Talk:Hard disk drive/Archive 12

From Wikipedia, the free encyclopedia
Jump to: navigation, search
Archive 11 Archive 12 Archive 13

Disks replacing tape

Magnetic tape was the dominant secondary store through most of the 1960s, and into the 70s. Early disks were expensive and had nowhere near the capacity needed for most business applications. Most data processing was batch, not real time, and tape densities kept improving from 200 bytes per inch to 6250 bpi, the latter allowing up to 180 megabytes on a single 2400 foot reel of tape. I don't think a 2004 Toshiba press release is a reliable source to the contrary.--agr (talk) 11:48, 19 October 2012 (UTC)

It's not clear to me what you are referring to in this talk section. I agree with your observations except that I would note there is a substantial semantics difference between "secondary store" as ordinary technical terms and secondary storage as a term of the art. We might argue whether or not in the early diskless systems tape was secondary storage in that heirarchy or not, but it seems pretty clear to me that with the advent of disk operating systems, tape became off-line storage. Tom94022 (talk) 16:49, 19 October 2012 (UTC)
Sorry, somehow I missed the latest edit to the lede to which I will respond directly. Tom94022 (talk) 16:55, 19 October 2012 (UTC)

──────────────────────────────────────────────────────────────────────────────────────────────────── Arnold, will you agree to my returning to the previous wording not mentioning tape at all and including continuous improvements? Something like:

"Introduced by IBM in 1956, HDDs have been the dominant secondary storage device in general purpose computers since the early 1960s. They have maintained this position thru continuous improvements at a rate not unlike Moore's Law. More than 200 companies ..."

FWIW, the sentance on home computing is redundant since gp computing includes home computing so I think it can be stricken. I ask because Thumperwald seems to have a strong ownership of his specific edits so I hesitate to make any change without getting agreement from at least one other editor. Tom94022 (talk) 19:37, 19 October 2012 (UTC)

Average Access Time

I do not know for sure what "Average access time" means, but I know for sure our coverage is unreliable. We currently have:

An HDD's Average Access Time is its average Seek time which technically is the time to do all possible seeks divided by the number of all possible seeks, but in practice is determined by statistical methods or simply approximated as the time of a seek over one-third of the number of tracks.

However, we also have on Hard disk drive performance characteristics:

The access time or response time of a rotating drive is a measure of the time it takes before the drive can actually transfer data. The factors that control this time on a rotating drive are mostly related to the mechanical nature of the rotating disks and moving heads.

And later in this article:

Access time can be improved by increasing rotational speed (thus reducing latency) and/or by reducing the time spent seeking.

It is clear that access time is not merely seek time; it includes latency. Therefore, either Average Access Time means something other than the average of access times - which is highly dubious - or our definition of Average Access Time is wrong. However, we do have a non-insignificant source giving this definition:

The average time a drive requires to perform a seek operation, usually measured by one-third stroke.

I propose that Western Digital got that definition wrong. Note that this is not from a page specifically covering average access time, but from a glossary. I easily found 2 articles contradicting WD:

Average Access Time: The average time it takes for a disk drive to begin reading data placed anywhere on the drive. This includes seek time and latency.[1]
An average access time of 8.0 ms (including seek time and rotational latency) [...] [2]

Unless we can find more sources supporting WD's definition, we should fix our definition. Note that the lead refers to average access time. --Chealer (talk) 23:38, 4 January 2013 (UTC)

The WD definition is historically wrong; average access time as published by most if not all disk drive manufacturers would better be called average seek time. Unfortunately access time as a term of computer art,as in RAM, means the time to the start of the data transfer which in disk drive terms is seek time plus time to rotate the disk to the data which on the average is average access (seek) time plus latency. That is, in the disk drive world average access time means something other than average of access times! Thus I think the article is correct as written. Since this is an HDD article we should use the industry definitions and maybe clarify the distinction over the computer industry term access time in a footnote. Tom94022 (talk) 00:05, 5 January 2013 (UTC)
A simple solution is to purge the term access time replacing it throughout with access to data. Tom94022 (talk) 00:10, 5 January 2013 (UTC)
I'm afraid it's not that simple. Access to data doesn't refer to a time. --Chealer (talk) 14:41, 9 March 2013 (UTC)
OK, how about replacing access time standing alone with time to access data? Tom94022 (talk) 19:24, 9 March 2013 (UTC)

────────────────────────────────────────────────────────────────────────────────────────────────────Actually I am not sure what u are disputing since the referenced WD dictionary page separately defines access time and average access times as two different things consistent with our article. However, since there is obvious confusion I think should either purge access time or footnote its usage with an explanation of the non-obvious different meanings. Tom94022 (talk) 00:52, 5 January 2013 (UTC)

I am not disputing a specific sentence, as I do not know what Average Access Time means for sure. If you are right, then this page is not incorrect, but if Average Access Time really means something other than average of access times, we shouldn't use the term Average Access Time (we already use "Average seek time" in some cases), or we must have a huge warning about the "confusion". --Chealer (talk) 14:41, 9 March 2013 (UTC)
The page is correct, Average Access Time (initial caps) is a term of the disk drive industry art which is not the average of all access times as the term access time is used in computer art. We should use the industry term of art in this article about the industry's product. I am going to purge standalone access time and see what happens. Tom94022 (talk) 19:24, 9 March 2013 (UTC)
We don't have to use industry's jargon to designate its products. Whichever term is best should be our choice. Anyway, avoiding "access time" is neither desirable nor feasible. Can an expert confirm the meaning of "average access time" so we can decide how to move forward? --Chealer (talk) 03:24, 18 March 2013 (UTC)
While I consider myself an expert in the art, any expert's opinion is not necessarily relevant; what is relevant is that there are many reliable sources that confirm that in the disk drive industry Average Access Time is NOT the average of all access times, but rather the average of all seek times. It is a published term of the art, not jargon, and belongs in the article because that is the term the user will encounter if he/she should read any specification published for any disk drive, hard or otherwise. FWIW it turned out to be relatively easy to purge access time so as to avoid confusion. It might be worth adding a footnote that Average Access Time is NOT the average of all access times, but that is somewhat covered in Section 5.1. so it might be TMI. Tom94022 (talk) 05:44, 18 March 2013 (UTC)
While there may be many reliable sources supporting the problematic meaning, this article refers to a single one. And we have the opinion of a single presumed expert. Which is why I'd like at least the opinion of another expert. "Access time" may have been "purged" in a sense, but it remains in other articles, and even in this article as part of "average access time". If the situation is what it seems to be, a footnote is insufficient, the terminology must be explained in the body. I do not see what you say in section 5.1. I simply see the definition which conflicts with the definition of "access time", but I consider the contradiction as the problem, not as the solution. --Chealer (talk) 16:06, 23 March 2013 (UTC)

2012 sales

Hard drive shipments: 475 million in 2012 (TG Daily, May 8, 2013). The article says it was 673 million units in 2012 and 624 million units in 2011. —  Ark25  (talk) 07:45, 9 May 2013 (UTC)

The article says forecast for 2012. Given that TG Daily, May 8, 2013 doesn't state its source it is not likely a reliable source. I suspect it comes from a misreading of iSuppli's PC HDD market size and not the total market size. Most reliable sources put the the actual total market size for 2012 at about 600M down from previous forecasts. Tom94022 (talk) 17:29, 9 May 2013 (UTC)

material stores the bits

What material stores the bits? This article vaguely mentions a "magnetic material" and more specifically says it is a "ferromagnetic material". I suppose in theory any magnetic material could in principle be used to store bits. But is there one specific material that is actually used in practice? Or do many hard drives use one material, and many other hard drives use some other material? --70.177.113.174 (talk) 18:09, 7 June 2013 (UTC)

Great question. Initially the magnetic layer was gamma iron oxide particles in an epoxy base. Beginning in the late 1980s, the industry gradually moved to various cobalt alloys (Co-Cr-Ta, Co-Pt-Cr, Co-Pt-Ni). I'll see if I can find a suitable source to add this to the article or maybe to the Magnetic Disk article. Tom94022 (talk) 22:26, 7 June 2013 (UTC)

The dates on this article are off by 200 years

It says that the hard drive was invented in 1754. The dates on the rest of the article seems to be 200 years off. That is a lot of editing to go through, so I thought I would report it on here. THanks, 86.45.150.169 (talk) 20:07, 17 June 2013 (UTC)Terri

There's a persistent vandal who's been vandalizing pages since 2013-06-12 by editing the dates. I just reverted his/her latest "contribution" in its entirety. Guy Harris (talk) 20:31, 17 June 2013 (UTC)

Silent Data Corruption

It is true that HDDs have a hard error rate but dealing with it is beyond the HDD and therefore beyond the scope of this article so I suggest the section Silent Data Corruption is not particularly relevant to this HDD article and should be incorporated in the RAID article or elsewhere. The hard error rate is already covered in the prior section Error rates and handling so I removed the section leaving the text there for someone to incorporate into RAID or elsewhere. All that is needed in this article is a See Also or link to where ever Silent Data Corruption winds up. Tom94022 (talk) 23:32, 1 January 2014 (UTC)

I think much of this material, especially that added today by an IP, reads like OR. It is not for Wikipedia to say that anything is "evident" (though we can quote a RS that says so), nor to declare that one thing or another is "one solution to his problem". Extensive references are needed if this paragraph is to be retained without severe editing, whichever article it ends up in. Jeh (talk) 23:42, 1 January 2014 (UTC)
On the other hand, I feel that information from the first paragraph of this former section does belong here, within the existing "Error rates and handling" section. This would put the error rates into context. Jeh (talk) 00:30, 2 January 2014 (UTC)
I agree that some of the first paragraph can go into the existing "Error rates and handling" section, but perhaps nothing more than a hard error (uncorrectable and unrecoverable) can be dealt with by redundancy at the subsystem level (i.e. RAID) or at the system level but even then "Silent Data Corruption" may occur leading to loss of data. BTW, I am not even sure he had a good definition of "Silent Data Corruption". Why don't u take a try at adding RS material to the existing paragraph, his ACM article seems like a good place to start, but again, IMO most of it belongs in the RAID article not here. Tom94022 (talk) 01:18, 2 January 2014 (UTC)
A quick look at the referenced CERN presentation suggests that in their context "Silent Data Corruption" is end-to-end, including far more than disk and thus supporting my contention that this is not appropriate for an HDD article, maybe beyond a RAID article. Tom94022 (talk) 01:32, 2 January 2014 (UTC)

──────────────────────────────────────────────────────────────────────────────────────────────────── Hello there! It was me initially moving these "Silent data corruption" and "Error rates and handling" sections over here, so please allow me to present an overview of the background.

In a few words, that move was a rushed "rescue" of the content from the ZFS article, back at the time when that article went under a heavy "surgery". This article looked like a somewhat acceptable fit, at least for some time, while I do agree that "Silent data corruption" is far beyond what HDDs are supposed to be dealing with. On the other hand, "Error rates and handling" section totally belongs to this article.

I'm going to move the currently commented-out content to Data corruption article, and that should be the final destination for "Silent data corruption" section, if you agree. — Dsimic (talk) 03:24, 2 January 2014 (UTC)

Done, please check it out: edit #1, edit #2, edit #3, and edit #4. — Dsimic (talk) 05:19, 2 January 2014 (UTC)
Thanks for cleaning up the article; I'm not sure a "See also" is exactly the right link but I need to work on the "Error rates and handling" section at which time I will see if I can find a better way to link.

────────────────────────────────────────────────────────────────────────────────────────────────────Moved remainder of discussion about definition of Silent Data Corruption to Talk:Data corruption Tom94022 (talk) 18:29, 2 January 2014 (UTC)

External portable drives? What about "external desktop drives"?

The section on "External removable drives" was renamed to "External portable drives", but I'm not sure how portable all the drives that attach via hot-pluggable connections are. Seagate's page for external drives has separate sections for "Portable Hard Drives"] and "Desktop Hard Drives", as does Western Digital's page for external drives, and LaCie's site has "MOBILE STORAGE" and "DESKTOP STORAGE". Guy Harris (talk) 21:15, 12 January 2014 (UTC)

Hello there! I agree that both old and this new section title aren't perfect... Though, naming it "External desktop drives" would be slightly worse, as that way we'd be ruling out external laptop drives while they're covered in the section; that's probably that manufacturers are trying to accomplish by separating external HDDs into those two categories (3.5-inch vs. 2.5-inch external HDDs). Personally, calling something "external desktop drive", and something else "external laptop drive" (or "external portable drive") is pretty much plain PR talk to me, as that would mean external 3.5-inch HDDs aren't portable; sure, they are less easily portable, but they still are very portable. — Dsimic (talk) 21:25, 12 January 2014 (UTC)
I split the difference and renamed it to "External desktop and portable drives".
I think "portable", in manufacturer-speak, means "you can carry this drive fairly conveniently" (e.g., "tucks into a laptop bag", as is the case for the current backup drives for our household), and "desktop" means "you could carry this drive, but it's a bit of a pain (so that it won't tuck into a laptop bag without pain, as was the case for our older backup drives). Our "desktop" backup drives also included power supplies and wall-warts; our "portable" backup drives get their power from the USB connection. So I think the distinction isn't just PR talk; there really are differences in how the drives are used (at least in our experience - we take the "portable" backup drives along with us and do backups when away from home, whereas we didn't bother doing so with the "desktop" backup drives.
I don't know how the manufacturers intend "desktop" to be interpreted, but I view it as referring to the drive, not to the machine into which the drive was plugged; you can plug "desktop" external drives into a laptop (our computers are both laptops, so that's what we were doing with the old "desktop" backup drives), but they're more likely to be sitting on a desktop than carried with you. (And you can plug "portable" backup drives into desktop computers, as my sister and brother-in-law did.) Guy Harris (talk) 21:43, 12 January 2014 (UTC)
Well, back at the time (not so long ago) I was extremely happy to have an external USB 2.0/Firewire enclosure holding a 3.5-inch 80 GB HDD, and I considered it to be super portable. :) I agree that portable vs. desktop distinction makes sense, but that's still primarily the way for manufacturers to name their product lines and make choosing them easier for Joe Averages. To me, it would be more important to know whether is it a 5,400 or 7,200 rpm HDD inside the enclosure, and how much cache it has (what isn't published, as Joe Averages have no clues about those numbers), than to have those two obvious form factor–based product categories. :)
What about enclosures holding two HDDs, exposing them via eSATA or USB (no Ethernet, so they're not NASes), while doing some kind of RAID or JBOD? Should those be categorized as "external enterprise drives"? They're still portable, fit on a desktop, and can be connected to laptop and desktop computers. :)
I've edited this section further, please check it out. It should be acceptable, and the section title is now more universal. — Dsimic (talk) 21:57, 12 January 2014 (UTC)
Your edit looks great, this additional clarification fits really well. I've just cleaned up the containing paragraph, making it slightly more clear and readable; please check it out. — Dsimic (talk) 22:12, 12 January 2014 (UTC)
Your edits look good.
As for the two-HDDs-in-a-box drives, I might be inclined to call them "desktop" drives. I think of "enterprise" external drives as being Fibre Channel or Serial Attached SCSI drives in an external enclosure (or, in the olden days, parallel SCSI drives in an external enclosure; we had parallel SCSI and FC, but not SAS, enclosure drives when I left NetApp), i.e. the sort of drives that get attached to servers. (Perhaps they'd be "enterprise" when attached to workstations.) Guy Harris (talk) 22:24, 12 January 2014 (UTC)
Thanks. Agreed, calling those boxed-two-drives enclosures "desktop" as well should be good. For a moment, I forgot those SCSI enclosures that were highly popular back in the days of Sun workstations, for example; there was quite a variety of those, with some bordering with external backplanes, or however we're going to recall them. :) Nowadays, there are rack-mounted SAS-based storage arrays intended to be attached to SAS controllers installed in servers, for example. Though, most of them are (again) more of a kind of external backplanes, as majority of the control logic is within the servers.
Shall we improve Backplane § Backplanes in storage section as well, so those external backplanes are also mentioned? — Dsimic (talk) 22:40, 12 January 2014 (UTC)

spare sectors

How many spare sectors does a typical hard drive have? I hope that re-posting this question here will get a better response than the (lack of) response at Talk:Bad sector#Spare sectors. --DavidCary (talk) 19:19, 27 January 2014 (UTC)

I believe the remaining three HDD manufacturers no longer publish this information, treating the number of spare sectors, their locations and allocation algorithms as trade secrets. It used to be published; e.g. a IBM 4 GB 2½-inch HDD circa 1997 had specified a minimum of 1230 spare sectors. Alternatively, the number could be derived from knowledge of the drive's geometry (zones, tracks per zone, sectors per track) but that is also no longer published. BTW, there is an interesting paper out of Korea on a software program that can be used to determine the geometry of an HDD; I don't think anyone ever used it to determine the number of spare sectors. Perhaps it also could be estimated from changes to SMART data. So the lack of response is likely a lack of public information. Tom94022 (talk) 22:56, 27 January 2014 (UTC)

2.5" HDD Z-height max is really 19mm

I noticed the form factor table limits the Z-height of 2.5" HDDs to 15mm with a source reference from a 2010 online article. I personally poses a Quantum 2.5" 80MB Go Drive from the early 1990s that is 19mm high. I have been unable to find an online copy of the data sheet, but I have found a few "sources" listing the 18.8+-2mm height from a copy of the documentation. http://www.4drives.com/DRIVESPECS/QUANTUM/3364.txt http://museum.ttrk.ee/th99/h/txt/3368.txt

I have also found a few photos of it online that you can easily measure the height, compare it to the length, and scale that to the known length of the 2.5" form factor. http://i.pchub.com/i/Quantum-Go-Drive-120AT-HDD-IDE-2-5-1GB-Below-Go-Drive-120AT-GO12A011-b-24302.jpg

I am not sure if these sources are legitimate enough or if maybe the combination of them is sufficient. I don't recall any other drive names to look up to search for spec sheets. § Music Sorter § (talk) 03:43, 5 March 2014 (UTC)

Hello there! In my opinion, the combination of these three sources (two text files and one picture) should be more than enough for adding 19 mm as one more possible height for the 2.5-inch HDDs. — Dsimic (talk | contribs) 02:08, 6 March 2014 (UTC)