HDDs for backups

Elledan

[H]ard|DCer of the Month - April 2010
Joined
Oct 18, 2001
Messages
15,913
We use magnetic media (FD, Zip, tape, etc.) and optical media (CD-R(W), DVD+/-R(W), DVD-RAM, etc.) on a (hopefully) daily base to make backups of important data. However, the data capacity of those technologies has lagged behind that of HDDs. No one in his right mind would use CD-Rs to make a backup of the 200 GBs of data on a system. Even DVDs lack sufficient capacity for this task, requiring multiple disks.

Tape storage is affordable only by individuals and companies with a sufficiently large budget, or those willing to settle for an older generation of tape storage, in the latter case negating the advantage tape storage has over other storage methods in regard to capacity. Among the disadvantages of tape, aside from price, is the speed (it's not random-access capable).

Moreover, CDs, DVDs, FDs, tape and similar are quite fragile. The dye used for CDs and DVDs consists out of organic compounds, which are everything but stable, as well as the target of at least one type of fungus (which thrives in warm, humid climates).
At the speeds with which CDs and DVDs are read in modern drives, even the tiniest hairline fracture may cause the disc to desintegrate, as well as the risk of scratching or in another way damaging either side of the disc, which might render (parts of) it unreadable.

FDs can stand a lot more abuse, but with some effort the magnetic disc inside the outer shell is warped or damaged to the point where it becomes unusable. Tape is more durable in this regard, but if it were to break at one point, it would be a very costly operation to repair this tape.

HDDs, on the other hand, are, when not powered, virtually indestructable when compared to the aforementioned storage technologies. The platters storing the data are spinning inside a sealed, metal shell, and in modern drives the read/write heads are unlikely to crash into the platters, especially so when the HDD is turned off.

The only ways to destroy a non-powered HDD is to either drop it on a hard surface with considerable force (over 10 Gs), so as to dislodge some of the internal components, or attack it with another brute force method, aimed at destroying the outer shell and the parts inside. Well, that and pouring corrosive acids on it. Frying the electronics of the HDD has little use, as this would not affect the data stored on the platters. Using a strong magnetic field to scramble the platter's magnetic patterns might also work, but requires a very strong field, something which is unlikely to occur purely by accident.

So, based on the above comparison, isn't it reasonable to state that HDDs are the ideal backup medium? Moreso if one looks at the $/GB comparison of currently available storage technologies.

Naturally, the HDDs used for backups are only to be powered when data is read from/written to them.

Any comments?
 
i used to work at a company where we had this server that needed to be backed up everyday. at night, someone would have to load a tape cartridge in and then run this backup software that more often than not, stopped half way through the backup.

i got to thinking there had to have been an easier way of doing this and you came up with the same idea. why not use a hard drive to back up large amounts of data? it would be much easier in my opinion, to setup a 'ghosting station'. but of course, this probably isnt the best solution in all cases.

just me 2 cents
 
I would think that this has always been the case... though I'm only 15... I don't even know what a tape backup is, lol.

Why would anyone not want to use HDDs as backups in the 1st place?
 
Originally posted by pbXassassinX1524
I would think that this has always been the case... though I'm only 15... I don't even know what a tape backup is, lol.

Why would anyone not want to use HDDs as backups in the 1st place?

Mainly because of the cost/Mb.
Also, if you do full backups now and then, and incremental backups (requiring much less space) every day, you either have to use a lot of HDs (waste), or use the same one for several days' backups (stupid).

I'm not saying that it's a bad idea, though, but tape does make sense in some situations.
(Besides, you get a lot of data on really expensive tape.)
 
Originally posted by HHunt
Mainly because of the cost/Mb.
Also, if you do full backups now and then, and incremental backups (requiring much less space) every day, you either have to use a lot of HDs (waste), or use the same one for several days' backups (stupid).
HDDs are relatively cheap, so it's not that crazy to use multiple HDDs. For the price of a single tape you can buy an HDD of equivalent size, without even taking in consideration the need to buy the streamer to use the tapes with (hundreds of US$).

I'm not saying that it's a bad idea, though, but tape does make sense in some situations.
(Besides, you get a lot of data on really expensive tape.)
Really expensive, yes, which is the reason why you see so few people use tape for backups.

I know that tape is used by many companies and some individuals, and I know that it works, but I fail to see why one should continue to use tape if HDDs are much cheaper per MB and arguably more reliable.
 
Originally posted by Elledan
HDDs are relatively cheap, so it's not that crazy to use multiple HDDs. For the price of a single tape you can buy an HDD of equivalent size, without even taking in consideration the need to buy the streamer to use the tapes with (hundreds of US$).


Really expensive, yes, which is the reason why you see so few people use tape for backups.

I know that tape is used by many companies and some individuals, and I know that it works, but I fail to see why one should continue to use tape if HDDs are much cheaper per MB and arguably more reliable.

Well, the tape media in itself isn't that expensive, it's the streamers that makes it insane.

Anyway. I guess it s a rather new thing that harddrives are cheaper/mb than tape, so most of the tape backup systems was created before this happened. It's also a mindset thing, lost of people are used to tape==backup.

Besides, keep in mind that tape has no movable parts [1], so theoretically it should last until it chemically detoriates.

I'd say that hotswap-carriages and harddrives still would be a very good solution for most, but not all, places that use tape today, and a lot that doesn't.
The ones where I wouldn't replace tape with HDs are where you need as high safety as possible, because a tape (being only media) can survive things a HD (Being complicated electronics and mechanics) can not. EMP pulses spring to mind :D



[1] Well, they do, but they are simple, and at worst replacable, parts.
The theory is that with tape you move most points of failure to the streamer, which is replacable.
 
Originally posted by HHunt
Well, the tape media in itself isn't that expensive, it's the streamers that makes it insane.
True. I pointed out in an earlier post that the price per MB for tape and HDDs is about the same, as long as you ignore the price of a streamer.

Anyway. I guess it s a rather new thing that harddrives are cheaper/mb than tape, so most of the tape backup systems was created before this happened. It's also a mindset thing, lost of people are used to tape==backup.
Good point. Things certainly have changed since 10 years or so ago, haven't they? :)

Besides, keep in mind that tape has no movable parts [1], so theoretically it should last until it chemically detoriates.
Well, that's the theory. In reality there's the issue of the tape 'stretching' and similar, whereby the length of the tape changes, making it difficult or impossible to read the data stored on it. Further, while little can go wrong when the tape is just stored in a climate-controlled room, when it's being read, the tape is forced through a rather torturous construction in order to get the tape near the read/write heads.

So, while tape does last quite a long time in storage (in the right conditions), it can only be rewritten quite a limited number of times after which it becomes too unreliable to be used again.

Keep in mind that tapes are basically just audio tapes on steroids ;)

I'd say that hotswap-carriages and harddrives still would be a very good solution for most, but not all, places that use tape today, and a lot that doesn't.
The ones where I wouldn't replace tape with HDs are where you need as high safety as possible, because a tape (being only media) can survive things a HD (Being complicated electronics and mechanics) can not. EMP pulses spring to mind :D
I wouldn't be so certain. If the EMP is capable of scrambling the data on the HDD's platters, it most certainly can scramble the data on tapes. If the controller board of the HDD is fried, it can be replaced, or the data on the platters recovered without any loss of data.
 
Good point. Things certainly have changed since 10 years or so ago, haven't they? :)
Indeed. I'm even old enough to remember these days (It's a sign of the times when you speak of "the old days" and 1) it's 10 years ago and 2) you're 20 now :D )


Well, that's the theory. In reality there's the issue of the tape 'stretching' and similar, whereby the length of the tape changes, making it difficult or impossible to read the data stored on it. Further, while little can go wrong when the tape is just stored in a climate-controlled room, when it's being read, the tape is forced through a rather torturous construction in order to get the tape near the read/write heads.
So, while tape does last quite a long time in storage (in the right conditions), it can only be rewritten quite a limited number of times after which it becomes too unreliable to be used again.

True. Makes tape ok for archive-type storage, though.

Keep in mind that tapes are basically just audio tapes on steroids ;)
Good steroids, but yes :D
(I wonder just how much you could get on a 70s-style data tape reel with modern read/write-equipment?)

I wouldn't be so certain. If the EMP is capable of scrambling the data on the HDD's platters, it most certainly can scramble the data on tapes. If the controller board of the HDD is fried, it can be replaced, or the data on the platters recovered without any loss of data.
I was mostly thinking about the electronics. Keep in mind that HD controller boards aren't really meant to work with anything but the exact drive model and revision they were shipped with, while tape streamers are meant to be slightly more compatible.
("Meant to be" :D )
 
The Medium is not the massage or Indefinately long term storage for Leeds University
Hard Drives Instead of Tapes? 70 TB Backup RAID at the University of Tübingen
CDR Archival Storage
CDR Media Longevity
Handle with care Video rm
Proper Handling Guide


IMO

HDD are more fragile than you have credited
CDR\DVD more robust than you have credited
Data Loss from infection or Power Issues more pervasive
but yes with proper care and infrastructure, its being done
but that includes UPS and Power Filtering as well as serious security and RAID
 
depending on what your budget is why not go with a NAS setup that is raided and be done with it..that way all the data is saved at a single location and backed up on the fly...they are a little expensive (or very depending on what you get) but you can get into the multi TB range with them
 
Originally posted by Ice Czar

IMO

HDD are more fragile than you have credited
In what sense? I already pointed out that it's relatively easy to destroy the controller board. I also looked at non-powered (off) HDDs only, in which case the issue of a head crash is more than just unlikely.
CDR\DVD more robust than you have credited
CDs and DVDs can last a long time in storage, but this is only valid for the more expensive discs which use dyes which remain stable for 100-200 years (in theory) in the right conditions. As pointed out on one of the pages you linked to, sunlight is one of the worst things which can happen to a CD, as the energy contained in sunlight destabilizes the organic components in the dye. CD-RWs are explicitly refered to as only being suitable for short-term storage, making them unsuitable for backups.

I may have exaggerated some things while trying to get the point across, but everything I wrote is factually correct, to the best of my knowledge.
Data Loss from infection or Power Issues more pervasive
but yes with proper care and infrastructure, its being done
but that includes UPS and Power Filtering as well as serious security and RAID
I assume that you're referring to active arrays of HDDs, like in the article at Tom's Hardware to which you linked?
 
Originally posted by gigglebyte
depending on what your budget is why not go with a NAS setup that is raided and be done with it..that way all the data is saved at a single location and backed up on the fly...they are a little expensive (or very depending on what you get) but you can get into the multi TB range with them
I have little to no experience with NAS. Do you have any good examples of it?
 
even with the head parked, Id say it takes less than 10G
FDB HDDs and smaller form factor being better
(Seagate is introducing 2.5" enterprise storage in a few months)

I was definately refering to Phthalocyanine dye CDR or DVDR
if its for archival backup, buying in bulk makes it far more affordable

a NAS or SAN would typically be always on, but Ive had great reliability with my IDE RAID 5, through 3 years of intermittent power
I still back it up however to hard media
Are SCSI drives more reliable than ATA drives? < a consideration


Originally posted by Elledan
I have little to no experience with NAS. Do you have any good examples of it?

NAS | SAN linkage
http://www.enterprisestorageforum.com/technology/features/article.php/947551
http://www.extremetech.com/search_results/0,3964,,00.asp?qry=SAN&site=ExtremeTech
http://www.extremetech.com/search_results/0,3964,,00.asp?qry=NAS&site=ExtremeTech
http://www.enterprisestorageforum.com/sans/features/article.php/2195011
http://www.networkcomputing.com/1109/1109ws1.html
http://www.enterprisestorageforum.com/sans/features/article.php/1583811
http://www.enterprisestorageforum.com/sans/features/article.php/1467691
 
Originally posted by Ice Czar
even with the head parked, Id say it takes less than 10G
Seagate claims 63 Gs and 350 Gs (operating and non-operating, respectively): http://www.seagate.com/docs/pdf/datasheet/disc/ds_barracuda7200.7.pdf

Of course, the angle at which the impact occurs will play a role in the amount of Gs necessary to inflict damage.
FDB HDDs and smaller form factor being better
(Seagate is introducing 2.5" enterprise storage in a few months)
Less mass == less inertia?

SCSI drives may be more reliable, but as long as one stays away from the bleeding edge, one can buy a number of IDE drives for the price of a single SCSI drive, allowing one to have more copies of the same set of data and thus effectively making the chance on data loss even smaller.


Thanks! At least I'll have something to read tonight :p
 
Originally posted by Elledan
Less mass == less inertia?
yup plus
smaller platter more rigid less flex, in the head arm as well
Enhanced Rigidity: Smaller drives use smaller platters, which are less susceptible to damage as a result of shock, always a concern for a drive that will be moved around (often while operating!)
2.5" drives are pretty much entrenched as the standard for laptop machines. They are also used occasionally in industrial applications, where the smaller size and increased ruggedness of portable drives is important.
as well spin at a higher speed (less mass)
Platter Size
Enhanced Rigidity: The rigidity of a platter refers to how stiff it is. Stiff platters are more resistant to shock and vibration, and are better-suited for being mated with higher-speed spindles and other high-performance hardware. Reducing the hard disk platter's diameter by a factor of two approximately quadruples its rigidity.

Mass Reduction: For performance reasons, hard disk spindles are increasing in speed. Smaller platters are easier to spin and require less-powerful motors. They are also faster to spin up to speed from a stopped position.

Improved Seek Performance: Reducing the size of the platters reduces the distance that the head actuator must move the heads side-to-side to perform random seeks; this improves seek time and makes random reads and writes faster. Of course, this is done at the cost of capacity; you could theoretically achieve the same performance improvement on a larger disk by only filling the inner cylinders of each platter. In fact, some demanding customers used to partition hard disks and use only a small portion of the disk, for exactly this reason: so that seeks would be faster. Using a smaller platter size is more efficient, simpler and less wasteful than this sort of "hack".
Seagate 10krpm 2.5"ers



SCSI drives may be more reliable, but as long as one stays away from the bleeding edge, one can buy a number of IDE drives for the price of a single SCSI drive, allowing one to have more copies of the same set of data and thus effectively making the chance on data loss even smaller.

actually I linked that for the >
"Server usage patterns differ from desktop usage patterns in that the former expects 24/7 operation, with minimal power cycles and many seeks, while the latter is close to the opposite. As such, SCSI drives are typically tested (and designed) for 24/7 operation, but a relatively low number of power cycles as compared to IDE drives, and vice versa.
 
Originally posted by Elledan
So 2.5" HDDs are pretty much better in every respect when compared to 3.5" HDDs?
except capacity, at whatever level of current areal density, a larger platter will always hold more, in addition while access times drop, the lmissing tracks on the outside of a larger platter will have a higher density passing under the head for the same rotational speed (better Sustained Transfer Rate) in reality however, because its smaller it can spin faster, so instaed of a 7200rpm its upped to a 10000rpm, offsetting that loss


If IDE HDDs are designed for frequently being turned on/off, then it makes them the perfect choice for off-line backups, doesn't it? :)

Indeed
 
Originally posted by Ice Czar
except capacity, at whatever level of current areal density, a larger platter will always hold more, in addition while access times drop, the lmissing tracks on the outside of a larger platter will have a higher density passing under the head for the same rotational speed (better Sustained Transfer Rate) in reality however, because its smaller it can spin faster, so instaed of a 7200rpm its upped to a 10000rpm, offsetting that loss
So STRs should be equal or better than those of 3.5" HDDs. What about manufacturing costs? Will 2.5" HDDs be cheaper to manufacture or is the difference insignficant?

Also, how long will it take for 2.5" HDDs to reach the data storage capacity of 3.5" HDDs?
 
well they will never match 3.5" form factor drives per se
that is in a drive by drive, platter by platter comparision,
(for the largest capacities)
the measure is in storage density in a rackmount where thay are already equal
to similar performing drives
and at less power\heat per square inch,
the primary advantage for enterprise storage adoption

manufacturing costs will be initially higher until volume picks up
like any new model introduction

the initial release is a 36GB Platter, in a single and double platter version, at 10,000 rpm, Serial Attached SCSI
 
Originally posted by Ice Czar
manufacturing costs will be initially higher until volume picks up
like any new model introduction

the initial release is a 36GB Platter, in a single and double platter version, at 10,000 rpm
And how does a HDD using those platters compare in terms of $/MB with, say, SATA drives? How will this change in the future (how low will it go? =) )?
 
Originally posted by Elledan
And how does a HDD using those platters compare in terms of $/MB with, say, SATA drives? How will this change in the future (how low will it go? =) )?

poorly for desktop, would be my guess
if your running a data center however reduced cooling and power factor large in that equation
"the Savvio drives use 40 percent less power"
probably near the same in heat generation

a major cost of ownership\operation factor if your running a stack of blade servers

"Storage arrays with Savvio drives would allow 70 percent more input-output operations per second in a smaller space than many existing arrays, Seagate said."

plus I dont really see desktop adoption of SAS (Serial Attached SCSI) the whole SATA vs SAS comparision is shaping up as a true dividing line between desktop and enterprise use, just dont see any aps that need that extra bandwidth on the desktop, near impossible to saturate the current 64bit PCI bus now with the current crop of desktop aps. SATA is making great inroads on the typical SCSI territory, and 64bit SATA is readilly available, if the access is really needed for realtime editing or something the current SCSI standards are upto it, but with 64bit computing in the works that will eventually change, but until those aps are in place...

looks like Maxtor and Hitachi will also be introducing models within a year or so


*desktop = single user workstation :p in this case
 
Originally posted by Ice Czar
poorly for desktop, would be my guess
if your running a data center however reduced cooling and power factor large in that equation
"the Savvio drives use 40 percent less power"
probably near the same in heat generation
Considering that the IDE drives (5400 and 7200 RPM) in my systems would die a fiery death if it weren't for the airflow forced over them I consider this to be a really good thing (combined with the lower electricity bill ;) ).

OT: imagine the coolness (literally) of a solid-state 'HDD' based on MRAM or similar. Now that's something all big companies with large arrays of HDDs would buy instantly.


plus I dont really see desktop adoption of SAS (Serial Attached SCSI) the whole SATA vs SAS comparision is shaping up as a true dividing line between desktop and enterprise use, just dont see any aps that need that extra bandwidth on the desktop, near impossible to saturate the current 64bit PCI bus now with the current crop of desktop aps. SATA is making great inroads on the typical SCSI territory, and 64bit SATA is readilly available, if the access is really needed for realtime editing or something the current SCSI standards are upto it, but with 64bit computing in the works that will eventually change, but until those aps are in place...
IDE is dead, long live IDE? :)
 
Back
Top