Review of Seagate's 750Gb Hard Drive

Zoxed writes "The Tech Report have a comprehensive review of Seagate's Barracuda-7200.10 'perpendicular' drive, including a primer on the technology. They ran performance tests against 10 other drives, checking the noise and power consumption levels. The Seagate fared pretty well, even on cost (per Gigabyte)." From the article: "Perpendicular recording does wonders for storage capacity, and thanks to denser platters, it can also improve drive performance. Couple those benefits with support for 300 MB/s Serial ATA transfer rates, Native Command Queuing, and up to 16 MB of cache, and the Barracuda 7200.10 starts to look pretty appealing. Throw in an industry-leading five year warranty and a cost per gigabyte that's competitive with 500 GB drives, and you may quickly find yourself scrambling to justify a need for 750 GB of storage capacity."

Comments

Get perpendicular :D

[LiquidCoooled]

http://www.hitachigst.com/hdd/research/recording_h ead/pr/PerpendicularAnimation.html

Watch out for the superparamagnetic effect though.

Re:Big HUGE warnings

[Dan Ost]

No, this isn't true. If the failure rate of drives is constant (pretty close to reality), then
if you've got 7 drives and I've got 1, you're seven times more likely to lose a drive than
I am.

Granted, you only lose 1/7th if your drive fails, and I lose all of it, but since we're both
making backups (you ARE making backups, right?), you're paying 7 times the space, electricity,
heat, and noise costs for less reliable storage than I am. Assuming that we both run out systems
long enough for drives to fail, you're also paying 7 times as much of your time replacing drives
than I am.

What sense does that make?

Why do airplanes only have 2 engines?

[Marx_Mrvelous]

Because with only 2, there is *less* risk of engine failure.

Having 7 drives increases your risk of failure by a factor of seven. Unless you mirror every drive, but then, you now have 14 disks v 2...

Re:Whoah

[Martin Blank]

You just have to look for them a bit. I just picked up a 300GB Maxtor SATA-2 with 16MB cache and NCQ that has a 5-year warranty, and it only cost me about $6 more than the 3-year warranty version with identical specs. Other companies may also offer them. (Of course, Maxtor is now a part of Seagate.)

Bad math..

[JMZero]

If you've got 7 drives and I've got 1, you're seven times more likely to lose a drive than
I am.

Let's say each drive has a 20% chance of failing. So if you have seven of them, do you have a 140% chance of one failing? Of course not. What you really have is 80%^7 percent chance of them all remaining OK. 80%^7 = 21%. Thus you have around a 79% chance of failure with 7 drives (if they all have 20% failure rate).

Your point still stands - but I noticed pretty much all of the replies to this guy used the same bad math.

Re:Bad math..

[enrevanche]

Not true, if you flip a coin 3 times are you 3 times as likely to get a tail. i.e. do you have a 150% chance of getting at least one tail? no you have 1 - 0.5^3 = 1 - 0.125 = 87.5% chance. There is 12.5% chance that you would get no tails. With drive failure it works the same way, you have a chance at no failures and also a chance of multiple failures.

7 times as many failures (over a large number of samples) is not the same as 7 times the chance to have a failure.

Re:Bad math..

[pla]

Thus you have around a 79% chance of failure with 7 drives (if they all have 20% failure rate).

IF you have a 20% failure rate.

It cheats somewhat to use that as an example, however, because with the real probabilities involved, you approach a linear trend with the number of drives.

Let's try an MBTF of 50k hours. That gives us a 0.002% chance of failure per hour. Take 0.99998 to the seventh, and we get 0.9998600084... Or "seven times as likely", accurate to better than one part in a thousand.


Though, I will admit doubt that the GP explicitly took that into consideration in his statement. ;-)

Re:Big HUGE warnings

[mrchaotica]

Plus who has room (and power) in their case for 7 drives? Datacenters, sure, but not home users.

I've seen cheap ($30) mid-tower cases that had about 8 internal 3.5" drive bays -- they just had the mounting rails go the entire height of the case. Combine that with a decent power supply and you're set.

I just wish they made high-quality cases with that many drive bays, but I haven't found any for some reason.

Re:That's a lot of DVDs

[Anonymous Coward]

For home/home-office use there is really only one practical way to back up huge hard drives - and that's onto other hard drives. However, for good security, you really need the backup drive(s) to be in another PC - and preferably in another building - and the network performance ends up killing you there.

So right now - I use sledded drives that I can remove from the PC and put somewhere safe.

Re:That's a lot of DVDs

[Anonymous Coward]

Assuming there is no loss in formatting (which we know isn't true) it would take 160 DVDs (4.7 GB). If you have dual layer disks, it would "only" take 80 disks to back up 750 GB. Because of marketing, this drive is probably closer to 700 GB in size, which would take about 150 (4.7 GB) DVDs. Time to go buy a couple spindles of DVDs, eh? Or...by two of these, and stick the second one in another computer somewhere and pray you have a nice upload speed.

Re:Scrambling?

[drinkypoo]

Any firewire bridge that has the right interface to speak to the drive should be able to talk to it just fine. This isn't the old dark days of DOS where you needed extender software just to talk to fancy new drives. Since drives use logical geometry to talk to the host adapter, this just isn't an issue any more.

Re:Scrambling?

[Fweeky]

Anything supporting LBA48 should handle it just fine, although we're rapidly approaching the 2TB limit many controllers have on a single disk/array. LBA48 supports drives up to 128PB (512 byte blocks * 2^48), but of course we're still in a largely 32bit world, so it's more like 512*2^32 unless you're careful.

Re:Big HUGE warnings - Not quite true

[swillden]

RAIDing 7 or 8 drives would be quite a task, while doing 2-4 drives is relatively easy.

For those who don't see the difference: Most boxes don't have controller capacity for more than four drives (two PATA channels and two SATA channels) and seven or eight drives will also strain your PSU and your cooling capacity. Might be hard to fit in your case, too.

Of course, once you solve those problems, actually setting up the RAID is no different whether you have three drives or 30. A little more typing, maybe.

My file server has six disks in it, BTW, so I've worked through all of this. I can easily add a seventh without any trouble. An eighth would require a new controller card. I'm not sure how many drives I can add before my 550W PSU starts to have trouble. My cooling solution is low-tech, loud and very effective: The side of the case is off and I have a 30-inch box fan (the kind you mount in a window to cool your house) blowing into it.

One nifty trick I discovered is that if you slice all of your disks up into many small partitions, then create many RAID-5 arrays (using partition 1 on each disk to create the first array, etc.), then use LVM to bind all the arrays together you can add additional disks and rebuild the arrays without having to find some way to back up all of the data first.

I just added a 500GB drive to my system and I'm in the process of changing all of my four-disk RAID-5 arrays to five-disk RAID-5 arrays. The process works like this:

1. Use pvmove to migrate all of the data off of an existing four-disk array.
2. Use vgreduce to remove the now-unused array from the volume group.
3. Use pvremove to remove the LVM superblock from the array.
4. Use mdadm to stop the array and clear the md superblocks on the partitions.
5. Use mdadm to construct a new five-disk array from the four partitions that made up the old array, plus a fifth partition from the new disk.
6. Use pvcreate to add an LVM superblock to the array.
7. Use vgextend to add the array into the volume group.
8. Go back to step 1 with the next four-disk array, until they've all been converted.

This assumes Linux, obviously, is a bit tedious and requires that your LVM volume have enough free space so you can drop an array out of it. It's a whole lot easier than trying to figure out how to back up a TB+ of data so that you can rebuild your array, though. In my case, there's an additional step right after step four -- because my new drive is SATA and Linux doesn't support more than 15 partitions on an SATA drive, I'm moving from using 20GB partitions to 40GB partitions. So after I kill each pair of four-disk arrays, I repartition the drive to merge the partitions.

Let me tell you... repartitioning all of the disks holding my data made me more than a little nervous at first :-) I kept backups of the partition tables, just in case, but it actually worked just fine. Next time, though, I think I'll just create a single partition and use LVM to chop it into pieces which I can RAID together. So I'll have LVM over RAID over LVM. Sounds weird, but it makes a lot of practical sense.

Re:you failed prob and stat didn't you?

[adamdeprince]

Now be nice. He is right, 0.07% is correct.

The chance of any one failing is the same as 1 - the chance of none failing. If you have n gadgets with a risk of failure of m, the risk of any one or more failing is 1-((1-m)^n). For 0.01% (0.0001) and 7, that puts your risk of any disk failing at about 0.07%. 1 - (1 - 0.0001 ) ^ 7) = 0.0006997901 ... or about 0.07%.

The idea is the chance of any one or more failing is really the flip side of the chance of none failing. If any item has a p chance of happening, then the chance of them all happening is p1 * p2 * ... pn. So, if we have a chance of one drives survival at .9999, the likelyhood that all will survive is that 0.9999 ** 7.

Re:Big Big Drives are great...but backup is a prob

[tomstdenis]

RAID-5 [or 6]. If you're running something where you have 750GB of information chances are you can justify spending 2-3K on reliable storage.

3x750 in RAID-5 would net you about 1.3TiB of storage and would allow upto one drive to completely die without losing data. If you're more paranoid you could use 4x750 and have upto two drives die.

The RAID access will be automatic so effectively you're always backing data up.

Tom

Re:Big Big Drives are great...but backup is a prob

[nuggz]

Buy 2 drives, use the second one for backup.

Put it in a USB adapter and use rysync.
Quick, easy, cheap.

Re:Big HUGE warnings - Not quite true

[drinkypoo]

you need to order new drives and ship them overnight as soon as you have problems.

Not only that, but you NEED a hot spare. If you have a RAID made up of a bunch of the same drive, and one fails, it should be a warning sign to you. The other drives are substantially the same as that one, and they should now be considered likely to fail, unless it failed WELL within the MTBF window.

Re:The real reason why this drive is great

[ars]

I'll give tou two tips:

1. Preview

2. Go into your prefs and change Comment Post Mode to 'Plain Old Text'.

Re:only 187 million times cheaper per bit

[WuphonsReach]

$413 sounds pricey until, as you noted, you do the math for the $/GB amount. For being a leading-edge drive, the price per GB is rather competitive.

The following prices are estimates based on www.pricescan.com. There could be as much as +/- 10% variation in prices.

PATA drive prices
120GB $64 - $0.53/GB
160GB $70 - $0.44/GB
200GB $75 - $0.38/GB
250GB $80 - $0.32/GB
300GB $105 - $0.35/GB
400GB $195 - $0.49/GB
500GB $260 - $0.52/GB
750GB $490 - $0.65/GB

SATA Drive prices ($/GB)
120GB $68 - $0.57/GB
160GB $65 - $0.41/GB
200GB $76 - $0.38/GB
250GB $80 - $0.32/GB
300GB $105 - $0.35/GB
400GB $175 - $0.44/GB
500GB $250 - $0.50/GB
750GB $434 - $0.58/GB

Re:Big HUGE warnings - Not quite true

[swillden]

Any modern nForce4-based motherboard handles raid arrays on 4 SATA drives, any additional chip is gravy (my 1 year old DFI handles 4 form the nForce plus 4 from a Promise chip)

This is perhaps the case. I don't use modern motherboards and processors for home file servers. Old stuff works fine -- but suffers from the drive controller capacity limitations.

Seven or eight drives won't strain a PSU (unless it's a cheap no-brand piece of crap)

My experience is different. I had a decent (but non-server) 350W PSU and when I added the fifth drive to my box I began getting intermittent failures reported by the drives. When pushing all of the drives hard, the failures were frequent and sever. I put a good 550W PSU in the box and the problems went away.

Cooling is likewise, while hard drives don't cool very well (they don't have heatsinks or anythink) they produce very few heat, just put a low-speed 120mm fan (low speed as in under 1kRPM, I'm talking Papst or Nexus here) in front of your drives (a fan for each 3 or 4 drives) and they'll keep well under 40C in a room at 25C.

But only if your case has airflow that is designed for that many drives. My case doesn't, and most cases not specifically designed for lots of drives don't, so they end up being placed in all sorts of odd locations, which means that cooling becomes problematic.

So, to sum up, if you're (a) using a newish motherboard, (b) a PSU designed for server, or at least high-end workstation use and (c) a case designed to hold and cool lots of drives, then you don't have any of the problems I mentioned, because you've already bought the components to solve them. If you're building a home file server the way most people do, though, out of the machine that was your desktop box several years ago, they're real problems.

Some controllers are also able to extend (or even fully replace) arrays out of the box. You usually don't find them on consumer-grade motherboards though.

Not to mention the fact that if you use hardware RAID you're tied to that particular hardware's implementation of RAID. If that controller dies (unlikely, but possible), you need to replace it with the same sort of controller. For a data center that's not an issue, since you just purchase a service contract that ensures the components will be replaced as needed. For my home use situation, it make me nervous to tie my data to any particular controller. Software RAID is the safer bet.

Re:Big Big Drives are great...but backup is a prob

[Anonymous Coward]

You can also leave it in the case, and then periodically spin up-rsync-spin down. Here's my crontab:

0 4 * * 0 sdparm --command=start /dev/sdc ; mount /dev/sdc1 ; rsync -av --delete /sdb/* /sdc ; umount /dev/sdc1 ; sdparm --command=stop /dev/sdc

NewEgg

[blogeasy]

Only $440 from NewEgg.Com with 3-day shipping.

These drives are needed for a PetaByte array

[Terje Mathisen]

Oil companies which do a lot of Pre-Stack processing, i.e. raw seismic data, need an awful lot of disk storage: We're currently in the ~50 TB (geographically mirrored RAID5 servers) range, and this is with Post-Stack only.

Going to Pre-Stack will generate 10 to 100 times as much data, which means that 500 TB to multi-PB is where we'll be in a couple of years. Having 750 GB SATA drives in a Nexsan SATABeast http://nexsan.com/products/products/satabeast/sata beast.html enclosure results in about 27-28 TB of usable disk space in a single 4U rack unit.

Very nice!

Terje

Re:Big Big Drives are great...but backup is a prob

[jabuzz]

How about a DLT-S4 tape then. They hold 800GB of data native. At no time to the best of my knowledge in the last 10 years has the largest hard drive ever held more than the largest tape. Before that I don't know.

Yes they are expensive, but that is because people don't backup so the volumes are to small. Chicken and egg situtuation really. Only something like two million DLT drives of all types have ever shipped. In the same period it is probably more like two billion hard drives that have shipped. Hard drives are so cheap because the volumes are there to make it so.

Re:That's a lot of DVDs

[QuoteMstr]

I actually just implemented a DVD-based backup solution for my employer. While DVDs are much smaller than tapes, they're much cheaper, especially when buying in bulk. They're so cheap, in fact, that we can afford to do nothing but level-0 backups, which eliminates a great deal of complexity. We deal with the optical degradation problem by using error-correcting codes to gaurantee against the loss of one disc from any given backup set --- and we usually make multiple sets anyway. All in all, it's not a terrible system.