My new 600 Petabyte drive

So I inserted the spawn of the devil itself into my unraid server 2 weeks ago. I was doing drive expansion on my unraid, and I inserted this drive to preclear, and lo and behold, it somehow threw a grenade down my sata bus (x10sl7-f mobo w/ hp SAS expander card to 24 slot case). It gave 4-5 of my good drives 'Command Timeout' SMART defect and some 'pending sector' stats to a couple of other drives. I took my array off and did a lot of offline copying and checksumming to verify the drives were ok. Apparently they were all alright. The 'command timeout' parameter never increased on any of the other drives, and the 'pending sector' count reset to zero, without increasing the 'reallocated sector' count. Weird huh??

So anyway, this is the SMART report of the little ****. Check out the 'user capacity' field and how it's ashamed of its own make and model by blanking those fields.

root@Tower:~# smartctl -a -A /dev/sdc
smartctl 6.2 2013-07-26 r3841 [x86_64-linux-3.17.4-unRAID] (local build)
Copyright (C) 2002-13, Bruce Allen, Christian Franke, www.smartmontools.org

=== START OF INFORMATION SECTION ===
Vendor:               /5:0:0:0
Product:              
User Capacity:        600,332,565,813,390,450 bytes [600 PB]
Logical block size:   774843950 bytes
scsiModePageOffset: response length too short, resp_len=47 offset=50 bd_len=46
scsiModePageOffset: response length too short, resp_len=47 offset=50 bd_len=46
>> Terminate command early due to bad response to IEC mode page
A mandatory SMART command failed: exiting. To continue, add one or more '-T permissive' options.

root@Tower:~# smartctl -a -A -T verypermissive /dev/sdc
smartctl 6.2 2013-07-26 r3841 [x86_64-linux-3.17.4-unRAID] (local build)
Copyright (C) 2002-13, Bruce Allen, Christian Franke, www.smartmontools.org

=== START OF INFORMATION SECTION ===
Vendor:               /5:0:0:0
Product:              
User Capacity:        600,332,565,813,390,450 bytes [600 PB]
Logical block size:   774843950 bytes
scsiModePageOffset: response length too short, resp_len=47 offset=50 bd_len=46
scsiModePageOffset: response length too short, resp_len=47 offset=50 bd_len=46
>> Terminate command early due to bad response to IEC mode page

=== START OF READ SMART DATA SECTION ===
Log Sense failed, IE page [scsi response fails sanity test]
Error Counter logging not supported

scsiModePageOffset: response length too short, resp_len=47 offset=50 bd_len=46
Device does not support Self Test logging

That's it, nothing more. If I run badblocks on it, the middle counter increases so very very fast.

Has anyone seen a messed up drive like this? What is wrong with it?

(It's WD RED 4TB, straight from a sealed blister pack).

What about other drives connected to the same controller? Do they all report strange results? You may need to give smartctl a special "-d" parameter to work properly.

What about other drives connected to the same controller? Do they all report strange results? You may need to give smartctl a special "-d" parameter to work properly.

The other drives are 'scarred' that they have a few 'command timeout' bits and some 'realloc event error' or smething, but after running numerous checks on them (including about 4 parity checks since it happened), not a single problem showed up, so I assume those counts were generated due to this 600PB stink box.

I don't think I'll bother trying to extract anymore info from it anymore, I'll just RMA the bastad.

What about other drives connected to the same controller? Do they all report strange results? You may need to give smartctl a special "-d" parameter to work properly.

 

The other drives are 'scarred' that they have a few 'command timeout' bits and some 'realloc event error' or smething, but after running numerous checks on them (including about 4 parity checks since it happened), not a single problem showed up, so I assume those counts were generated due to this 600PB stink box.

 

I don't think I'll bother trying to extract anymore info from it anymore, I'll just RMA the bastad.

200 - 220 days per preclear cycle, lol, go for it..........

200 - 220 days per preclear cycle, lol, go for it..........

I'm pretty sure the unnecessary stress and annoyance from the incident has aged me by at least a couple of years..

Just ask WD for another drive of the same size 

An UnRAID server with 2 600PB drives would give you 600PB of fault-tolerant storage => I suspect that would last you for quite a while  8)

I'd connect it to a motherboard port and see what it shows. The drive should not do that, even if their are issues with the magnetic media. The only time I've seen this type of problem is with an incompatible controller. I guess it is possible that the firmware is hosed, but I have never heard of such a problem.

Just ask WD for another drive of the same size 

 

An UnRAID server with 2 600PB drives would give you 600PB of fault-tolerant storage => I suspect that would last you for quite a while  8)

Heheh put it on a colo, and if you can sell it for just $1 per 10GB (lifetime!), you'll make $60 million dollars.

I'd connect it to a motherboard port and see what it shows. The drive should not do that, even if their are issues with the magnetic media. The only time I've seen this type of problem is with an incompatible controller. I guess it is possible that the firmware is hosed, but I have never heard of such a problem.

That was actually direct from a mobo. I've already moved it to a spare server to test (and trash).

Just ask WD for another drive of the same size 

 

An UnRAID server with 2 600PB drives would give you 600PB of fault-tolerant storage => I suspect that would last you for quite a while  8)

Actually not.

One PB=1000TB.

So a 600PB drive would be 100,000 times larger than a 6TB drive.

Something like a format, that you would expect to take say 1 minute on a 6T drive, would take 70 days to finish on a 600 PB drive.

A disk operation that would take 1 hour on a 6TB drive would take 11.4 years on a 600PB drive. So a zeroing operation (8 hours on 6TB?) would take over 90 years.

I'd say you'd die long before you got an unRAID or even RAID1 configured on 600PB drives.

Actually not.

 

One PB=1000TB.

 

So a 600PB drive would be 100,000 times larger than a 6TB drive.

 

Something like a format, that you would expect to take say 1 minute on a 6T drive, would take 70 days to finish on a 600 PB drive.

 

A disk operation that would take 1 hour on a 6TB drive would take 11.4 years on a 600PB drive. So a zeroing operation (8 hours on 6TB?) would take over 90 years.

 

I'd say you'd die long before you got a RAID1 configured on 600PB drives.

So, are you saying unRAID doesn't scale? 

Just ask WD for another drive of the same size 

 

An UnRAID server with 2 600PB drives would give you 600PB of fault-tolerant storage => I suspect that would last you for quite a while  8)

 

Actually not.

 

One PB=1000TB.

 

So a 600PB drive would be 100,000 times larger than a 6TB drive.

 

Something like a format, that you would expect to take say 1 minute on a 6T drive, would take 70 days to finish on a 600 PB drive.

 

A disk operation that would take 1 hour on a 6TB drive would take 11.4 years on a 600PB drive. So a zeroing operation (8 hours on 6TB?) would take over 90 years.

 

I'd say you'd die long before you got an unRAID or even RAID1 configured on 600PB drives.

Hogwash !!

IF there was indeed a 600PB drive available, it would be due to some significant increase in areal density -- NOT due to building a disk with 600,000 platters !!  Your calculations assume the same rotational rate; same areal density; etc. of the 6TB drive would apply to the 600PB drive -- clearly NOT the case.

By way of historical example, my first hard drive was a 26MB 14" platter unit.  It was a single platter, so clearly it had an areal density of 26MB/platter ... but that was a 14" platter, not a 3.5" platter, so it had 16 times the area of a modern 3.5" platter.    A modern 6TB WD Red has an areal density of 1.2TB/platter ... this is 46,153 times the density of that old 26MB drive ... or an areal density that's 738,461 times that of the 26MB drive (considering the much larger area of the early drive).

Increasing the areal density by a factor of 738,461 is a rather stellar improvement in storage technology 

And that comparison is between two drives where the amount of storage in the later drive was 230,769 times as much as the early drive.  A 600PB drive is "only" 100,000 times larger than a current 6TB unit ... so it's a smaller evolution than what has already occurred  :)

Just ask WD for another drive of the same size 

 

An UnRAID server with 2 600PB drives would give you 600PB of fault-tolerant storage => I suspect that would last you for quite a while  8)

 

Actually not.

 

One PB=1000TB.

 

So a 600PB drive would be 100,000 times larger than a 6TB drive.

 

Something like a format, that you would expect to take say 1 minute on a 6T drive, would take 70 days to finish on a 600 PB drive.

 

A disk operation that would take 1 hour on a 6TB drive would take 11.4 years on a 600PB drive. So a zeroing operation (8 hours on 6TB?) would take over 90 years.

 

I'd say you'd die long before you got an unRAID or even RAID1 configured on 600PB drives.

 

Hogwash !!

 

IF there was indeed a 600PB drive available, it would be due to some significant increase in areal density -- NOT due to building a disk with 600,000 platters !!  Your calculations assume the same rotational rate; same areal density; etc. of the 6TB drive would apply to the 600PB drive -- clearly NOT the case.

 

By way of historical example, my first hard drive was a 26MB 14" platter unit.  It was a single platter, so clearly it had an areal density of 26MB/platter ... but that was a 14" platter, not a 3.5" platter, so it had 16 times the area of a modern 3.5" platter.    A modern 6TB WD Red has an areal density of 1.2TB/platter ... this is 46,153 times the density of that old 26MB drive ... or an areal density that's 738,461 times that of the 26MB drive (considering the much larger area of the early drive).

 

Increasing the areal density by a factor of 738,461 is a rather stellar improvement in storage technology 

And that comparison is between two drives where the amount of storage in the later drive was 230,769 times as much as the early drive.  A 600PB drive is "only" 100,000 times larger than a current 6TB unit ... so it's a smaller evolution than what has already occurred  :)

Hogwash yourself.

You were jokingly saying that a pair of these supposedly 600PB drives would last you forever, and I said they'd be too slow to use even if you had them.

All true.

Your premise that a 600PB disk will become practical with aerial density increases is also flawed. Magnetic spinning disks are NOT going to get us to 600PB. SSDs and successor technologies will lead to some futuristic superconducting storage model that will achieve these higher densities. But we shouldn't be jealous. They'll be contending with 3D holographic media files in the 100s of TB and will be dreaming about 600ZB drives!

I wasn't saying that a 600PB disk would use higher areal density platters ... just that whatever storage media they use will NOT be a result of moving to a huge number of platters.    I suspect that when we get those kind of densities (and we will ... or at least our children will) ... it will be very useable and not take "90 years" to clear 

It's clear this level of density won't be using the magnetic platters of today -- it'll be some other technology that will indeed have an amazing areal density on whatever the medium may be ... whether it's solid state; optical; or whatever.

We've seen areal density increase by a factor of 738,461 on magnetic media alone in the past 34 years ... to see another 100,000 times increase in the future -- likely on some other media -- is hardly a big stretch.

I wasn't saying that a 600PB disk would use higher areal density platters ... just that whatever storage media they use will NOT be a result of moving to a huge number of platters.    I suspect that when we get those kind of densities (and we will ... or at least our children will) ... it will be very useable and not take "90 years" to clear 

 

It's clear this level of density won't be using the magnetic platters of today -- it'll be some other technology that will indeed have an amazing areal density on whatever the medium may be ... whether it's solid state; optical; or whatever.

 

We've seen areal density increase by a factor of 738,461 on magnetic media alone in the past 34 years ... to see another 100,000 times increase in the future -- likely on some other media -- is hardly a big stretch.

That would be 100,000x improvement from where we are today. After a 738,461x improvement from whatever starting point you are declaring. That would mean you are expecting us to move from 738,461x to 73,846,100,000x is not a stretch. I think it is.

We'll argue about it in the afterlife. (Assuming we both go to the same place )

FWIW, since Gordon Moore stated his hypothesis in 1971 that transistor counts in integrated circuits would double every year for the foreseeable future, that has proven true for over 50 years ==> the current gain in transistor count since he stated that is well over 5,000,000,000 times the 1971 starting point.

While clearly areal density on magnetic media doesn't grow at this rate (It doesn't even quite reach the projections by Mark Kryder known as "Kryder's Law"),  the general idea is the same -- that the capabilities of electronics are growing exponentially.

I do NOT think it's a stretch that over the next 30-40 years capacities of storage devices will be in the PB range.  It's been 34 years since I bought that 26MB drive.    At the same time, Seagate was selling 5 1/4" hard drives with a capacity of 5MB.    I suspect most folks during that time would have thought it'd be a real stretch to imagine that a significantly smaller drive could hold 1,600,000 times as much for less than 1/15th of the price ... yet that's exactly what a modern 8TB drive does