Slashdot Mirror
Seagate Releases Hybrid Hard Drive
An anonymous reader writes to tell us Seagate has released a new hybrid hard drive. This new drive adds the speed of a solid state drive to the conventional hard drive. Originally designed for laptops this new drive comes in 80, 120, and 160 GB flavors and features 256MB of flash memory.
Didn't samsung or some such outfit already do this?
I've seen one of these at a trade event in Atlanta earlier this year. The idea is great, and after much strenuous testing, seemed to still work great. I can't wait to get my hands one some!
"Snatching defeat from the mouth of victory on a daily basis."
If I'm not mistaken, the flash memory serves to save things in a more quickly accessible memory when your computer goes to sleep or hibernates, allowing for an extremely quick "awakening". Hence these were designed with the laptop user in mind.
We figured out a long time ago that it's easier to elect seven judges than to elect 132 legislators.
http://www.pcworld.com/zoom?id=138102&page=1&type=table&zoomIdx=2 -attached to- http://www.pcworld.com/article/id,138102-c,harddrives/article.html
Both hybrids, Samsung AND Seagate were not only more expensive, they were considerably slower in tests vs. a traditional harddrive. I understand the drive to be green, but I think I'm going to wait a few years before jumping on this bandwagon!
this?
I have excellent Karma and I am not afraid to Troll it.
Can't read the article but this will help understand about the Hybrid drives.
Since laptops can't support the faster speeds that their desktop brethren, any access time improvement is desirable. You can keep your most frequently used data on the Flash or as a buffer, such as during a movie. Another benefit is that flash takes less energy to read than a HDD.
Here's also a review of the drive itself
import system.cool.Sig;
For Hybrid Hard Drives to live up to promises. After a bit more digging - There is still a lack of results from this drive, although boot time and power savings are starting to show up. RAM caches have been around for years, and getting even 1 GB of flash memory is getting down to pretty reasonable levels. Why is this commanding a 30% premium and delivering unspectacular benefits? Unless there's a solid standard behind addressing for HHD's exists, there's no point in blaming BIOS or Vista for a problem that could also be addressed in on-drive logic.
Meh.
http://www.seagate.com/ has a press release on their home page.
Actually, you have: http://msdn2.microsoft.com/en-us/library/ms940846.aspx (it's the first thing I do on a new installation of Windows)
Integration can solve a key point which is data integrity during an abrupt power event.
(see above).
AIK
I read an article somewhere that showed how a flash based drive could outlast a platter drive by efficient use of an algorithm that rotated through the bits. I don't recall any further information on this though, such as performance impact. Sorry for the lack of a link. I am sure you can google it though. :)
I do not respond to cowards. Especially anonymous ones.
I'm imagine they'd use this as a write-through cache. When you write data to the disk, it stores it in flash. Because a write-through cache can be quite effectively implemented in a ring-buffer (with reordering within a moving window for efficiency), you get perfect wear levelling without any complex controller logic. That means that it will work for writing 256MB times the number of rewrite cycles. Cheap flash has 10,000 rewrite cycles. My current laptop has been on for 30 days and has written 172.85GB to disk in this time. That gives 5.76GB/day of writing, or 23 complete write through the cache per day ignore, for now, that some of those were large linear writes, which would probably want to bypass the cache). For 10,000 rewrite cycles, with this usage pattern, it would take 435 day (1.2 years) to wear out the flash. This is, as I mentioned, assuming very cheap flash. Slightly more expensive stuff can get 100,000 rewrites, giving 12 years. If the mechanical parts of a laptop hard drive lasted 12 years, I would be very impressed. They should last longer with this kind of system, because it can batch writes a lot, and reduce the frequency of spinning the drive up and down. You also won't need to spin up the drive to read back data that you've only just written, which could help some poorly performing swapping algorithms (i.e. all of the ones used by 'modern' operating systems).
By the way, flash has a slight weird characteristic that you can write to it with a byte granularity, but only erase it with a block granularity, and it's the number of erases that cause the problems.
I am TheRaven on Soylent News
What is ReadyDrive:
http://www.microsoft.com/windows/products/windowsvista/features/details/performance.mspx
I'm summarizing what I learned from the German c't computer magazine, which has tested the various new technologies like ReadyDrive and others in Vista and also tested Flashdrives and Flash memory in general. Read the current issue of this magazine for in-depth analysis.
1) Pure Flash disks have only ONE advantage over harddisks: they are less sensitive to mechanical stress. In real-life scenarios, they don't safe power, and they are most definitely not faster than 2.5 inch drives. They ARE faster than 1.8 inch ones often used in ultra-mobile PCs, so there they indeed provide a benefit. For everyone else: especially write performance sucks compared to modern 2.5 inch disks, and read performance is at most en par. True, they don't need to position any heads so random access should save time - but according to the real-world tests made by c't that benefit isn't noticeable.
2) c't testers were very suspicious about how long Flash memory could survive as HD replacement where writing happens all the time, and yes, Flash cells have a limited lifetime, one cannot write too often. That's the theory. In practice c't testers were unable to make even the cheapest Flash USB stick show any sign of memory loss even after something like 16 million write cycles, when they gave up further testing because that's many many years of real-work usage. (pg. 104 of c't 21/2007)
3) Intel TurboMemory or MS Vista SuperFetch, ReadyBoost or ReadyDrive were shown to provide no measurable benefit AT ALL.
Suspicion of Hitachi and others seems to be that the current implementation in Vista isn't quite finished and SP1 should provide an update, and second the amount of Flash memory is waaaaaay too small.
Original article (German): http://www.heise.de/ct/07/21/100/
http://en.wikipedia.org/wiki/Wear_levelling
Nah, it's not worth it. Turning off last access time (BTW, it's turned off by default in Vista :) ) cuts my C++ project building time by 30%. I don't think any kind of intelligent defragmentation will be better.
Actually, this is a persistent read cache. As mentioned in another reply, flash memory is much slower at writing but very fast at reading and it doesn't need to be spinning to be read. This allows for much faster boot times in laptops/etc because it doesn't have to wait for the disk to spin up.
And don't SATA drives typically have better than 100MB/s write speeds?
That seems fast to me (copying one drive to another maxed out at about 40 MB/s, but I had an older SATA drive so maybe newer ones are faster), but that's for sustained transfer. If you're doing stuff like metadata operations, which are all journaled, it looks more like "write a journal block saying you're updating this bit of metadata, seek to the inode or indirect block you're modifying and write it, seek back to the journal and write a commit message". IRL these are batched, so the overhead isn't horrible, but there is still plenty of seeking involved.
In fact, ideally the drives would ensure that streaming writes don't push out the metadata from the flash cache.
That said, letting the drive run slower and stuff for longer life and better power consumption is very attractive. I don't actually know what exactly their goal is.
No, flash memory can be quite a bit faster than that. Most of the time the limiting factor is something other than the flash. USB2 can only do 480megabits/s, and that is bursts, using something faster than that would be a waste. Even having something that can do 480 tends to be a waste as most of the time the transfer is much slower.
As for SATA drives they don't normally do 100MiB/s unless the information is already in cache on the driven. The flash memory is basically there to be a larger cache which is persistent across boots, allowing for the bootloader, kernel and a few essentials to be guaranteed a faster access time. Any additional items that go in there depend upon what the specific manufacturer specific algorithm does.
The size of the flash is like the size of cache on a harddisk, bigger isn't necessarily better. You could give a HD 30mb of cache, but if it is using a poorly designed caching algorithm, the difference can be nonexistent if the important stuff isn't in it.
In this case, 256 ought to be enough for present day computing. Even Linux is a fraction of that size, sure you could include a few utilities that regularly run during start up along with the kernel, but when you start to get beyond a quarter gig, you are beginning to get into things that run less predictably.
Here is an alternate article for the slashdotted original:
This really isn't about reads but writes. By using this they can collect writes better (so they have to move the spindle less), cache the writes here (so they can avoid spinning up the disk longer), and protect writes (write in to this, power goes out, data still safe... RAM wouldn't do that). There isn't really much point to this for reads, as just sticking a little more cache (say 64MB) on the drive would work just about as well there.
Comment forecast: Bits of genius surrounded by a sea of mediocrity.
Having spent hours, days, years studying the effects of hard drive defragmentation, let me put the kibosh on 'intelligent defragmentation' here and now.
Defragmenting the files themselves gives about 20% of the potential benefits of defragmentation.
Defragmenting the file allocation table (FAT on FAT/FAT32 file systems, or MFT on NTFS file systems) gives the remaining 80% of the performance boost potentially given by defragging.
In the big scheme of things, it honestly doesn't matter whether the most recently used files are at the beginning of the drive, next to each other, or on opposite sides of the drive - if the file allocation table (or MFT) is sufficiently fragmented. Frag out the FAT/MFT bad enough over time, and simply defragging the MFT/FAT will make your computer run an order of magnitude faster.
Want the bad news? Windows doesn't ship with a FAT/MFT defragger (well through XP. Not sure about Vista.)
Only way I know to do it is with aftermarket software like Diskeeper (excellent product, BTW, 99% of the time.)
Glonoinha the MebiByte Slayer
Normally flash has a 10^7 erase/rewrite cycles. (from text book) It might be more for the new-age tech. But it is far lower than magnetic drive anyway.
It works on any Windows since Win2k.
You can also use fsutil utility to do the same thing: http://www.microsoft.com/resources/documentation/windows/xp/all/proddocs/en-us/fsutil_behavior.mspx?mfr=true
Better not tell my yonder XP box, then. Its HDs have been asleep all day, the lazy things.
But I turn off indexing service, which doubtless makes a big difference.
~REZ~ #43301. Who'd fake being me anyway?
No, that's not correct. Samsung is shipping hybrid hard drives for over half a year now (see, for example, http://www.engadget.com/2007/03/07/samsungs-hybrid-hard-drive-hhd-released-to-oems/).
Recently, they even blamed Microsoft for the poor performance of hybrid hard disks on Windows Vista (in German, http://www.heise.de/newsticker/result.xhtml?url=/newsticker/meldung/97021&words=Samsung%20Hybrid&T=samsung%20hybrid)
I'm not sure that would be good. If you believe Google's findings, you'll kill your hard drive's bearings doing that sort of thing.
If you go back and (re)read Google's study, you'll find that Google has little to say about power cycling harddrives, as they let them run continously. While Google did note a weak correlation, they speculated that this might be caused by already problematic machines that needed to be powered down and repaired more often.
I agree that convential wisdom does say that lots of power cycles is bad though.
Keep every disk less than 70% full and you'll cut way down on fragmentation. Go above 85% full even once, and Windows will hold a grudge. Diskeeper helps calm it down.
Get two or more drives (budget permitting). Partition drives into fast/slow areas. Thousands of small files (random seeks) are a lot faster on the first partition of the second drive (i.e. temp files incl. web browser cache). For Windows, set up the TEMP and TMP environment variables to that partition. I usually do 20-50% for random seeks on the outside of the platter and the rest for larger files that don't seek a lot.
Split the swap across all drives, and increase RAM out the wazoo; file cache is always good. Windows still uses the swap file even with 1GB free... I don't understand why, but the guys in Redmond thought it was a good idea.
Caveat: Move stuff back to the primary drive if you ever need to remove that second drive, or you may get boot time errors.
Pay attention to your own usage patterns (i.e. random or sequential)... if your disk is busy a lot, consider moving an app or two to another physical disk. Development environments can be heavy on random seeks.
I run Visual Studio 2005 Team System at my office; load times with a two-drive setup like that are about 3x faster than a one-drive setup. If your work is video editing or large sequential reads, a RAID setup might be better.
Don't upgrade a 4-year old machine to Vista. If you want Vista, just buy a new one.
The ideas behind this are applicable to any O.S. and there are proposed standard ATA commands to manipulate the Non-Volatile cache, see http://www.t13.org/Documents/UploadedDocuments/docs2007/D1699r4b-ATA8-ACS.pdf. I hope Linux and Mac hackers are working on it.
I'm not sure if the drive takes advantage of NV cache without specific O.S. support. Even without O.S. support, the drive could decide "You keep reading blocks X Y and Z, so I'll store them in NV cache" (drives already do this with their conventional RAM cache, typically 8MB) and "I'll keep your pending data writes in NV cache while waiting for the disk to spin up".
Windows Vista's ReadyDrive takes specific advantage of this feature: "During shutdown or hibernate all the disk sectors needed to boot or resume are pinned into the NV cache... Offsets within files and/or specific LBAs can be specified by the PC OEM in registry for pinning in the NV Cache". I converted the MS PowerPoint presentation http://docs.google.com/Present?docid=dcvvrqtp_13gj635t&fs=true (yay Google Docs, die PowerPoint DIE!).
=S