backtop

Print E-mail del.icio.us 39 comment(s) - last by mindless1.. on Nov 16 at 10:51 PM
New file system ups the performance of flash drives greatly, according to SanDisk

Solid state drives (SSDs), despite a rough economy, have carved out a solid niche, thanks to plummeting NAND prices and the promise of improved reliability and performance.  With the upcoming Windows 7 promising an even sweeter boost for flash-drive equipped PCs, it seems a good time for the disk hard drive alternative.

Now SanDisk Corp., a major player in the flash drive market, has unveiled a new file system fine-tuned for flash drive performance.  The new system ExtremeFFS, according to SanDisk, offers 100 times faster random write speeds.  The system is set to ship on SanDisk products in 2009.

SanDisk has also created two new metrics -- vRPM and LDE -- as part of its solid state campaign.  The vRPM measure details performance between a SSD and a hard disk drive (HDD) or another SSD.  The LDE metric measures drive lifespan.  SanDisk believes both metrics will help it show off its new file system.

According to SanDisk, the new file system and new benchmarks are "critical" to understanding and implementing solid performance in Windows Vista.  Rich Heye, senior vice president and general manager for SanDisk's Solid-State Drive (SSD) Business Unit remarked, "SSDs will revolutionize client storage, but we need new benchmarks that allow them to be treated differently than HDDs."

SanDisk has worked on flash memory file systems since 1994, when it debuted the TrueFFS file system.  The file system was used on mobile devices and incorporated into Windows at one point.

However, despite this experience, SanDisk's drives have been performing very poorly in Windows Vista where competitors shine.  The company is hoping that its new file system will change this and boost its fortunes.  According to Mr. Heye, the large number of random writes in Windows Vista are a critical issue to performance.  He stated, "The mismatch to block size is significant."

The company described their new system, explaining:

[It uses] a page-based algorithm, which means there is no fixed coupling between physical and logical location. When a sector of data is written, the SSD puts it where it is most convenient and efficient. The result is an improvement in random write performance -- by up to 100 times -- as well as in overall endurance.

The new system allows independent NAND channel operation, allowing some channels to be reading, while others are employed with writes or garbage collection.  The new system also "learns" usage patterns and groups data together on the drive smartly to improve performance among most commonly used applications.  Mr. Heye continued, "This feature might not show up in benchmarks, but we believe it is the right thing to do for end-users."

The new offerings may help SanDisk, which has struggled economically of late, despite demand for solid state drives.  The problems led to Samsung walking out of talks about a potential acquisition last month.

Comments     Threshold


This article is over a month old, voting and posting comments is disabled
DOA
By mikefarinha on 11/6/2008 11:46:37 AM , Rating: 2
This new file system is DOA, at least for Windows installs. Windows relies heavily on the built-in security features of NTFS, coming up with a new fangled file system is a non-starter in the Windows world.




RE: DOA
By Klober on 11/6/2008 12:32:48 PM , Rating: 4
By default, flash drives use FAT or FAT32 for their file system because NTFS has so many extra write operations associated with it. Unless you change some advanced settings in Windows, every time you read or write a file there is at least 1 additional write because of logging. This obviously equates to measurably higher wear and tear on flash drives with their limited writes.

In other words, for the vast majority of users (99.9+%) NTFS plays no role what-so-ever in the use of their flash drives, and unfortunately makes the argument you base your opinion on a non-issue. :)


RE: DOA
By Smilin on 11/6/2008 1:33:55 PM , Rating: 3
I think his argument has merit.

Abandoning MFT journaling because of limitations in one company's SSD drives wouldn't be wise.


RE: DOA
By Klober on 11/6/2008 1:59:14 PM , Rating: 2
Everyone is entitled to their opinion. In mine, journaling is one of the, if not the biggest, reasons not to use NTFS on an SSD. You're welcome to keep your NTFS with journaling on your SSD and kill it 2-4 times as fast. Me? I'm either going to use FAT32, ExtremeFFS (even as dumb as the name sounds, if it works I'm down for faster writes - all else being equal of course), or NTFS with journaling and indexing disabled.


RE: DOA
By mikefarinha on 11/6/2008 2:00:21 PM , Rating: 5
um.. 99.9%?

You cannot install Windows on anything other than an NTFS partition. So for 100% of Windows users NTFS plays a very significant role.

And without a damn good reason they're not about to change because some company cannot get their act together.


RE: DOA
By Klober on 11/6/2008 4:08:41 PM , Rating: 2
My apologies. You're right, I should have put that differently. I was speaking mostly of USB flash drives for much of that post - specifically when talking about flash drives using FAT or FAT32 by default and 99.9+% of users not being effected by NTFS on their flash drives.

However, that does not invalidate my argument completely. Flash drives of any kind still pay a high price when using NTFS with its journaling because of the constant writes for pretty much every read or write operation.

In my opinion, forcing use of NTFS under Vista for any "hard drive" (quotes because the lines are now skewed between the classic hard drive and the new SSD drives, although the term is used interchangably by Windows) is not right considering the setting to disable journaling is global (afaik - if I'm wrong please correct me). Either it's on for the OS or it's off, it's not a per device setting. Now, if Windows dynamically enabled journaling based on whether the device is a mechanical drive or an SSD, that I could understand. Unfortunately, if NTFS is in its default state it is currently a really bad idea for use on SSDs (or any other flash based storage device).


RE: DOA
By mikefarinha on 11/6/2008 4:47:05 PM , Rating: 2
Every flash drive I've used comes pre formatted in some form of FAT. But that isn't the issue at hand.

SSDs will be the replacement of HDDs. So for them to take the place of current HDDs they will need to accommodate everything the HDDs do, including journaling. NTFS also has important security characteristics built into it which also carry a little overhead. This is the nature of the current IT landscape. IT cannot move backwards in these areas and thus if a new technology cannot unequivocally replace an existing technology then it should fail in the market place. We shouldn't make excuses for these issues.

Obviously NTFS was created years ago with only magnetic drives in mind. With each release of Windows NTFS has also been revised and updated. Eventually NTFS will have to cope with the characteristics of SSDs, hopefully the transition will be seamless but we also should hold hardware manufactures to comply with these realities in order to move forward.

Or, working together, hardware manufactures and OS makers(eg Microsoft) need to come together to agree on a new format standard going forward.

I don't believe the SanDisk policy of blaming others for their problems is the most effective way of moving a technology forward.


RE: DOA
By Klober on 11/6/2008 6:07:08 PM , Rating: 2
I'm not arguing whether SSDs can or can't support the full feature set of NTFS since obviously it can, I'm arguing that with the current incarnation of NTFS and the way it works with the full feature set enabled it is a really bad idea for use on the current and near future generation of SSDs.

You mention IT a lot in your post, and I agree that NTFS has many necessary functions for the IT sector. However, many, if not most, of the "features" of NTFS are not necessary or even utilized by the common home user, and most of those features reduce SSD life (and also reduce available storage area, which won't be a problem in the future, but definitely is for the current generation of SSDs). The extra features should have better explanations and easier options for home users to disable them at the very least.

I realize that NTFS has been revised over the years through the various OS releases, but maybe it's time for an all new file system. Can they really revise NTFS enough with it's current feature set to make it SSD friendly and still backwards compatible? My guess is probably not, in which case we might as well go with a new file system if backwards compatibility can't be maintained.

To be clear, I fully agree with you on your last statement. SanDisk's poor performance is fully their own fault and they shouldn't be blaming anyone else. All the other SSDs work fine so obviously the common denominator is SanDisk.


RE: DOA
By TomZ on 11/6/2008 6:21:20 PM , Rating: 4
quote:
However, many, if not most, of the "features" of NTFS are not necessary or even utilized by the common home use

Are you sure about that? What mechanism would be used to prevent a virus from infecting system files - you guessed it - NTFS file permissions. That's a super common home use case.


RE: DOA
By mindless1 on 11/8/08, Rating: 0
RE: DOA
By whiskerwill on 11/8/2008 12:03:46 PM , Rating: 2
What fantasy land are you living in? There are a lot of viruses that infect system files. In fact MOST viruses used to do this until Windows began implementing better system security, due in part to the security built into NTFS.


RE: DOA
By mindless1 on 11/16/2008 10:51:14 PM , Rating: 2
Most don't infect system files. Take a look at all the popular ones in the last few years, you will clearly see they don't. Some might replace files, but mostly they are their own discreet code loaded at boot-time or through a plugin, script, etc.

You wrote "until windows began... NTFS", but that was several years ago now. Even if you were correct with your initial statement (which you weren't), the second part pretty much invalidates it too if talking about anything post-9x era.


RE: DOA
By mikefarinha on 11/6/2008 6:24:08 PM , Rating: 2
TomZ has already proven this argument fallacious but it's fun banter :)

The one thing I'd point out as a flaw in your argument is that Microsoft has intentionally merged the Windows code base in order to streamline development. This goes from the basic home user to their top-of-the-line server OS. The extra effort to allow windows to sit on top of multiple file systems wouldn't be worth the cost in effort to maintain.

So using this development process, which I happen to agree with, will cause many important business features to 'trickle down' to home users. I also have a hard time believing that home users would be better off without many of these business features despite whatever performance penalties they might incur.


RE: DOA
By mindless1 on 11/8/2008 10:32:45 AM , Rating: 1
Then you'd be wrong. The vast majority of home users can't even tell you what filesystem they use, let alone know anything about it.


RE: DOA
By Clauzii on 11/7/2008 9:50:58 AM , Rating: 2
My XP machine runs fine on FAT32. I guess You mean 'Windows Vista' then?


RE: DOA
By mikefarinha on 11/7/2008 10:17:20 AM , Rating: 1
Of course I mean Windows Vista... who would want to run a 7 year old OS?


RE: DOA
By mindless1 on 11/8/2008 10:33:30 AM , Rating: 1
Anyone who isn't stupid (wants a mature patched OS instead of a beta toy).


RE: DOA
By Silver2k7 on 11/7/2008 5:15:09 AM , Rating: 2
Yes but since FAT32 can't handle larger files, atleast i tend to reformat flashdrives to ntfs.. im sure other people do the same..

Kind of backwards to have a 2008 storage device wich is limited to a 4 gb filesize.


RE: DOA
By TomZ on 11/6/2008 5:47:32 PM , Rating: 5
quote:
This new file system is DOA, at least for Windows installs. Windows relies heavily on the built-in security features of NTFS, coming up with a new fangled file system is a non-starter in the Windows world.

No, that's not correct. Sandisk is not proposing to replace NTFS and FAT with their new file system.

Instead, what Sandisk is describing as a "file system" is something implemented within and contained by their SSD. NTFS (or FAT) sits logically "on top of" the Sandisk file sytstem. Think of layers - NTFS on top and Sandisk's system beneath it.


RE: DOA
By mikefarinha on 11/6/2008 6:00:28 PM , Rating: 2
Really?

It didn't clarify that in the article with all the reference's to Windows. You're a smart guy TomZ so I'll take you at your word.

Thanks for the clarification.


RE: DOA
By Klober on 11/6/2008 6:11:19 PM , Rating: 2
Is it what you describe, or is it more of a firmware implementation (or is firmware implementation what you're describing)? Now that you've mentioned this possibility it makes me wonder how this truly works... Thanks for kicking my brain back into gear TomZ and making me think about the possible implementation a bit more. :)


RE: DOA
By TomZ on 11/6/2008 6:24:28 PM , Rating: 2
As I understand it, this new FFS would be implemented within the firmware of the SSD. This is necessary because the interface to a SSD is the same as a HDD - ATA command set over SATA or PATA. Basically the SSD "looks" the same to the computer as a HDD - tracks, sectors, etc.

I'm not any kind of expert in this stuff, but that is how I understand it to work.


RE: DOA
By BikeDude on 11/7/2008 7:36:55 AM , Rating: 2
quote:
Sandisk is not proposing to replace NTFS and FAT with their new file system.


So why all the hoopla? Flash drives have always done stuff 'under the hood'. E.g. Reordering blocks to make sure that the same blocks do not see 90% of the writes.

Basically then, Sandisk are saying that they finally sorted out the problems that their competitors sorted out a long time ago?

In that case, Sandisk are using Windows Vista as the fallguy. Very cheap.


RE: DOA
By Cerberus90 on 11/7/2008 8:00:41 AM , Rating: 2
I'm actually surprised that no one else has come up with a new system for SSDs sooner. We all know that NTFS apparently shortens the lifespan of SSDs, surely the manufacturers know this, so you'd have thought they could have come up with something better.

Maybe they don't really see SSDs as the future yet, just a gimmick, so they didn't put the money into developing a new system for something they think is a gimmick.

And if you think about it, what uses SSDs the most? Netbooks? And most netbooks are running Linux, which I'm sure, if it isn't what TomZ has said it will be (which I'm not saying it isn't, just an if), then it should be easy for Linux to support it.


Nice Try SanDisk
By isorfir on 11/6/2008 11:29:35 AM , Rating: 5
I'm not going to bite until they come out with the UltraFluxLaserMissleNinjaTurtle File System. Until then, my inner 8-14 year old won't be satisfied with a name like ExtremeFFS.




RE: Nice Try SanDisk
By FITCamaro on 11/6/2008 12:23:53 PM , Rating: 5
What about FTWFFS?


RE: Nice Try SanDisk
By incompleteunit on 11/6/2008 1:22:45 PM , Rating: 5
I'd go for WTFFFS.


RE: Nice Try SanDisk
By chmilz on 11/6/2008 4:31:08 PM , Rating: 2
OMGWTFBBQExtremeMountainDewFFS (please recycle)


RE: Nice Try SanDisk
By Schrag4 on 11/6/2008 3:05:46 PM , Rating: 2
If they (and all tech companies) wanted to be accurate, all products would be name something along the lines of 'SlowByTodaysStandardsxxxx'. Every tech is only 'fast' for a few months, or in rare cases, a couple of years, so the name would be more accurate for a great part of the product's lifespan.

I can see it now. 3 years from now I'll spend an hour on the phone with my grandmother trying to explain to her that the Extreme option is actually the slowest option available. Thanks SanDisk.


RE: Nice Try SanDisk
By mikefarinha on 11/6/2008 3:57:53 PM , Rating: 5
Are you telling me that my Super Nintendo isn't so super anymore?!?!


Extreme
By jiteo on 11/6/2008 11:41:15 AM , Rating: 4
for f*ck's sake!




RE: Extreme
By SpaceRanger on 11/6/2008 12:13:40 PM , Rating: 2
That's what I initially thought when I saw ExtremeFFS..


RE: Extreme
By PhoenixKnight on 11/6/2008 7:16:45 PM , Rating: 2
At least they realize that the word extreme doesn't start with an X.


RE: Extreme
By AnnihilatorX on 11/7/2008 10:41:05 AM , Rating: 2
FFS some kind of swear word?
To be honest XFFS looks and writes better than ExtremeFFS


Headline FIXED
By therealnickdanger on 11/6/2008 12:06:00 PM , Rating: 5
How about something more accurate:

SanDisk Hopes to Remedy Woeful Sandisk SSD Performance




RE: Headline FIXED
By SavagePotato on 11/6/2008 1:56:50 PM , Rating: 4
They could just stop using those crappy jmicron controllers instead in my opinion.


RE: Headline FIXED
By Nihility on 11/8/2008 7:04:11 PM , Rating: 1
Yea, no kidding.
So they're claiming 100x write speeds? That claim would be valid by marketing standards (what standards?) even if it applied to only 4KB random writes. Considering Anandtech showed Sandisk as having a 7000ms latency while doing multiple write ques at the same time, we're basically talking about 70ms latency which isn't even considered good by todays high quality SSDs. So big whoop Sandisk. How about you fix your shit instead of blaming an OS which your competitors seem to be running fine on? (All numbers in this post are correct according to marketing standards)


for those who dont know
By shin0bi272 on 11/7/2008 6:35:38 AM , Rating: 2
quote:
An important feature of NTFS is journaling.

Before a change is made to the metadata of a file, a transaction is logged in the $LOGFILE. These transactions list the operations required to redo or undo the changes. Once the transaction has been logged, the file system can go ahead and perform the change. Once it has completed the actual change to the data, a commit record is added to the log to show that has been successful.

With this transaction log, it is possibly for a NTFS to quickly recover from a system failure by replaying the redo or undo transactions that do not have a commit record.

Also to keep the size of the $LOGFILE to a minimum, a checkpoint record is written to the log every so often (for example every five seconds). The system will ignore any records before the checkpoint record.

Please note that the journaling system only protects file system metadata, it does not actually protect the data stored within files.


So while it is important it isnt the only solution. Even if the extremeffs is a firmware layer system and ntfs sits on top of it there can still be a better idea for meta data handling than journaling. We used to get along fine without it so im sure we can put up with losing it either all together or in place of something better.




RE: for those who dont know
By fteoath64 on 11/9/2008 1:06:52 AM , Rating: 2
Yes, the OS only cares about the LOGICAL disk. The physical disk that really implements the storage can be radically different. This isolation is important and I think the work is done mostly on the physical layout/write/read schemes. Whatever fancyful schemes that can derive true read/write speeds as Intel has demonstrated it can do. And what Jmicron has really screwed up in their ASIC controller design.

Retaining NTFS with journaling is mandatory. How that is implemented on the physical block with fixed or varying block sizes is the challenge. The SATA interface speed is the limitation. The fact that flash NAND cannot saturate SATA today shows how badly, the physical is implemented.


"A lot of people pay zero for the cellphone ... That's what it's worth." -- Apple Chief Operating Officer Timothy Cook











botimage
Copyright 2009 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki