as one array) but they do it right/better I won't bore you with the details. I always hated drive extender but bought into the whole WHS thing. When MS killed DE that was my turning point to put it all on raid array;s like I should have in the first place.

[gplasky]

Originally Posted by PLUCKYHD

Raids give you the same ability (see all drives as one array) but they do it right/better I won't bore you with the details. I always hated drive extender but bought into the whole WHS thing. When MS killed DE that was my turning point to put it all on raid array;s like I should have in the first place.

I'm with you there. I'll probably build a new server with hardware RAID. I'm also looking at UNRAID and also Virtual server to see what may work best. But I' thinking from a stabilty standpoint do the hardware RAID and be done with it.

Gerry

[Nelbert]

H/W raid will be the most stable.

Unless you really have a need to plug in random sized disks found down the back of the sofa I'd not bother throwing money at unRaid and just use mdadm on Linux for software raid. Then again I only have 4 disks and don't want a pre-packaged system with the limitations they bring.

[jaminben]

I've been thinking of doing something with my system recently (bored with current setup) and had thought about using a raid rather than WHS but don't really know where to start..... so I have a few questions if you would be so kind as to answer some

1. Whats a good hardware card to get? Not too expensive but not too cheap either.....basically middle of the road.

2. Is a raid expandable? e.g can I start with 2 or 3 disc's and add more later or do I need to start with the total that I want in the future?

3. Does the operating system need to be installed on the raid or does it use a seperate disc?

4. Is it better to keep the raid exclusive for kept media (ripped movies etc) and not use it as a recording drive?

5. Is there a limit to the amount of TB's a raid can have?

6. Can a system have more than one raid card installed at anyone time?

7. Do raids require much CPU usage or will my old'ish Intel core 2 duo 2.4Ghz be ok?

8. I've been looking at this card but I see it doesn't do raid 6..... is this good or bad? What would raid 6 give me over raid 5?

Pretty basic questions I guess and Google would probably provide the answer's but I'd rather get advice than try to work it out myself and make a mistake.

Cheers

Ben

[mikejaner]

I will try to answer these:

1- This comes down to many different personal prefrences. IMHO LSI Megaraid cards are good, and Dell PERC 6i Cards are good. I have both, and they work well. With LSI you are looking at $300-$500 range for 4-8 port cards with battery backup etc. Dell PERC 6i (not 6ir), you can get for about $150 without the battery backup, and probably around $200 with it on Ebay.

2- RAID is expandable in certain situations. Generally in $1000 RAID cards, or Software RAID like mdadm. I will probably be corrected on this because I don't keep up with all the models. There could be a $700 card too.
I did this my self twice with my mdadm array on my Ubuntu server. Once, upgraded my 4.5TB RAID5 array to 6.0TB, and the second time reshaped my RAID5 array to a RAID6 array, all while it was live and recording etc.....

3- No, you can have an OS drive, like a single SATA drive, and then your RAID array, like a d:\ drive etc... You can also RAID your os drive, like a Mirror, and then have a RAID5 array for your data.

4- You can do whatever you want. I record to my RAID Array. As long as the drives can keep up with the demand. In most cases, including recording multiple HD streams, and playback to multiple clients, speed should not be an issue. Too many instances of Comskip can kill that performance though.

5- No, just make sure you use the right type of partition/file system which supports larger partitions, like a GPT Partition formatted in NTFS for Windows for >2TB, and ext4 or XFS in Linux.

6- Yes

7- No Core2 is more than enough for a Software RAID, and Hardware RAIDs don't use any CPU

8- If a card is that cheap, steer clear unless you plan on doing software raid.

[gplasky]

If you are looking for best performance in a RAID configuration, RAID 10 will be better performing than RAID 6 and RAID 5. According to manufacturer specifications and official independent benchmarks, in most cases RAID 10 provides better throughput and latency than all other RAID levels except RAID 0 (which wins in throughput).

It is the preferable RAID level for I/O-intensive applications such as database, email, and web servers, as well as for any other use requiring high disk performance

Our standard at work is mirror the OS drive(s) (Think a couple of SSDs.) and RAID the rest for data.

Areca and 3Ware are your better known, high-end RAID cards.

Gerry

[jaminben]

Thanks for the quick replys

I've found a card on ebay.... its used but its also about 1/3 of the retail price.

Adaptec AAR-2820SA 8 port SATA 2 PCI-X RAID controller

Look anygood? The other ones you mention don't really seem to be very available in the UK..... I could import but I would rather buy from the UK.

Cheers

Ben

[PLUCKYHD]

For home use you really don't need a high end card. I use the ones that come with sans digital external esata enclosures for like $200 have not had a issue running raid 5 on them yet. I can't think of the raid card it comes with right now but I have a good history with them.

[stanger89]

Originally Posted by Nelbert

H/W raid will be the most stable.

Unless you really have a need to plug in random sized disks found down the back of the sofa I'd not bother throwing money at unRaid and just use mdadm on Linux for software raid. Then again I only have 4 disks and don't want a pre-packaged system with the limitations they bring.

First off, I'm not picking on anyone, your posts just has all the stuff I want to address.

I really liked RAID, I had an 8-port RAID-5 array and it was great. Now, this was an "early" 3ware 7506-8, it was PATA. For a long time I was very happy with it, that was, until I ran out of space. This is basically where my love affair with RAID ended, even though I didn't realize it at the time.

The problem for me comes down to cost. Not so much the cost of building an array, that's palatable (I did it once), but the cost of rebuilding it when you run out of space. Let me explain.

Now part of my problem was that it was an "early" card, so it didn't have OCE (Online Capacity Expansion) and it was the last of the PATA, so I was stuck at the time replacing the card and all the drives. That can get to be a very expensive proposition.

Now, if you were buying today, you have things like OCE, which helps. But for OCE to be very useful, you need a card with quite a lot more ports than you initially need, and these 12, 16, 24 port cards are very expensive. And beside that it can get logistically troublesome running that many drives.

Here's the other issue. When you do fill up your array, you've basically got to replace all the drives. You can play some tricks like replacing two, and building a second array on the free space on those drives, and expanding the second array as you replace drives, but then you lose that nice single volume.

Basically I've faced this "to RAID or not to RAID" question for quite a long time now. The first time I answered it by getting a ReadyNAS X6. There's a lot of stuff I really like about it, but it wasn't big enough, and I filled it up twice (replaced the drives in it once).

Most recently I answered the question by replacing my RAID array in my SageTV system with an unRAID system. I built it with a Supermicro X7SPA and 2TB drives.

I really like unRAID for media storage. The reason it it's basically free from all the "trouble" of RAID-5 or RAID-6. I think some here (as I did) underestimate the value of not being tied to one size drive. IMO the benefit isn't when you first build it, it's when you start to fill it up. I've got 5 drives in my unRAID right now, it amounts to 10TB of usable space. It will take me a while to fill that, but when I do, I can just replace a couple of the 2TB drives with 3TB ones (or add 3TB drives).

The real benefit of the drive size freedom is you can continually buy the largest and/or most cost efficient drive sizes, and not get stuck needing a plethora of small disks. Looking forward I can see where an unRAID Plus (6 drive) license will fill my needs for the forseable future, where if I were building a new RAID system, I'd be looking at something much larger than that so I could add more, smaller drives further into the future.

Of course there are other benefits like unRAID can spin down unused discs and only spins up the ones you're using. So for a media server where you're mostly reading, you can save a bit of heat/power/usage.

[Nelbert]

1. Look to be spending ~ £200

2. Depends on the cards. HP have a £350 that supports online capacity expansion and raid migration. Look for OCE as a feature

3. Take your pick. If you don't put the OS on a raid system what happens when the OS drive fails? For home use using the same array as the data shouldn't be an issue, but I tend to put the OS on different disks. That way you can move the data to a different machine easily.

4. Depends on usage profiles, but for home use shouldn't matter.

5. Depends on the card. Most no longer have a 2Tb limit, and many proper cards support backplane daisy chaining to silly amounts.

6. Becareful with cheap cards. Promise have some cheap cards which are fakeraid and sometimes it'll work with 2 cards sometimes it won't. Hardware cards shouldn't be an issue.

7. Software raid won't tax your cpu and a hardware card will have 0 cpu usage.

8. It's interesting, but I wouldn't touch it. It's part of a new cheap breed. Not fakeraid but not a full hardware raid card. The card only has 2 SATA channels. In one channel they then attach a port multiplier with a cheap hardware raid chip on it. The 5 raid drives are all actually connected to the PC via 1 sata port. With 3 drives you should get full speed to the drives, 4 or 5 drives and you'll saturate the sata port between the PC and the raid chip.

One issue with cheap cards is they have no status lights or buzzers to warn you of a disk failure. At best, they rely on software often running in the desktop system tray to tell you. If the card goes in a headless server you'll never know when a drive fails until you lose enough to kill the array.

Don't forget to also consider HP, SUN, IBM, SuperMicro etc... They often rebrand cards, especially at what they consider entry level. When they refresh their range you can get branded cards cheaper than the original manufacturer versions.

HP and IBM often rebrand LSI. Sun use LSI, Emulex, Intel cards.

For a 3ware card in the uk something like this 9650 would be much better than the one linked to.

Are you looking for a PCI-X or a PCIe card?

[PLUCKYHD]

Originally Posted by stanger89

First off, I'm not picking on anyone, your posts just has all the stuff I want to address.

Nice post Stranger. To me I guess having everything in one array isn't as important to me. To me sage shows them all combined so it doesn't matter to me. So basically what I am doing it 8 slot arrays at a time. I usually start with 4 drives and expand online until I fill the 8 slots (granted online expansion is slow 3 to 4 days but still possible). Then when it fills up I build another array.

To each their own Unraid is nice but nothing is perfect unfortunately and all options are better then DE from WHS.

[Nelbert]

Originally Posted by stanger89

First off, I'm not picking on anyone, your posts just has all the stuff I want to address.

No worries, Raid conversations are always based on personal experience and what works for one person doesn't work for another. Without discussion about things we never find out more.

Technology progresses and things change. I'm not going to dispute your dislike for expanding hardware raid arrays, but time has moved on. It's still something that is necessary when you fill all the ports. Either need to do the disk shuffle or just add another raid card.

I'm very curious as I've no idea with unRaid:
How long is it going to take to remove a 2TB drive, replace it with a 3 or 4 Tb drive, and rebuild your parity partition into your new drive?
If you have no free sata ports and want an extra 4Tb of usable space how many new disks are you actually going to need to buy? As I understand it, you can't just plug a 4Tb disk in and gain 4Tb of free space because it's has to become the parity drive as it's bigger than all the other drives?
How long is it going to take to rebuild?
Does the system have parity protection while the parity drive is being rebuilt on the larger disk?

I understand the appeal of unRaid and what they're trying to achieve, but things like lack of native UPS support, and only having APC support via a plugin, no stable NFS support in what is suppose to be a media nas, needing a cache drive as the number of disks increases at version 4.5 is simply not acceptable for me. On top of that forum responses along the lines of "your nfs client is the wrong version and I hate nfs anyway" from the dev are hardly professional when a modern Linux nfs client won't work. Being rejected for acceptance into the Linux kernel also has to raise questions about maturity, performance and stability of the code.

From Lime's website unRaid is an appliance, so to use unRaid with Sage you have to run 2 servers, one for unRaid and another for SageTV? Surely that starts to negate the power benefits of running Sage on the storage server itself?
I know green power is a great marketing pitch at the moment and we're all trying to reduce our power footprint, but Lime haven't overlooked that either. If you put the Server OS and Sage install on different disks to the data raid then the data disks should still power down when not in use.

The best solution I've used for management, realtime backup, stability, speed and expansion is ZFS. Completely free and no problem running large numbers of hard disks. I could keep have kept that in my solaris nas or switch to FreeNAS (just took the disks from the solaris box and plugged them into the FreeNas) but that still means running 2 servers. One NAS and one Sage.

For sage that leaves Linux or Windows to run a single server.

mdadm works well until it fails and then it becomes a right royal pia. While people seem to think x-raid is something new, mdadm has had the ability to perform the same thing for years, along with some emc hardware cards.

BeyondRaid looked interesting until drobo priced it out of sensible reach.

Windows software raid? lets just not even bother going there.

I think the future for data protection especially for soho and home users needs to move from being disk based to content based. unRaid, FlexRaid, greyhole, DNAS devices are steps in that direction, but still have along way to go. DE was a good start, but v2 was a mixed bag, moving in the wrong direction as it basically required mirroring all data for protection.

Having dealt with all sorts of wierd and wonderful disk systems and technologies in everything from single machines to highly available fault tolerant systems over the past 15 years, I'd still currently take h/w raid over the others for performance and stability in Windows and Linux servers.

If I was running a 2nd storage server then Freenas or Solaris on a usb stick/cf drive with zfs would be my first choice, especially having deployed and managed zfs in a high usage, highly available clustered server environment.

Ask me in another year and I might have changed my mind, as things like unRaid, FlexRaid and greyhole mature, but for now to run Sage and storage in one box I'll take h/w raid if possible.

[stanger89]

Originally Posted by PLUCKYHD

Nice post Stranger. To me I guess having everything in one array isn't as important to me. To me sage shows them all combined so it doesn't matter to me. So basically what I am doing it 8 slot arrays at a time. I usually start with 4 drives and expand online until I fill the 8 slots (granted online expansion is slow 3 to 4 days but still possible). Then when it fills up I build another array.

Problem I have with that is I don't want 5 arrays floating around. If I'd done that I'd be on my 4th now.

To each their own Unraid is nice but nothing is perfect unfortunately and all options are better then DE from WHS.

Agreed, it always amazes me how MS always seems to fall just short of making the product that would clobber the competition. I mean WHS was just so close. If they'd have put WMC's backend in it, and put an unRAID (ie parity vs duplication) like redundancy mechanism... But alas.

Originally Posted by Nelbert

Technology progresses and things change. I'm not going to dispute your dislike for expanding hardware raid arrays, but time has moved on. It's still something that is necessary when you fill all the ports. Either need to do the disk shuffle or just add another raid card.

I'm very curious as I've no idea with unRaid:
How long is it going to take to remove a 2TB drive, replace it with a 3 or 4 Tb drive, and rebuild your parity partition into your new drive?

Well, I had to swap out a couple of my 2TB drives (had to change the jumper), I think it took about 15 hours or so for the resync to happen.

If you have no free sata ports and want an extra 4Tb of usable space how many new disks are you actually going to need to buy? As I understand it, you can't just plug a 4Tb disk in and gain 4Tb of free space because it's has to become the parity drive as it's bigger than all the other drives?

True, worst case you'd have to get 2 drives. But if you already had an extra-large parity drive, you could just get another one the same size.

How long is it going to take to rebuild?

Depends on the drives but like above, I think mine was about 15 hours.

Does the system have parity protection while the parity drive is being rebuilt on the larger disk?

No it wouldn't, I'm not sure if there are plans for 2 parity drives at some point or not.

I understand the appeal of unRaid and what they're trying to achieve, but things like lack of native UPS support, and only having APC support via a plugin, no stable NFS support in what is suppose to be a media nas, needing a cache drive as the number of disks increases at version 4.5 is simply not acceptable for me.

For something mission critical I'd agree with you on some points (UPS), but at least for me, it's just for my media, which is all either backed up (my photos, etc) or replaceable (I have the discs) if something catastrophic happens.

NFS, I really don't see a need for a media server, most devices use Samba/CIFS.

And regarding the Cache drive, you don't need one, no matter how many drives you have. If you're going to spend a lot of time writing to it, it improves performance though.

From Lime's website unRaid is an appliance, so to use unRaid with Sage you have to run 2 servers, one for unRaid and another for SageTV? Surely that starts to negate the power benefits of running Sage on the storage server itself?

I thought the same thing, and spend a good bit of time trying to avoid that. But you can build an unRAID box that pulls less than 50W idle when you factor in disk spindown. The drives on a RAID array alone will pull more than that.

I'll put it this way, I replaced my old HTPC which had an 8-drive RAID-5 array with an unRAID box and my new HTPC and I think I reduced my usage by around 100W.

For sage that leaves Linux or Windows to run a single server.

You can actually go virtualized. unRAID runs pretty well in a VM, as does Sage. I would have done that but my R5000 just will not work in that environment.

I think the future for data protection especially for soho and home users needs to move from being disk based to content based. unRaid, FlexRaid, greyhole, DNAS devices are steps in that direction, but still have along way to go. DE was a good start, but v2 was a mixed bag, moving in the wrong direction as it basically required mirroring all data for protection.

What do you mean "content based"? How would you protect "content" without either using discs (RAID, unRAID, etc) or files (FlexRAID) or duplication?

Having dealt with all sorts of wierd and wonderful disk systems and technologies in everything from single machines to highly available fault tolerant systems over the past 15 years, I'd still currently take h/w raid over the others for performance and stability in Windows and Linux servers.

Personally, I don't have any concerns about unRAIDs stability. it's got far too many fans and too few 'opponents' (besides those who dislike it's philosophy), and I really don't need the performance, I'm just serving media to a couple clients (usually only 1 at a time even).

Ask me in another year and I might have changed my mind, as things like unRaid, FlexRaid and greyhole mature, but for now to run Sage and storage in one box I'll take h/w raid if possible.

I agree, or would have, it's only recently I've decided a second (or third in my case) box running unRAID made more sense for my use.

[PLUCKYHD]

Originally Posted by stanger89

Problem I have with that is I don't want 5 arrays floating around. If I'd done that I'd be on my 4th now.


.

So you have 32 drives based on a 8 drive array I understand you couldn't add drives online then like you can now.

[jptheripper]

I am shocked plucky. I love WHS. Love DE. it just works for me. I have 4 2tb drives on a PCI SATA card, 1 2tb and 2 1,5tb's on the onboard satas, plus a 2x1tb onboard raid for the OS. I have lost a few drives over the year or two i have been running it, and it is rock solid without ever lossing data. Plus it is an all-in-one solution for client backups that is flawless. I lost my HTPC to a failed OS drive and was back up and running on a spare drive in less than an hour.

Just because M$ is dropping DE in VAIL doesnt mean i have to upgrade to it

Originally Posted by PLUCKYHD

JBOD is just what the name stands for "just a bunch of disks". There is no redundancy or drive pooling.

I would recommend a raid 5 or raid 6 (if you can afford it) hardware not software configuration before I could ever recomend WHS again. I am currently moving everything off WHS bad drive extender onto raid arrays. Raids give you the same ability (see all drives as one array) but they do it right/better I won't bore you with the details. I always hated drive extender but bought into the whole WHS thing. When MS killed DE that was my turning point to put it all on raid array;s like I should have in the first place.

[stevech]

Originally Posted by PLUCKYHD

JBOD is just what the name stands for "just a bunch of disks". There is no redundancy or drive pooling.

NTFS mounts under folders...

Then JBOD + software such as SecondCopy would yield the same as MS's abandoned Drive Extender?

I've read many horror stories about consumer-priced RAID-5 failing to cope with a simple drive failure.

[PLUCKYHD]

Originally Posted by jptheripper

I am shocked plucky. I love WHS. Love DE. it just works for me. I have 4 2tb drives on a PCI SATA card, 1 2tb and 2 1,5tb's on the onboard satas, plus a 2x1tb onboard raid for the OS. I have lost a few drives over the year or two i have been running it, and it is rock solid without ever lossing data. Plus it is an all-in-one solution for client backups that is flawless. I lost my HTPC to a failed OS drive and was back up and running on a spare drive in less than an hour.

Just because M$ is dropping DE in VAIL doesnt mean i have to upgrade to it

The backup system of whs is about the only thing of any value to me any more. De is painfully slow at read/write rates. I guess it still will serve a purpose buy to me if ms dropped it they arentgoing to support it all that well anymore. Also to me 1gb to 1gb duplication it too much data for me to give up.

Originally Posted by stevech

NTFS mounts under folders...

Then JBOD + software such as SecondCopy would yield the same as MS's abandoned Drive Extender?

I've read many horror stories about consumer-priced RAID-5 failing to cope with a simple drive failure.

Not to the full extent of de. DE allows you to make any shares duplicated at will if you have the space. So like me for instance I never duplicates mymovies since I could just copy them again. The way you say would still offer no replication.

[panteragstk]

I've been doing a lot of thinking on this exact subject. I like the idea of how unRAID does things. I was really looking into a RAID 5 or 6 setup, but after doing some research I've found that one issue people seem to have is not when the drive dies, but when the RAID card dies. Example: if I purchase a high end areca card with 24 ports and have a RAID 6 array and the card dies I have to find the exact same card or my array is useless. I could be wrong about that and I'd really like to know if that is the case.