Jump to content


Photo

IBM Deskstar 120GXP


  • You cannot start a new topic
  • Please log in to reply
90 replies to this topic

#26 knut

knut

    Member

  • Member
  • 23 posts

Posted 24 February 2002 - 01:30 PM

http://forums.storag...=461&highlight=

Seems like the smaller wd drives uses 40gb pr. platter.

#27 Shark

Shark

    Member

  • Member
  • 166 posts

Posted 24 February 2002 - 05:19 PM

Seems like the smaller wd drives uses 40gb pr. platter.


Hard to argue against people having the drive and benchmarks to back it up, but I am surprised. I can't find any mention from WD regarding this drive configuration.

It would make sense, though, to offer more than just the 120GB capacity. It could be that making the 40GB platters costs them enough that, until now, they haven't wanted to sell them in anything but the most expensive drives.

Here's the real dilemma: now how do we know which drive we are buying??

#28 Mickey

Mickey

    Member

  • Patron
  • 2,274 posts

Posted 24 February 2002 - 11:00 PM


Seems like the smaller wd drives uses 40gb pr. platter.


Hard to argue against people having the drive and benchmarks to back it up, but I am surprised. I can't find any mention from WD regarding this drive configuration.

It would make sense, though, to offer more than just the 120GB capacity. It could be that making the 40GB platters costs them enough that, until now, they haven't wanted to sell them in anything but the most expensive drives.

Here's the real dilemma: now how do we know which drive we are buying??


Serial number prefix and the suffix on the model number. Or, if you have a way of scanning the drive to see how many physical heads it has, that'd be another way to do it.

#29 Shark

Shark

    Member

  • Member
  • 166 posts

Posted 24 February 2002 - 11:36 PM

Serial number prefix and the suffix on the model number. Or, if you have a way of scanning the drive to see how many physical heads it has, that'd be another way to do it.


Um... Yeah.... I think that's what Knut meant when he pointed out the different revision numbers.

I'd just like to take this moment to be a complete jerk and point out that most of the time, online retailers will not read you serial numbers off of hard drives. Also, most brick and mortar stores will not let open up a box, "scan the drive to see how many physical heads it has" and then tell them, "Sorry, this isn't the one I wanted."

CLARIFICATION: How would people know BEFOREHAND what drive a reseller had when purchasing online (outside of the reseller reporting the serial numbers, which is not likely from most resellers)?

FURTHER CLARIFICATION: That is mostly a rhetorical question to show the potential problems as well as my displeasure with WD using the same model number for (in my view) substancially different HDDs.

#30 entity80

entity80

    Member

  • Member
  • 1 posts

Posted 25 February 2002 - 06:54 AM

I was looking at the 40gb 120gxp and noticed that it was the exact same size as the 40gb 60gxp..I was thinking they were the same drives until I read that 120gxp uses 1 40gb platter instead of 2 20gb platters. So what did they do with the 60gb 120gxp? 1 40gb platter and 1 20gb? If so, how does this affect the performance..will the 40gb or 60gb perform on par with the larger 120gxp's?

#31 Shark

Shark

    Member

  • Member
  • 166 posts

Posted 25 February 2002 - 07:47 AM

I was looking at the 40gb 120gxp and noticed that it was the exact same size as the 40gb 60gxp..I was thinking they were the same drives until I read that 120gxp uses 1 40gb platter instead of 2 20gb platters. So what did they do with the 60gb 120gxp? 1 40gb platter and 1 20gb? If so, how does this affect the performance..will the 40gb or 60gb perform on par with the larger 120gxp's?


I think that they use two 40 GB platters, but only "finish" one side of the second platter, and they do not put a read/write head on the "unfinished" side.

"Normally", each platter has a read/write head on both sides.

It should not affect performance a great deal. It can vary somewhat, depending on the drive's controller code, but in general a drive with more platters (of the same size) will perform better than his "little brothers". I don't know of any manufacturer's that are using drives with different sized platters in the same drive.

I think most drives now record a few of the outer tracks on the first side, then move to the second side, record a few more tracks, and so on. Then they go back to the first track and begin all over. In this way, they record data first on the outside of the platters, where the sustained read/write speeds are faster. The alternative is to write an entire side, then move to the next side.

With either method, though, you simply have more "outside edge" to work with, which should yield a slight performance increase.

Ok, so I'm bored and rambling.

#32 JJR512

JJR512

    Member

  • Member
  • 6 posts

Posted 25 February 2002 - 03:33 PM

Data is recorded cylindrically, that is, in cylinders. A platter is formatted with a series of concentric circles, or tracks. A cylinder is all the tracks from the same position, for example the 100th track from the center, on each platter. Data is recorded starting with the top of platter 1, then the bottom of platter 1, then the top of platter 2, then the bottom, and so on and so forth. (Or maybe it starts at the bottom, but that's not important.) The idea is to write and read as much as possible with as little head movement as possible.

#33 JJR512

JJR512

    Member

  • Member
  • 6 posts

Posted 25 February 2002 - 03:38 PM

Since there seems to be no way to edit my message (that I can see), let me add on that yes, recording starts at the outside of the drive first, and as a drive fills up, recording moves inward. Programs like Norton Speed Disk can move infrequently used files to the inside of the platters, though, thus freeing up more outside room for the more frequently used files, the theory being that the more frequently you use a file, the faster it should be for you.

As far as I am aware, data is never recorded by first filling up an entire side of a platter before moving to the next side, filling that entire 2nd side up before moving to the next platter, etc. I believe it is always done in cylinders.

#34 Steel

Steel

Posted 25 February 2002 - 05:14 PM

As far as I am aware, data is never recorded by first filling up an entire side of a platter before moving to the next side, filling that entire 2nd side up before moving to the next platter, etc. I believe it is always done in cylinders.

If I remember correctly, one of Western Digital's SCSI models worked this way, not sure why they did it though.

#35 Shark

Shark

    Member

  • Member
  • 166 posts

Posted 25 February 2002 - 08:36 PM

As far as I am aware, data is never recorded by first filling up an entire side of a platter before moving to the next side, filling that entire 2nd side up before moving to the next platter, etc. I believe it is always done in cylinders.

If I remember correctly, one of Western Digital's SCSI models worked this way, not sure why they did it though.


As far as I am aware, data is never recorded by first filling up an entire side of a platter before moving to the next side, filling that entire 2nd side up before moving to the next platter, etc. I believe it is always done in cylinders.

If I remember correctly, one of Western Digital's SCSI models worked this way, not sure why they did it though.



Well, I thought some drives had done this in the past (long long ago), but no drives still do this, do they? Truth is, I was up all night and wrote that this morning just sort of rambling off the top of my head.

Thinking out loud (again): if you were going to fill a drive with data that will only be read sequentially (like a 4 GB drive filled with a single DVD movie), would this method (fill one side at a time) be the fastest?

Probably not, since it should be faster to switch heads rather than switch tracks.

I searched pretty hard through the SR reference section, and couldn't find any mention (though I could have simply missed it) of these different methods.

My guess is that doing cylinder by cylinder would yield the best overall performance for a different application types.

#36 Mickey

Mickey

    Member

  • Patron
  • 2,274 posts

Posted 26 February 2002 - 12:12 AM

I'd just like to take this moment to be a complete jerk and point out that most of the time, online retailers will not read you serial numbers off of hard drives.  Also, most brick and mortar stores will not let open up a box, "scan the drive to see how many physical heads it has" and then tell them, "Sorry, this isn't the one I wanted."

CLARIFICATION:  How would people know BEFOREHAND what drive a reseller had when purchasing online (outside of the reseller reporting the serial numbers, which is not likely from most resellers)?

FURTHER CLARIFICATION:  That is mostly a rhetorical question to show the potential problems as well as my displeasure with WD using the same model number for (in my view) substancially different HDDs.


Fair enough. There is no easy way to tell from the outside of a box what kind of drive is inside. I suspect WD switched to the current naming convention because it simplified things for certain large customers, although it would confuse things for the more discriminating endusers (like those that frequent SR). That, and it gives WD more flexibility in selling, as they can sell whatever configuration they want as long as the total capacity is correct.

Still doesn't make me happy about it, though, so you're not alone. :(

#37 Shark

Shark

    Member

  • Member
  • 166 posts

Posted 26 February 2002 - 06:10 AM

Still doesn't make me happy about it, though, so you're not alone. :(


I reread my post that you quoted, and I just wanted to apologize for the rude tone.

#38 sprockkets

sprockkets

    Member

  • Member
  • 27 posts

Posted 28 February 2002 - 01:26 AM

From reading the guide here sometime ago IBM makes 60GB drives cause one side of the platter failed, so instead of tossing a half good platter away, simply get a 60 instead of 80GB drive. The 60GB 120GXP should be faster, quiter, cooler, more reliable since it has half the amount of heads and one less platter.

Hey maybe IBM learned it's lesson about reliability, that's why the 60GXP has only 3 platters, and it took it's time to release the 120GXP. I had my 75GXP fail twice, but I've used 6 60GXPs (one was a replacement for the 75GXP from IBM under warranty service, nice guys) and no problems whatsoever. Heck, I'm bidding on ebay on a returned no warranty 60GXP hard drive, simply cause IBM doesn't care who bought the drive, they are easy to send it and repair.

#39 DenisS

DenisS

    Member

  • Member
  • 2 posts

Posted 02 March 2002 - 06:24 PM

I have read some specs for both IBM 60GXP and 120 GXP drives.
Here we go:
Projected term of life:
5 years when operating in average desktop
i.e. average 333 POH a month (power-on hours).
It means roughly 10 hours a day. :?
Yes, it does not look for the servers. It means 5/2.5 = 2 years of projected life for the servers.
Now, total hours are 333 a month * 12 months * 5 years=roughly 20,000 hours.

From the Western Digital site, for WD 600BB, the component design life is 5 years too. However, the MTBF is 500,000 POH.
I don't have a clue what MTBF is, but 500,000 POH is 25 times more than 20,000.
For the WD1200JB they do not tell POH at all, just mention 5 years.

What do you think about all these numbers?

#40 JJR512

JJR512

    Member

  • Member
  • 6 posts

Posted 02 March 2002 - 07:06 PM

MTBF stands for Mean Time Before Failure. "Mean" is a mathmatical term that basically means "average". MTBF is the average amount of time something will last before it fails. An MTBF of 500,000 hours means most should last to about 500,000 hours, but some will last more and some less.

#41 Shark

Shark

    Member

  • Member
  • 166 posts

Posted 04 March 2002 - 03:26 PM

I have read some specs for both IBM 60GXP and 120 GXP drives.
Here we go:
Projected term of life: 
5 years when operating in average desktop
i.e. average 333 POH a month (power-on hours).
It means roughly 10 hours a day. :? 
Yes, it does not look for the servers. It means 5/2.5 = 2 years of projected life for the servers. 
Now, total hours are 333 a month * 12 months * 5 years=roughly 20,000 hours.

From the Western Digital site, for WD 600BB, the component design life is 5 years too. However, the MTBF is 500,000 POH.
I don't have a clue what MTBF is, but 500,000 POH is 25 times more than 20,000.
For the WD1200JB they do not tell POH at all, just mention 5 years.

What do you think about all these numbers?



I just want to make a distinction between recommended use and MTBF--that is not an apples to apples comparison.

An analogy: I use synthetic oil in my car--some brands say I can drive 50,000 miles before needing an oil change. However, my mechanic might recommend that I continue to change the oil every 3,000 miles--over 16 times more often.

A recommendation is not anything like MTBF.

Frankly, MTBF figures on new products are based on (allegedly) educated guessing, not actual history. Hopefully, the engineers used appropriate or even conservative modeling to derive their numbers, but I'm the temptation is surely to inflate them. It would certainly seem so in the case of Western Digital; do the math: 500,000 hours is just over 57 years. Obviously, they have no proof--their drives haven't been around long enough to be powered on for 57 years.

I like WD's drives, but I think it's pretty bold for them to estimate that only half their drives would fail after 57 years of use--and that's not even a SCSI drive! hehehe

At least IBM was honest enough to write this:

Reliability has been measured in the industry with Mean Time Between Failure (MTBF), a term or claim that is easy to advertise (a higher number is better), difficult to explain, and nearly impossible to prove or guarantee.

(funny, given the 75GXP series issues, but IBM wrote an article about hdd reliability, from which the above quote was taken: http://www.storage.i...ary/reliab.htm)

#42 rcfa

rcfa

    Member

  • Member
  • 2 posts

Posted 05 March 2002 - 07:38 PM

How can a drive be labelled a good server drive when the manufacturer recommendation is 333 power-on hours a month (~11-12 poh/day). In my experience, servers are pretty much required to be up 24/7.  :roll:


Well, I discovered that early on, and my suspicion is that it might have something to do with the fluid bearing, viscosity breakdown, or something like that...
Anyway, I asked IBM, and here the result of that exchange:
(I cut my e-mail address and the name of the IBM rep, because
I don't want to get him fired if he disclosed more than he was supposed to...)

Me, myself and I: I have an IDE-to-SCSI RAID which
takes 6 EIDE HDs, creates a RAID-5
array from them, and presents the
resulting storage to the host
computer as if it were a single, large
SCSI drive.
Currently this array is loaded with 6
IBM 34GXP 34 GB Deskstar drives,
however I want to increase the
capacity, and was thinking of
replacing these drives with 6 of the
120GXP Deskstar drives.
The recent troubles with the 60GXP
series drives made me wonder
about potential reliability problems.

Then I read the technical specs of
the 120 GXP series, and there I find
the following item, notably under the
title "Reliability":
"Recommended power-on hours
(monthly) 333"

Does this mean this drive is not a
reliable choice to be used in a RAID
that's running 24*7 ?
I cannot find similar exclusions on
either Maxtors or WesternDigital's
drive specs, and the later specifies a
non-recoverable error rate that's an
order of magnitude better, and
claims a minimum of 50k start/stop
cycles, while the IBM drive claims an
average of 40k start/stop cycles.

Are IBM's specs just particularly
conservative in their estimates or is
the side effect of the drive being
more silent and marginally faster
than the WD1200JB, that the
reliability suffers and that it loses the
ability to be used in a 24*7 set up?

What about all the people who'd like
to use a big drive like this in various
slimline rackmount internet servers?
Will they have to go elsewhere?

IBM:
No IBM ATA drive has been rated for use
in a 24/7 environment. In most cases, only
SCSI drives are rated for this type of usage.
ATA RAID is a relatively new market. It may
work, but it is not tested by IBM in this
capacity.

Me, myself and I:
thanks a lot for your answer. I would strongly suggest that IBM
rethinks their strategy, however. While ATA RAIDs are relatively
new (at least in terms of common visibility, they have been on
the market for at least four years now), rackmount LINUX servers
are not. There must be thousands, if not millions of 1U or2U
PC based rackmount Linux servers in server farms that are rented
out to people who want to host their own web site or e-store
at a managed site. This was even acknowledged in the product
description of your new 60GB TravelStar drive, which is now rated
for "continuous use" in "blade servers".
Similarly many offices that just need a workgroup e-mail and
intranet server for low data volume applications commonly use
a desktop PC, often even one that's no longer a current model,
because serving a few intranet web pages is a low CPU power
job, so even a relatively outdated machine can take care of this.
Hardly anyone uses SCSI drives in such circumstances, particularly
since Microsoft software makes it a pain to deal with SCSI, while
ATA is comparatively speaking a breeze.

IBM:
This is not something that we can change.
This is a barrier that is part of the IDE infrastructure There is no way
for us, nor any other manufacturer, to overcome this.
This is now, and probably will be for quite some time, the primary
difference between IDE and SCSI.
The Travelstar 60GB seems to be an exception to the rule. How this was
accomplished I do not know.

Me, Myself and I:
with all due respect, but I don't understand your answer.
I don't see how an electrical interface specification and
data transfer protocol (EIDE/ATA) has any influence on the
reliability of the drive, which is largely a matter of
mechanics, read/write heads, electronics, lubrication, etc.

The fact that one drive uses and EIDE/ATA interface doesn't
make it necessary that it can only endure 333h power-on per
month, while a SCSI drive can work in a 24*7 environment.

That SCSI drives may be faster due to the built-in smarts,
etc. is a completely different story.

What seems to be much more believable is that in order to be
able to compete on price, which is key in the desktop market,
not the same amount of quality engineering goes into the
mechanics and electronics of the EIDE/ATA drives that goes into
the more expensive SCSI drives.

The fact that the TravelStar 60GB drive can withstand
a 24*7 operational environment shows only that on a drive like
this the margins are fat enough to put the relevant quality
into the mechanism.

Several other manufacturers advertise their higher-capacity
EIDE/ATA drives for use in work group servers (which run 24*7)
or for the use in rack-mount internet appliances (which also run
24*7). So either IBM is more cautious, or their quality is better.
Needless to say, without access to data proprietary to each of
these companies, there is no way of telling which is the case.

Now should I have missed a point as to *why* EIDE/ATA drives
cannot be built as reliable as SCSI drives, I'd sure like to know.

IBM:
*Silence*

Me, Myself and I:
I think I'll have to go with a WD drive, because I can't gamble
with my server's reliability, and so access speed will have to take
second priority. After all, the WD1200JB is no slouch either.

Ronald

#43 rcfa

rcfa

    Member

  • Member
  • 2 posts

Posted 05 March 2002 - 07:47 PM

My guess is that doing cylinder by cylinder would yield the best overall performance for a different application types.


Surely is much faster, if the electronics can keep up. Theoretically you could even do parallel reads off the various platters, and then achieve similar performance gains as in RAID arrays in comparison to reading
only one platter's data at a time. A striping RAID reads multiple
drives at once, so it's the same thing, just one step up the abstraction
hierarchy.
Data rates depend on bit density, rotational speed. If you then read N
platters concurrently, with other parameters the same and no interface
bottlenecks, you Ntuple the data rate.

#44 Future Shock

Future Shock

    Member

  • Member
  • 2,885 posts

Posted 06 March 2002 - 09:38 PM

Now should I have missed a point as to *why* EIDE/ATA drives
cannot be built as reliable as SCSI drives, I'd sure like to know.

Ronald

From what others have posted here WRT IDE using the NEWEST technologies to achieve higher (and thus cheaper) areal densities on each platter, I would have to guess that it is simply higher density that causes IDE to have the lower MTBF ratings. After all, with SCSI's lower densities comes the ability to have a higher flying head, which makes the drive less suceptible to particle contamination, shock, heat, etc. Is that all there is to it? Don't know, but I put it forward for group consideration...

#45 drees

drees

    Member

  • Member
  • 84 posts

Posted 08 March 2002 - 10:56 AM

Based upon IBM's recent history, I no longer recommend IBM drives for my company.

For high performance IDE RAID setups, we're now using the WD WD1200JB drives with their 8MB cache.

For budget IDE RAID setups and desktop machines (NONE of our desktops remain on only 8-10 hours/day, they are all 24/7 machines) we're now using either Seagate Barracuda IV or Maxtor drives.

After seeing numerous IBM drive failures over the last year while seeing 0 drive failures from other manufacturers, the decision was pretty much made for us!

#46 sprockkets

sprockkets

    Member

  • Member
  • 27 posts

Posted 08 March 2002 - 03:47 PM

Drees is right; although I've never had problems with the 60GXP, and mine runs 24/7, you can't say let's use the 120GXP with this limitation to your IT manager.

Let's put it this way, IBM doesn't guarrantee or stand behind 24/7 use of it's drives, for the same reason AMD says you shouldn't use XP instead of MP processors in a dual configuration. Of course there is nothing wrong with using XPs in a dual config, but you don't have that guarrantee, and for businesses won't cut it.

#47 sprockkets

sprockkets

    Member

  • Member
  • 27 posts

Posted 08 March 2002 - 07:25 PM

Just a reminder, isn't WD been part of IBM for some time? Just like to confirm what I thought I read here or somewhere long ago.

#48 Zwoe

Zwoe

    Member

  • Member
  • 3 posts

Posted 08 March 2002 - 08:38 PM

Hope your wonderful knowledge can give me an answer for this:
I ordered the IBM deskstar 120 G hdd and going to instal it on my P4-1.5 G cmputer to do video editing stuff. You know video loves big and fast hdd, if I instal it as the second one (slave setup), will it perform less than it is alone by itself? I'm currently have a 40 G IBM deskstar.

Thx in advance for any advice!
Zwoe

#49 jehh

jehh

Posted 08 March 2002 - 09:20 PM

Just a reminder, isn't WD been part of IBM for some time? Just like to confirm what I thought I read here or somewhere long ago.


No, WD was never part of IBM.

WD licensed some IBM technology about 3 years ago, but they haven't worked together for a few years.

IBM is now behind in technology from the other drive companies for the most part, they have been coming to the party last for a year or two.

Jason
There are two theories to arguing with women. Neither works.

#50 sprockkets

sprockkets

    Member

  • Member
  • 27 posts

Posted 09 March 2002 - 01:12 AM

Re: It will perform just fine if it has it's own cable. Then again, if you have 2k or XP pro and have one hard drive on each cable, stripe a portion for RAID performance.

I'm not too familiar with video editing, but you probably don't need a stripe config if follow the hd per cable, one drive is used for access to the program and virtual memory, and then one hd is used for the temporary video file.

Just a little note, I took apart a defective 75GXP I got cheap from ebay. Maybe I'll take a pic soon. Apparantly the drive heads park and stay there by a magnet holding it in the park position.



0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users