What is JBOD?

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

Eli wrote:

> More correctly, how is JBOD typically implmented? Are all
> implementations similar or is it open to a manufacturer's
> interpretation?
>
> My understanding of JBOD is that it permits a volume that spans the
> drives.
>
> But is JBOD implemented by writing data across the drives? That would
> be RAID 0 if no redundancy is offered.
>
> If data is _not_ written across drives, then what kind of algorithms
> are used to decide where a file is written?

RAID 0 stripes the data across the disks to gain a performance improvement.
JBOD just tacks one disk onto the end of another without striping. There's
minimal performance improvement, the only significant gain is that you have
one big volume instead of a bunch of little ones.

--
--John
to email, dial "usenet" and validate
(was jclarke at eye bee em dot net)

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

"Eli" <nospam@thanks.com> wrote in message
news:36m8d1d22mh69ete8arjp11gbt3s9sb0rs@4ax.com...
> More correctly, how is JBOD typically implmented?


Just a Bunch Of Disks(that includes just one). The way all standard single
or multiple disks are connected to a PC is JBOD. AKA not RAID and not
spanned.

> Are all
> implementations similar or is it open to a manufacturer's
> interpretation?


Basic ATA spec and basic SCSI spec and they mostly work interchangeably.

> My understanding of JBOD is that it permits a volume that spans the
> drives.

NO! That's spanning and not JBOD.

> But is JBOD implemented by writing data across the drives? That would
> be RAID 0 if no redundancy is offered.
>
> If data is _not_ written across drives, then what kind of algorithms
> are used to decide where a file is written?


You got a bum steer on the definition of JBOD.

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

On Wed, 13 Jul 2005 00:10:16 GMT, "Ron Reaugh"
<ron-reaugh@worldnet.att.net> wrote:

>
>"Eli" <nospam@thanks.com> wrote in message
>news:36m8d1d22mh69ete8arjp11gbt3s9sb0rs@4ax.com...
>> More correctly, how is JBOD typically implmented?
>
>
>Just a Bunch Of Disks(that includes just one). The way all standard single
>or multiple disks are connected to a PC is JBOD. AKA not RAID and not
>spanned.
>
>> Are all
>> implementations similar or is it open to a manufacturer's
>> interpretation?
>
>
>Basic ATA spec and basic SCSI spec and they mostly work interchangeably.
>
>> My understanding of JBOD is that it permits a volume that spans the
>> drives.
>
>NO! That's spanning and not JBOD.
>
>> But is JBOD implemented by writing data across the drives? That would
>> be RAID 0 if no redundancy is offered.
>>
>> If data is _not_ written across drives, then what kind of algorithms
>> are used to decide where a file is written?
>
>
>You got a bum steer on the definition of JBOD.


Here's at least one explanation of JBOD that claims it's the
equivalent of spanning.

http://www.pcguide.com/ref/hdd/perf/raid/levels/jbod.ht...

Beyond the petty semantics, are there any standards for the
implementation of spanning? Would I be correct in assuming that files
are not written across physical drives in such an arrangement?

Ultimately what I'm trying to figure out is whether there are any
advantages to JBOD (spanning) vs RAID 0. I'm looking at building a
small (1/2 TByte) drive array in which neither speed nor redundancy
are important. But I _would_ like it to be addressable as a single
drive. RAID 5 would be my first choice, but I can only have four
drives in the array and don't want to give up 25% of the available
capacity.

Normally, I'd never use RAID 0 for storing data, but this data is
static and is well backed up on other media. Still, it would be
time-consuming to have to restore the array due the failure of a
single drive.

If a JBOD/spanning set were set up, would there be any advantages over
RAID 0? Say I lose a drive and lose only the files on that drive.
That's great. I still have the rest of the array running and only
need to restore the single drive from backup media. But... How do
you know what data was actually lost? Are you then in exactly the
same position as if you'd used a RAID 0 array - that of having to
restore then entire volume?

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

Previously Eli <nospam@thanks.com> wrote:
> More correctly, how is JBOD typically implmented? Are all
> implementations similar or is it open to a manufacturer's
> interpretation?

> My understanding of JBOD is that it permits a volume that spans the
> drives.

> But is JBOD implemented by writing data across the drives? That would
> be RAID 0 if no redundancy is offered.

No. RAID0 does interleaving, i.e. small slices of space are
taken from the drives in a round-robin fashion (it can be done
with more than 2 drives). JBOD is just the spaces appended.

> If data is _not_ written across drives, then what kind of algorithms
> are used to decide where a file is written?

Take the sectors of the first dosk, then the second , then the third,
...., and renumber linearly. JBOD is about the least reliable option
besides RAID0 and as slow as a single disk in many cases. Its only
reason for existing is that it allows disks of different size to be
combined. I would stay well clear of it. Disks are unreliable enough
by themselves.

Arno

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

On 13 Jul 2005 01:22:29 GMT, Arno Wagner <me@privacy.net> wrote:

>Previously Eli <nospam@thanks.com> wrote:
>> More correctly, how is JBOD typically implmented? Are all
>> implementations similar or is it open to a manufacturer's
>> interpretation?
>
>> My understanding of JBOD is that it permits a volume that spans the
>> drives.
>
>> But is JBOD implemented by writing data across the drives? That would
>> be RAID 0 if no redundancy is offered.
>
>No. RAID0 does interleaving, i.e. small slices of space are
>taken from the drives in a round-robin fashion (it can be done
>with more than 2 drives). JBOD is just the spaces appended.
>
>> If data is _not_ written across drives, then what kind of algorithms
>> are used to decide where a file is written?
>
>Take the sectors of the first dosk, then the second , then the third,
>..., and renumber linearly. JBOD is about the least reliable option
>besides RAID0 and as slow as a single disk in many cases. Its only
>reason for existing is that it allows disks of different size to be
>combined. I would stay well clear of it. Disks are unreliable enough
>by themselves.

Are files (potentially) split across disks?

Even if they are not, from a practical standpoint, would JBOD offer
any advantage overy RAID0? If a single drive fails then you might
still access the remaining data on the other disks, but I'm wondering
how in the world you'd restore the missing files. Seems like you're
nearly in the same boat as if the array were using RAID0 and that you
could only reliably restore the missing data by restoring the entire
volume.

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

"Eli" <nospam@thanks.com> wrote in message news:36m8d1d22mh69ete8arjp11gbt3s9sb0rs@4ax.com...
> More correctly, how is JBOD typically implmented? Are all
> implementations similar or is it open to a manufacturer's
> interpretation?
>
> My understanding of JBOD is that it permits a volume that spans the
> drives.

Yup, it's a virtual physical disk consisting of several concatenated physical disks.
http://www.wordiq.com/definition/Redundant_array_of_ind...

>
> But is JBOD implemented by writing data across the drives?

Obviously, but not necessarily at the same time unless a transfer
spans an area where one drive takes over from the previous one
or with multiple IOs each pointing to a different disk.

> That would be RAID 0 if no redundancy is offered.

Nope. JBOD is concatenated, not striped.

>
> If data is _not_ written across drives,

Which can mean anything.

> then what kind of algorithms are used to decide where a file is written?

As with all, destination block address.

>

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

Previously Eli <nospam@thanks.com> wrote:
> On 13 Jul 2005 01:22:29 GMT, Arno Wagner <me@privacy.net> wrote:

>>Previously Eli <nospam@thanks.com> wrote:
>>> More correctly, how is JBOD typically implmented? Are all
>>> implementations similar or is it open to a manufacturer's
>>> interpretation?
>>
>>> My understanding of JBOD is that it permits a volume that spans the
>>> drives.
>>
>>> But is JBOD implemented by writing data across the drives? That would
>>> be RAID 0 if no redundancy is offered.
>>
>>No. RAID0 does interleaving, i.e. small slices of space are
>>taken from the drives in a round-robin fashion (it can be done
>>with more than 2 drives). JBOD is just the spaces appended.
>>
>>> If data is _not_ written across drives, then what kind of algorithms
>>> are used to decide where a file is written?
>>
>>Take the sectors of the first dosk, then the second , then the third,
>>..., and renumber linearly. JBOD is about the least reliable option
>>besides RAID0 and as slow as a single disk in many cases. Its only
>>reason for existing is that it allows disks of different size to be
>>combined. I would stay well clear of it. Disks are unreliable enough
>>by themselves.

> Are files (potentially) split across disks?

Yes. The filesystem does not see were one disk ends and the other
starts. The RAID-drivers present the array as a single disk to the
higher OS layers. File fragmentation makes the effect worse, but
even large files may reside completely on one disk, contrary to
RAID0 where only files smaller than the stripe-size will get
stored completely on one disk.

> Even if they are not, from a practical standpoint, would JBOD offer
> any advantage overy RAID0?

Only that you don't loose space when disks are not the same size.
Nothing else. And it is conceptually simpler, so people that are afraid
or unwillinf to understand technology can just have several small
disks be magically combined into one larger one without understanding
what really happens. Of course they will also not understand the
risks. Watching the present tragedy with writable DVDs, I don't think
the storage industry cares....

> If a single drive fails then you might
> still access the remaining data on the other disks, but I'm wondering
> how in the world you'd restore the missing files.

You will likely loose metadata that was on the failed disk. That can
mean anything from lost directories to the remaining disks just turning
into a pool of sectors that are hard or impossible to attribute to
specific files or even tell whether they belong to a file at all.

> Seems like you're
> nearly in the same boat as if the array were using RAID0 and that you
> could only reliably restore the missing data by restoring the entire
> volume.

I think it is comparable. With RAID0 you loose about half of every
file in case of a lost disk. With JBOD you loose some files completely
a few parially and some are still present. Whether you can assemble them
from their parts is another question. Unless you are willing to spend
a lot of effort or money, I would say that a JBOD with a failed disk is
just as dead as a RAID0 with a failed disk.

Advice: Personally I have everything important on RAID1 and I have
backups nonetheless. Replacable stuff goes onto single native
partitions. JBOD is just a bad, bad idea IMO.

Arno

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

Ron Reaugh <ron-reaugh@worldnet.att.net> wrote
> Eli <nospam@thanks.com> wrote

>> More correctly, how is JBOD typically implmented?

Normally done at the OS level, called dynamic disks in NT/2K/XP.

> Just a Bunch Of Disks(that includes just one).

Wrong, as always.

> The way all standard single or multiple disks are connected to a PC is JBOD.
> AKA not RAID and not spanned.

Wrong, as always
http://www.pcguide.com/ref/hdd/perf/raid/levels/jbod.ht...

>> Are all implementations similar or is it open to a manufacturer's
>> interpretation?

The way the OS does it.

> Basic ATA spec and basic SCSI spec and they mostly work interchangeably.

Not a clue, as always.

>> My understanding of JBOD is that it permits a volume that spans the drives.

Correct.

> NO! That's spanning and not JBOD.

Not a clue, as always.

>> But is JBOD implemented by writing data across the drives?

Nope.

>> That would be RAID 0 if no redundancy is offered.

>> If data is _not_ written across drives,

It isnt.

>> then what kind of algorithms are used to decide where a file is written?

> You got a bum steer on the definition of JBOD.

Only from you, as always.

Archived from groups: comp.sys.ibm.pc.hardware.storage (More info?)

Eli <nospam@thanks.com> wrote
> Ron Reaugh <ron-reaugh@worldnet.att.net> wrote
>> Eli <nospam@thanks.com> wrote

>>> More correctly, how is JBOD typically implmented?

>> Just a Bunch Of Disks(that includes just one). The way all
>> standard single or multiple disks are connected to a PC is JBOD.
>> AKA not RAID and not spanned.
>>
>>> Are all
>>> implementations similar or is it open to a manufacturer's
>>> interpretation?
>>
>>
>> Basic ATA spec and basic SCSI spec and they mostly work
>> interchangeably.
>>
>>> My understanding of JBOD is that it permits a volume that spans the
>>> drives.
>>
>> NO! That's spanning and not JBOD.
>>
>>> But is JBOD implemented by writing data across the drives? That
>>> would be RAID 0 if no redundancy is offered.
>>>
>>> If data is _not_ written across drives, then what kind of algorithms
>>> are used to decide where a file is written?

>> You got a bum steer on the definition of JBOD.

> Here's at least one explanation of JBOD
> that claims it's the equivalent of spanning.

And all the others that show up using google do too.

> http://www.pcguide.com/ref/hdd/perf/raid/levels/jbod.ht...

> Beyond the petty semantics, are there any
> standards for the implementation of spanning?

Nope.

> Would I be correct in assuming that files are not
> written across physical drives in such an arrangement?

Thats up to the OS that implements it. Usually they do.

> Ultimately what I'm trying to figure out is whether there
> are any advantages to JBOD (spanning) vs RAID 0.

Yes, if you lose a physical drive, its not as drastic as with RAID0,
you should be able to recover most of the data on the surviving drive.

And RAID0 is much worse when the drives are quite different capacitys too.

RAID can be better with some of the cruder backup
software tho. Some dont support XP dynamic disks.

> I'm looking at building a small (1/2 TByte) drive array in
> which neither speed nor redundancy are important. But
> I _would_ like it to be addressable as a single drive.

Yes, that has real advantages, particularly free space
doesnt get scattered across the physical drives and its
much easier to implement stuff like digital TV capture etc,
it just keeps going until there is no free space left at all.

> RAID 5 would be my first choice, but I can only have four drives
> in the array and don't want to give up 25% of the available capacity.

> Normally, I'd never use RAID 0 for storing data, but
> this data is static and is well backed up on other media.

Then spanning would work fine.

> Still, it would be time-consuming to have to
> restore the array due the failure of a single drive.

> If a JBOD/spanning set were set up, would
> there be any advantages over RAID 0?

Yes, see above.

> Say I lose a drive and lose only the files on that drive. That's great.

Yeah, its better in that respect.

> I still have the rest of the array running and only
> need to restore the single drive from backup media.
> But... How do you know what data was actually lost?

Yeah, that can get a bit tricky.

> Are you then in exactly the same position as if you'd used a
> RAID 0 array - that of having to restore then entire volume?

Thats certainly the simplest and most automatic approach.