Slashdot Mirror
Hard Disk Sector Consolidates Amid Uncertain Future
Hugh Pickens writes writes "The WSJ reports that Western Digital will buy Hitachi Global Storage Technologies for about $4.3 billion in cash and stock, leaving only four key hard disk drive vendors — Seagate, Western Digital, Toshiba and Samsung. The hard drive world has been seen as ripe for consolidation, particularly as the rise of tablet computers such as the Apple iPad — which don't use hard drives for data storage — is casting doubt on the future of hard disks. Compared to hard drives, solid-state drives promise greater power efficiency, performance, resistance to physical shock, and run more quietly since they contain no moving parts. But one area that solid-state drives do not improve on their spinning predecessors is in their inevitable movement towards failure. 'SSDs are going to fail just like hard drives will,' says Chris Bross, Senior Enterprise Recovery engineer at Drivesavers Data Recovery. 'Every storage device will have issues regardless of their underlying technology.'"
This isn't all that different from when Seagate bought Maxtor. Back then, after the sale, Seagate controlled 44% of the market, compared to nearly 50 percent market share which this deal has bestowed upon Western Digital.
vos nescitis quicquam, nec cogitatis quia expedit nobis ut unus moriatur homo pro populo et non tota gens pereat.
There will be no hard drives because we'll just store all our data in the cloud. (ducks)
For the end-user, it's great that the average lifespan of a drive is measured in years. For the manufacturers, not so good.
Since upgrading my power supplies I've had very few drive failures over the past five years. I've purchased drives to expand storage, but rarely to replace. Across 10 laptops I have replaced two failed drives in two years. On the desktops, with about twenty drives between 5 machines, I've replaced maybe two units in two years. These run continuously, are rarely rebooted, and have semi-annual reboots to replace fans and clean out the dust.
>>SSDs are going to fail just like hard drives will
Like saying old stuff fail and get replaced. You can't, as hard disk producer, market products with high or infinite durability and be profitable. You want your shit to fail (within reason) so that people will spend more money on your products. We can very produce a lightbulb that never dies but we're not going to.
After it seems clear the rewrite count is going to hell - 5000/cell for 32 nm, 3000/cell for 25 nm, SSDs are going to have a helluva time catching up in cost/GB. People will still want huge storage disks, data centers still need storage, hard disks aren't going away. The SSDs do rock for speed and is making huge performance gains but that doesn't bring the cost down. The combination of a blazing fast 100GB SSD and huge, slow 2TB HDD seems to be the way forward.
Live today, because you never know what tomorrow brings
Who bought Fujitsu?
The "hard disk sector" consolidates, hmm?
For a moment, I did a double take and thought of Stac.
Be very, very careful what you put into that head, because you will never, ever get it out. -Thomas Cardinal Wolsey
"...and to commemorate their latest acquisition, Western Digital announces a new line of ultra-green drives...a spokesman had this to say..."
"Yep, these drives are so power-conservative, they actually stop consuming power permanently 30% faster than our previous line. We're calling them 'Hitachies'"
"When I am king, you will be first against the wall..."
HDDs are NOT going away any time soon.
I bought 8TB of storage over the weekend for $300. You can get what, 120GB of SSD storage for that? I use 6TB of it (RAID5) so if one fails, I'll live.
Can't beat that value proposition with SSDs at the moment. Also, as process sizes continue to shrink, PE endurance is only getting worse. A good hard disk will last 10 years in normal use. I doubt we'll ever see an SSD last that long.
Once upon a time, Seagate and WD both sold reliable disks in the consumer space. Then, Seagate bought Maxtor, and Seagate's consumer grade disks quickly fell to match the reliability of Maxtor's consumer disks. Seagate's "Enterprise" disks, at 2-3x the cost, are still decent, but why buy those when you WD's consumer-grade disks are still reliable? And now we won't be able to trust WD's lower-priced disks, and we'll have to buy REA3s or 4s (or 5s?) to be able to trust our spindles.
It looks like the price of hard disks just doubled. Impressive.
'SSDs are going to fail just like hard drives will,' says Chris Bross, Senior Enterprise Recovery engineer at Drivesavers Data Recovery. 'Every storage device will have issues regardless of their underlying technology.'
I do not see SSDs playing a major role anywhere near the traditional large database especially in financial institutions. In our trials with PostgreSQL that had 17 tables, the largest of which had 23.1 million records and 9 columns on an DELL notebook, these drives failed after about week of intense read/writes!
My former boss, who was a closed source stooge blamed the DB. Others like me knew these SSDs were not yet ready for prime-time. By the way all this was about 2 years ago. Technology could have changed for the better now.
I've never been able to get myself to buy Hitachi drives after the deathstar episode.
http://en.wikipedia.org/wiki/Hitachi_Deskstar
boycott slashdot February 10th - 17th check out: altSlashdot.org
"solid-state drives promise greater power efficiency, performance, resistance to physical shock, and run more quietly"
And cost 10 times more. I can buy a 2 terabyte hard disk drive for ~$100. Can I do the same with solid state? Nope. (That is why Nintendo and Sega moved from Solid state cartridges to discs - they cost less per megabyte.)
Information wants to be expensive AND wants to be free. So you have Value vs. Cheap distribution fighting each other.
Not really improved. I burned out a REALLY GOOD (best available) SLC SSD in 7 months with a mirrored production workload at a previous jobsite not that long ago.
Poof. All gone.
At the FAST conference, was yet another presentation on SSD lifetime burnout mechanisms, news not actually improving in the slightest so far on life. SLC is not good enough; MLC is toast in write-intensive apps.
Phase-change memory or one of the others, with millions of write cycles per bit, may pull this out, but Flash is not proving good enough for enterprises.
On a different note:
"But one area that solid-state drives do not improve on their spinning predecessors is in their inevitable movement towards failure."
I would argue that it's actually much worse. It is possible to recover most or all of the data from most hard drives that fail. Try that with the newer SSDs.
I disagree that "solid-state drives do not improve on their spinning predecessors is in their inevitable movement towards failure." SSDs wear out gracefully so that you can still read your data after many failures. Spinning drives just die and you go to a backup. To me, that's a major improvement.
The HDD death has been predicted a few too many times...
Its still the cheapest storage with easy access out there.
Consolidation is not only expected, but somewhat necessary.
I spent 15 years in the HDD industry, and some things to understand:
- It takes roughly 70 people and 6-9 months to design and develop a new disk drive.
- product lifetime has been as short as 2 months and as long as 1 year.
- typical product lifetime is 3-6 months.
- A company needs to have multiple design teams doing multiple product designs phased for phased product releases.
If the product is late, its already obolete, and will not sell.
If the product is slightly behind the times, it will not sell.
Because of the above NRE expenses are huge, so margins or volumes have got to be huge, to make any money.
Margins went to nothing many years back, so the volumes need to be huge. Thus fewer players are the results of all that.
Because of the above, dozens of companies that used to make disk drives are now long gone.
All of that said, the "death of the HDD has been greatly exaggerated"
- its cheap, high volume storage, and all in all "fairly" reliable.
www.effectiveelectrons.com "chips that work" Analog, RF, Mixed Signal
If my information is correct, the number of defects for SSD's are about the same as HDD's, with the exception of the Intel SSD's, which cut the number of returns in about 4 (can't find the article using Google, if anybody has a link?). I've returned mine because of a failed firmware update to remove a controller bug. I would not be amazed if the actual number of failed drives is about 8 times lower than HDD. So sure they fail, but I think that the failure rate will be more like that of DRAM than HDDs. And current HDD's already have a rather low return rate IMHO.
As for the early adopters: this is definitely the year of the SSD, but don't expect a super smooth ride yet. I've got to upgrade my Vertex 2 because of power management issues, but somehow I cannot flash the drive (cannot put it in secure mode). My USB to S/PATA connector does not even support SATA commands. Really annoying stuff - but you'll still have to pry the SSD's from my cold dead hands :)
I think both your boss and you need a refresher course in technology.
Two years ago, Consumer SSD didn't have proper TRIM features. The drive probably didn't die, just needed a wipe and rewrite. And you were probably using consumer grade OCZ drives, and not the better (and way more expensive) commercial drives, which had the better chips in them (and TRIM).
If you were doing Database transactions on SSDs, you'd realize that there is no way for HDD to compete with SSD in IOPS. If you really wanted and needed IOPS for your Database (doesn't sound like you did, since you're using a laptop), then the cost for the drive would barely register in the decision to buy the drive.
Suffice it to say, your anecdotal evidence means very little to those people who can see the benefit of SSDs. My professional opinion is that most of the SSDs being sold right now are going to Database Server farms, for the IOPS alone, and only the ones that don't make the grade end up on the consumer market.
Agent K: A *person* is smart. People are dumb, stupid, panicky animals, and you know it.
Yeah, ok.. How many factories or labs, etc are there competing against each other? This is a lot like petroleum. There's not much competition at all.
For justice, we must go to Don Corleone
considering SSDs have only a limited number of write cycles, database work with heavy writes would likely be better served by RAM disks if you want ridiculously low access times and very high speeds. Something along the lines of a HyperOS HyperDrive or an ACard ANS-9010 / 9010B would likely be better suited but those solutions also have their own issues (namely a very steep price and loss of data when the battery runs out)
Bad news, especially for the enterprise users:
- HGST drives quality is above the rest of the industry, it may easily change after the acquisition.
- HGST are often willing to invest in relatively niche products (recent example is 3TB 7.2k drives with SAS interface, no one else makes them). WDC will probably kill any product line that doesn't sell really huge quantities.
"My former boss, who was a closed source stooge blamed the DB. Others like me knew these SSDs were not yet ready for prime-time. By the way all this was about 2 years ago. Technology could have changed for the better now."
That is almost certain. First of all, everybody that is serious in the field will tell you that you that that kind of application requires a enterprise (read: SLC flash drive). Chances are that the SSD you've tested with was an older drive with the failed Micron chipset, or maybe even older tech. If you compare those niche products with e.g. a well tested Intel SLC drive with TRIM support, you'll see not just a huge performance boost over the older SSD. Don't forget that earlier flash drives were mostly of interest because of high reliability (as in: crash resistant) and power and weight ratios.
Of course, if you're a real player with some money to spend:
https://shop.sun.com/store/product/8be96180-a21a-11dd-a2a2-080020a9ed93
# Over 1 million IOPS
That oughta do it, especially made for DB applications :)
For home users, the IOPS of a Vertex 2 drive should be ample, for workstations I would still go for the G2 - also because of reliability.
While SSDs and HDDs serve the same function the technologies are pretty different, so it's much easier for Intel and various RAM manufacturers to start making SSDs than it is for WD to transition to them.
Last I checked WD's SSDs were just a rebranded product made by some other company.
I guess Hitachi Global Storage Technologies have all they need to manufacture SSDs in house and I'm assuming the other HDD companies will have to make some acquisitions of their own to stay competitive.
I've tried to do large database server farm tests on modern enterprise SSDs with TRIM, the best wear load leveling, SLC, etc. They go "poof" at moderate (few months, for my loads) lifetimes.
IOPS x Lifetime / price is a metric I find useful. Unfortunately, it makes SSD look even worse than it does just on a price basis 8-(
Before DNS you'd give the specific route you wanted your email to travel. So instead of a simple flanders@ersys.com, you'd address it like decrwl!alberta!aunro!ersys!flanders. Since there was no reason you couldn't send email to yourself, all you had to do to gank a little extra storage was uuencode your payload and mail it to yourself by the longest route possible. Then set up a .forward file to automatically re-send the email once it made it around the loop. Some email servers would transact just once a day, so you could really add to the latency if you included a couple of those in your address path. And, yeah, people actually did this.
...that the submission should have had a different title:
Hard Disk Sector Defrags ?
I use irony whenever I can, but my shirts are still wrinkled...
This brings up an interesting issue: will SSD manufacturers start pulling dirty tricks like shipping suboptimal wear-level algorithms and thwarting attempts to install or use third party firmware (that would presumably have better wear levelling and therefore increase the life of the drive)? It would be pretty easy for them to market "consumer" grade SSDs that break down faster than "industrial strength" SSDs. I could even see SSD manufacturers using the DMCA to try to stop hackers from distributing better firmware images (yes this is a bit of FUD, but it seems plausible).
Palm trees and 8
Enterprise SSD units used to be just lots of DRAM with a big battery and a spinny disk.
The big battery gave you time to flush the RAM to physical disk in the case of a power loss.
A Pirate and a Puritan look the same on a balance sheet.
Yes they will fail, but the failure modes will be different. No heads to crash into the spinning surface damaging the oxide layer and spreading bits of junk across the rest of the surface of the drive.
People will still fail to take adequate backups - so there will still be a market for data recovery from failed SSDs - I wonder what it will look like. Pulling the raw flash chips out of a failed SSD will most likely allow an enormous number of bits to be recovered. But unless the drive manufacturers cooperate with the data recovery companies to provide low level details of the wear leveling algorithms, then it will be astonishingly hard to turn those raw bits back into files.
Are any data recovery companies advertising recovery services for SSD yet?
Ben - Is that you? You're fired.
With data density approaching 100GB/inch^2 you would want to make sure the platters are shattered into what passes for dust in normal circles. Probably better to just use the security erase feature of newer ATA drives, or better yet only put encrypted data on the drives in the first place.
Raise your hand if when you first read the article title, you thought that "Hard Disk Sector" was a literal, physical, hard disk sector.
In the cloud, a smart system can store a single copy of any given file and some pointers to it. That way stuff that gets passed around via email, or even multiple people "downloading" stuff from someplace don't actually create copies in the cloud. Everyone just thinks they have their own copy stored in the cloud. Even online "content" can be compressed since every "news outlet" just copies a blurb from someone else - this is a bit more complicated since they usually make a couple minor edits to inject their own opinion. I think with efficient storage like this, the entire net should fit on a couple 2TB drives ;-)
Furthermore, HGSTâ¦has been a price aggressor in the HDD market, so its being taken over by WDC is an incremental positive for both WDC and STX.
IOW, Western Digital is spending 4.3 Billion dollars so that they can buy out their biggest competitor and raise their prices? Ruh Roh. If WD kills the nifty 3TB 7K3000 (which can already by had for $170 shipped when on sale) and doubles the price of their upcoming 3TB Caviar Black I'm gonna cry. There goes their only even somewhat reliable competitor in the large capacity 7200 rpm market segment. Now they will have a monopoly on that type of drive. If you want a 7200 rpm cutting-edge areal-density drive that isn't gonna break in the first few months it will now be WD or the highway. Samsung makes some very reliable drives, but they have given up on 7200 rpm spindle speeds and aren't even close to cutting edge in terms of high areal density + capacity drives.
Quite an experience to live in fear, isn't it? That's what it is to be a slave.
I see what you did there.
"Nine times out of ten, starting a fire is not the best way to solve the problem." - my wife
Depends what they were trying to do. If they were trying to actually run the database on the SSD through a filesystem then the row updates would likely create a very large degree of write amplification, particularly if the updates were fsync()'d on a fine grain (preventing the drive from doing any real write combining). Even aggregation wouldn't help much for random row updates. In simple terms, a 50-byte fsync'd update to a row would result in a 128KB erase/write on the SSD. With aggregation and a SSD which does proper write combining that 50-byte (random) row update would only cost ~512 bytes. Still a very high write amplification cost but nowhere near as high as 128KB:50B.
On the otherhand, just putting the database's replay log on the (MLC based) SSD and fine-tuning aggregation of the transactions to reduce write amplification effects would probably work quite well. No write combining would be needed and write amplification could probably be reduced to 4:1.
Either way SMART should have told (the original author) that the SSD was about to fail. If the SSD failed prematurely before SMART said it was going to then that's an issue with the firmware and not so much an issue with the base technology.
-Matt
I agree completely with this assessment. TRIM is no magic wand, or even close to one. You can get the equivalent simply by provisioning a little less space than the SSD reports. e.g. on a 40G factory-fresh Intel SSD I provision 32G and leave 8G untouched. For all intents and purposes that has the same effect as a heavily TRIMed SSD, and is more consistent to boot.
The performance issue with TRIM is actually related to the AHCI command set. TRIM commands cannot be natively command queued due to the massive brain damage Intel injected into the AHCI standard to differentiate it from the SAS standard. Because of that *ALL* active reads and writes have to complete and the TRIM command must be run synchronously with the drive basically quiescent. That makes TRIM a non-starter for anything but bulk clearing operations.
-Matt
I feel that the memory guzzlers are Multimedia files. The rest are insignificant. What if there were a better compression technique for Multimedia? Like say 50% on existing compression? How would a technology like this affect this discussion about harddisk and cloud storage? This would even reduce the strain on bandwidth.
(http://www.tsstorage.com/tsst/corp_e/indexe.html)
HermesPod: Free Podcast Download Manager for Windows