Seagate 7200.12 1 TB

[LiamC]

The 7200.12 put through it's paces.

2 plater 1 TB.

http://www.techreport.com/articles.x/16472

Hit or miss...

 

[Mercutio]

Hit or miss is very much my experience with recent Seagate drives. I've relegated the 750GB units I have exclusively to online backup. A low heat two platter unit sounds great and I hope a better drive company gets around to producing one soon.

 

[Stereodude]

Certainly a mixed bag from the review.

 

[Fushigi]

Certainly a mixed bag from the review.

Yes. It doesn't sound bad; just not exceptionally good. Basically average. Looking at the performance spread on all of the drives in the tests I really don't think any of them are clear losers. Average performance is pretty good and changing brands or models only helps (or hurts) by any significant amount for a very few specific things.

And it would seem that from the tests the Raptors aren't worth, or perhaps are no longer worth is better phrasing, much of a price premium.

Modern desktop drives seem to be evolving more and more slowly. Performance gets a little better over time; power consumption reduces a little over time; noise/heat are mildly reduced over time. What if anything can be done for a revolutionary - rather than evolutionary - leap in performance for platter-based drives? Smaller platters (2.5, 1.8")? Hybrids? I don't think spindle speed is the answer due to the higher power draw, noise, and heat.

 

[ddrueding]

If anyone is investing a significant enough amount into spinning platters to cause a revolutionary improvement, they should be fired. I would hope that everyone in that business is redirecting their R&D into something with a future (solid state).

 

[Santilli]

I wonder if that's true. I think the industry has worked together to only progress at a certain rate, a profitable rate. I think the great increase in capacity, and speed in platter drives is being driven by the threat of SSDs taking over the market. When you think about it, really isn't much of a reason for not producing SATA drives at 15K speeds, and, given the technology, I don't see why they can't move data at near 150 MB/sec, and, access times around 2 ms. You raid 0 two of those, and, you have a drive setup that's going to be very close to SSD's performance wise, at much less cost, with way more capacity...

 

[Mercutio]

really isn't much of a reason for not producing SATA drives at 15K speeds

Yes, there is. There's not enough demand for desktop drives that can reliably operate at 15k and still have the capacities and costs expected of mainstream products.

There's a reason that Raptors didn't take the world by storm, man.

 

[timwhit]

I wonder if that's true. I think the industry has worked together to only progress at a certain rate, a profitable rate. I think the great increase in capacity, and speed in platter drives is being driven by the threat of SSDs taking over the market. When you think about it, really isn't much of a reason for not producing SATA drives at 15K speeds, and, given the technology, I don't see why they can't move data at near 150 MB/sec, and, access times around 2 ms. You raid 0 two of those, and, you have a drive setup that's going to be very close to SSD's performance wise, at much less cost, with way more capacity...

How many people actually want to worry about setting up RAID 0? Not to mention that your failure rate is going to be terrible. So, now I have to worry even more about backups and restoration plans. I might also need to worry about keeping a spare drive on hand. If I can match that speed with a single SSD, that just makes much better sense.

 

[Santilli]

If you told me 5 years ago we'd have 110 mb/sec sustained data transfer out of a 7200 rpm drive, I'd have thought you had a screw loose. I really wonder what platter drives are capable of, if they really pull out the stops, and see what they can really do?

I suspect the cost points can be met, as well. I really think the cost of the drives is a semi-monopoly price fixing scheme, with limited players in the market. I can't help but think the actual cost of drive production is pennies, considering where they are made. There is a HUGE investment is drive production, and, I suspect we'll soon see what these companies are capable of producing. If a 7200 rpm drive can do 110 mb/sec, and the goal is around 200 mb/sec, I suspect that we will see developments that give platter drives performance around the same sustained data transfer rates, and, the cost point will be better.

Does anyone know if Seagate, or any of the drive producers are in SSD's yet?

I also think the product failure rates we are seeing is due to rushing to market new designs, before they are tested, due to the pressure from SSD's.

If David can build all his systems with SSD's, I see no reason the rest of the market won't follow soon. Laptops are begging for this kind of data speed, considering the processors and ram we now have.

 

[ddrueding]

The higher STR of newer drives is directly tied to the increases in aerial density, and could have been predicted easily. Just as easily, with basic understanding of physics, you know a larger platter cannot spin faster, and that rotation latency cannot be helped at a given speed.

I can actually see HDDs taking over from tape (cheap, slow storage), but I think tape might be better for that purpose.

 

[LunarMist]

I doubt that you want to increase aerial density. You mean areal density?

 

[ddrueding]

Yes, that would be the one.

 

[Handruin]

If a 7200 rpm drive can do 110 mb/sec, and the goal is around 200 mb/sec, I suspect that we will see developments that give platter drives performance around the same sustained data transfer rates, and, the cost point will be better.

Fast STR is great, but that's not the only thing to consider. When you combine fast STR with 0.1 ms access times in SSD, you can't beat the performance. I don't see spindle drives getting to 0.1ms access time in the near future...that's the performance benefit I like seeing in these new SSDs.

I would eventually guess that as the areal density of a platter grows so high that drives will eventually be prone to problems with even the most minuscule about of vibrations. I could be wrong, but I'm thinking about the situations that Intel and AMD have to deal with when produce 45nm and lower wafers. The machines creating these things need to be isolated from even the lowest of noises that come from the earth.

 

[Fushigi]

The only thing slow about tape is the seek time. STR for LTO4 is already 120MBps for uncompressed data and is slated to go up by 50% each generation for LTO5 and LTO6 along with the requisite doubling of cartridge capacity. Tape also travels a lot better for off-site storage that platter-based drives.

Tape is still a very good option for system backups and will remain so for a the moment, although if you do a lot of individual file restores, either disk based or Disk-to-disk-tape may be better.

 

[Mercutio]

End-to-end seek time on LTO4 is something like two and a half minutes. That seems pretty damned fast to me. I remember using a floppy-interface Travan drive as a random access device on my first Linux system. It was... substantially slower than that.

 

[P5-133XL]

My issue with tapes, is cost: Not performance.

Removable HD's with caddy's are outright far cheaper at least untill the data set is bigger than a single HD. Hd's are just as reusable as a tape.

I will agree however that HD's don't travel well and specificly require a kinder and a gentler world.

Recently, I've been playing around with mirroring on a removable drive as my backup. I then simply remove the drive and store; Then break the mirror, add a replacement drive and departition it; make a new mirror. My biggest problem has been the inability to script it so as to be impossible to mess up royally by a true moron: That departioning step is an unacceptable risk when a moron can pick any partition to kill and I havent been able to script choosing the correct drive to departition.

 

[snowhiker]

My biggest problem has been the inability to script it so as to be impossible to mess up royally by a true moron: That departioning step is an unacceptable risk when a moron can pick any partition to kill and I havent been able to script choosing the correct drive to departition.

Five or six years ago a friend of mine watched his boss FORMAT the production servers RAID array because he wasn't paying attention to which way the KVM switch was set.

 

[Mercutio]

I do mirroring with rsync. It's not perfect but mirroring is an availability solution not a true backup.

 

[P5-133XL]

It becomes a true backup, when you are removing the mirrored drive and storing it on a shelf. There is no effective difference between that and an image backup, other than it is much quicker to get it back up and running.

 

[blakerwry]

Yep, we have a few systems with RAID1 as a backup in addition to availability. We pull a drive and set it on a shelf. Insert a new drive and let the array rebuild. The drive on the shelf becomes our backup.

This type of backup can be performed online or off, and the rebuild rate can be set so as not to impact normal operation. Much faster and easier than using traditional imaging software.

Depending on your backup requirements you could fail and spin down the drive in software, rather than physically pulling it. This would allow automated backups without human intervention.

 

[P5-133XL]

Yep, we have a few systems with RAID1 as a backup in addition to availability. We pull a drive and set it on a shelf. Insert a new drive and let the array rebuild. The drive on the shelf becomes our backup.

This type of backup can be performed online or off, and the rebuild rate can be set so as not to impact normal operation. Much faster and easier than using traditional imaging software.

Depending on your backup requirements you could fail and spin down the drive in software, rather than physically pulling it. This would allow automated backups without human intervention.

How do you deal with the need for the replacement drive be unpartitioned? Of course, if the drive is new, that is not a problem but reusing old backups means that the drive has stuff on it. That then requires someone to pick the replacement drive from a list and de-partition. I am not willing for someone to do that because of the risk that they pick the wrong partition. I need a script to reliably determine the repacement drive out of the many drives on the server.

The only way I've figured out to do that is to do the departitioning on a totally unimportant machine before transfering to the server: that way if a mistake happens it won't be a disaster. I find that to be a totally un-elegant way that offends my efficiency sensability.

 

[blakerwry]

How do you deal with the need for the replacement drive be unpartitioned?

I don't believe that is a need that I have. The hardware RAID controllers we use will rebuild over a drive whether there is data on it or not...

Though I do keep it simple in these setups...

In the two situations where I've used RAID 1 as a backup mechanism we've used Windows and have 1 array which contains the OS+Data. Drive 0 always stays in the machine, Drive 1 is the mirror/backup.

I like to shutdown the server (usually while doing other maintenance that requires a reboot), pull Drive1, and boot the server back up. Insert the new drive, set as a hot spare or add it to the array and let the controller rebuild. Monitor from within the OS.

The new drive can be clean or dirty, doesn't seem to matter with Dell PERC/LSI controllers.

 

[LiamC]

To put the 7200.12 into perspective versus 13 other drives

http://www.xbitlabs.com/articles/storage/display/1tb-14hdd-roundup.html