Possible to Boot from 4 TB System Volume with BIOS?

[jaclaz]

Jaclaz, I fear using mount points as that would corrupt (or at least complicate) the ability to restore a system image backup.  I don't want to go there for that reason.  And besides, that still would not consolidate the free space into one chunk, which is one of the goals.  It's not about logical organization, it's about maximizing stability and utility.

Sure, I understand your point of view about having the largest possible free chunk and have it in one single place, I mentioned them only in the case (for some strange reasons) you wanted a single, fixed, drive letter.

 

But still you have to (actually you do not "have to", but you well may) realize how your approach (for which you surely you have your very good reasons) is very "uncommon" in the sense that usually people that care about their systems tend to have the OS volume (and possibly even "device") separated from Data.

 

Of course everyone has his/her own ways, and likes and dislikes, but personally I would not even think of having my data scattered over multiple devices in a RAID 0 (which is what you are planning to do if I get it right ).

 

I guess that ReFS and "Storage Spaces" is exactly what you need/want, but still it requires data being separated from boot/system.

 

jaclaz

[NoelC]

I realize my goals are not the norm, but I have very good reasons for them.  I know better than "conventional wisdom" with regard to configuring and running Windows systems.  It comes from a lifetime of experience and an understanding of the architectures that led to modern day Windows.

 

Regarding whether you'd trust RAID 0...  If you don't you don't know what you're missing.  It is very viable with good, high reliability hardware.  I'm already doing it, and have been for years with zero problems.  Backup mitigates the risk from data loss.

 

Most people only dream of the kind of reliability and performance I enjoy.

 

 

-Noel

Edited December 23, 2014 by NoelC

[jaclaz]

I realize my goals are not the norm, but I have very good reasons for them.  I know better than "conventional wisdom" with regard to configuring and running Windows systems.  It comes from a lifetime of experience and an understanding of the architectures that led to modern day Windows.

Let's say that your lifetime experience led you to have non-conventional opinions and approaches to configuring and running windows systems and to their data storage organization, as said no doubts about you having good reasons for your choices.

 

Regarding whether you'd trust RAID 0...  If you don't you don't know what you're missing.  It is very viable with good, high reliability hardware.  I'm already doing it, and have been for years with zero problems.  Backup mitigates the risk from data loss.

Sure .

Though RAID 0 by itself means "little", most of the tricks are in the choice of the "right" high reliability (and high performance) hardware..

 

Most people only dream of the kind of reliability and performance I enjoy.

Here you are definitely wrong , most people cannot even dream about the throughput of a bunch of SSD's in RAID 0.

jaclaz

[NoelC]

Okay, I'm here ONE MORE TIME to try to get a straight answer on whether I should be able to do something very specific...

 

I plan to increase the number of drives in my array from 4 to either 6 or 8.  I have already picked up a drive identical to the ones I already have.  They are now fairly inexpensive.

 

Now the RAID array totals 1920 GB.  When I add the drives and remake the array, the total composite RAID array will be in excess of the 2.2 TB "design limit" for MBR.

 

At least one time it has been said that MBR can create two partitions on a "drive" larger than 2.2 TB as long as the second one starts below the 2.2 TB limit. 

 

Jaclaz mentioned a "suffusion of yellow", which frankly confused me more than cleared things up.

 

Simple YES or NO question:

 

If I create a "drive" that looks like it has 3840 GB on it, using MBR can I make it have two 1920 GB partitions (acknowledging that the second one STARTS before 2.2 TB), and boot Windows 8.1 x64 to one of them?

 

In other words:

 

C: - 1920 GB

D: - 1920 GB

 

Thanks.

 

-Noel

Edited January 12, 2015 by NoelC

[dencorso]

No. No way. Forget about it. -- short answer.

 

Perhaps, in a very contrieved, roundabout, unsupported way, that may stop working at any point in time due to some update. -- long answer

 

TL;DR: Don't hold your breath.

[NoelC]

Thanks, though that's the answer I didn't want to hear.

 

I *could* just increase the total array space from 1920 GB to 2400 GB by adding the one more drive (though a 5 drive array just sounds wrong somehow).  Then I could increase the one big MBR partition to the 2.2 TB max.  That will leave a couple hundred GB of SSD space unprovisioned, but give me 280 GB more space on C: than I have now, as well as upping the performance as there will be one more drive running in parallel.

 

Perhaps my plans to take over the world will have to wait for my next EFI/GPT workstation, Pinky.

 

-Noel

[dencorso]

Or use 6 SSDs and make it RAID 5, instead of 0... it'll be even faster and much safer.

[bphlpt]

Or use 6 SSDs and make it RAID 5, instead of 0... it'll be even faster and much safer.

 

Best of what I understand about the various RAID versions, though never having run any of them myself, RAID 5 is safer than RAID 0, but NOT faster.  I believe RAID 0 is the fastest of all the various RAID versions, though I could be wrong.

 

Cheers and Regards

[dencorso]

The safest RAID is 6, the next safest is 5. Both use block-level striping with distributed parity, but 6 has two independent sets of parity stripes... Hence, during recovery from a single disk loss, 6 has 5 safety (i. e. can still resist the loss of one further disk), while 5 has no further robustness.

 

That said, let's consider RAID 0 vs 5: on a RAID 0 featuring 5 disks, capacity is the sum of the 5 disk and one reads from each in parallel, so that it can be up to 5 times faster than a single disk; now, a RAID 5 made from 6 disks (not 5) has the same capacity as the RAID 0 made from 5 disks, because the addicional capacity (i. e. the 6th disk) is virtually consumed by the parity stripes, but since the parity stripes are distributed over the 6 disks, instead of 5, the 6-disk RAID 5 will be reading from up to 6 disks in parallel, instead of just 5, being thus up to 6 times faster than a single disk. Obviously, when writing, the 6-disk RAID 5 will still be slower than the 5-disk RAID 0, because of the parity computation; but the added reliability can be seen as making it worth it.

 

Of course, NoelC does disagree with me about it, but most everybody else does not.

Then again, the majority opinion isn't necessarily the right one.

[bphlpt]

No argument at all that RAID 0 is the least safe of any of the RAID versions. But as to speed, if I understand what you are saying, then even though "the 6-disk RAID 5 will be reading from up to 6 disks in parallel, instead of just 5," you will still only be getting 5 disks worth of data since "the addicional capacity (i. e. the 6th disk) is virtually consumed by the parity stripes", right? And it will take some infinitesimal amount of time to reassemble the data utilizing the parity information, correct? If that is true, then it seems to me that a 5-disk RAID 0 should still be ever so slightly faster than a 6-disk RAID 5, or at least no slower. But I agree that a 6-disk RAID 5 should be faster than a 4-disk RAID 0.

But again, this is just my understanding from reading and not actual experience, which I fully understand should be more accurate.

Cheers and Regards

[dencorso]

And it will take some infinitesimal amount of time to reassemble the data utilizing the parity information, correct?

 

No. The parity data is updated on every write (thus making it slower), but ignored on reading, unless one of the disks has failed, when it becomes fundamental to recreate the lost disk (and that is a relatively slow process, spanning many hours). However, the parity strpes are evenly dispersed throughout all the 6 disks, so that when reading sequentially, even if one in every 6 blocks read is a (discarded) parity block, one is efectively reading from 6 disks in parallel, which should be somewhat faster than reading from just 5, in the most favorable conditions, and just as fast as reading from just 5-disks in a RAID 0, in the least favorable case. But this applies just for reading. 

[NoelC]

When you get into the gigabytes per second range, everything I've read says that the overhead for creating the parity data is significant.  In other words, you pay with every I/O operation for the safety from failure of a single device.  That being said, I didn't try it myself.

 

The extra overhead is less when the controller is sophisticated and implements the whole RAID management task in hardware.  Mine's not - it's a relatively inexpensive one.  Certainly if I were going to move up to a higher security RAID level (e.g. 5 or 6) I'd want to get a more powerful HBA.  Such devices also come with baggage - they present a greater load on the power supply and produce more heat, for example, both of which tend to reduce reliability.

 

Keep in mind I make very good (nightly) backups for safety, and I now have 2-1/2 years experience with the 4 drive RAID 0, with literally ZERO failures and ZERO downtime.  I realize that's just limited experience in the grand scheme, but with 2 million hour MTBF figures these SSD drives are inherently a good bit less likely to fail than HDDs that people rely upon all the time.  Increasing by one drive isn't going to change the math significantly.

 

-Noel

[jaclaz]

Jaclaz mentioned a "suffusion of yellow", which frankly confused me more than cleared things up.

The reference you miss on "a suffusion of yellow", meet Dirk Gently's holistic calculator:

http://www.thateden.co.uk/dirk/

 

jaclaz

[NoelC]

Thank you, that explains it.  Douglas Adams has never been that high on my list of literary favorites, though he's a hero to many of my technically-oriented friends.  I freely admit that it's my fault.  Perhaps I am missing the part of my personality that it would take to enjoy his particular level of tongue-in-cheek satire (though notably I did enjoy the point of view of Marvin the android).  My aunt Roberta always used to tell me I was too serious.

 

-Noel