Z Series SSD question and RAID/TRIM concerns

Creates a new copy of the same file name and starts over. It's on the third pass now.

 

Yay, batch files!

I remember writing the mother of all batch files using the NDOS extension back in BBS days.

 

ok. waiting eagerly for the results .. I am hooked to these forums now .. lol ..

 

I'm a bit befuddled. Ran the DummyFile script for a while, created a huge volume of unique writes.

Here's another shot of the thing working. This is on the second drive in my system. Check the Total under Write in HD Tune.


I ran some tests immediately after stopping the script


Here are some the same tests on the same drive, same config, taken earlier


Mixed results

 

So these drives that Sony are using seems to be giving slightly diferent results than the PM-800/PB-JJ2 220/200MB drives. I thought they were the same but may be that is not so. Those 4K speeds were around 18/7. These look better. Or am not correct? I'm interested in seeing the spec sheet. These drives not even in RAID seem to look pretty decent for the 4K speeds.

 

Could be.

Also keep in mind that I have been beating these poor drives up for days.

Either I'm not doing the "seasoning" correctly, or something is going on I'm unaware of, like FW-level garbage collection routines. Could also be the underpants gnomes, who knows.

 

now perhaps a similar test with these drives in RAID?

btw did you notice a significant drop in real world usage?

 

So if you are doing things correctly Zoinks, it would appear that Sony quite probably has thought about the degradation issue whilst developing these drives with Samsung, and that there probably is some cheeky under-the-hood cleaning activity going on...

If true, then that's reassuring news...

 

Wouldn't we need some other SSD to undergo the same tests for comparison purposes, to see if the Sony one is different? That way, we can tell whether the good results are because Sony pulled up its trousers, or because the tests weren't doing the right thing?

 

I'll test with what I have, which is a Sony Z591 with an OCZ Vertex 120GB.

It is actually a Turbo model, but I downgraded it to a normal Vertex in a firmware mis-hap (my bad). The Vertex and Vertex Turbo's are the same drives, just with slightly overclocked DRAM.

I'm not ready to kill off the OS yet. It's my daily driver until the other Z11 arrives (now bumped to 3/31). I will do the same tests to it eventually.

This said, here are some stats from this machine today. The Vertex is about 7 months old and has been running the 1.5 FW for several weeks (TRIM and GC).



I also have a new Vertex Turbo 120GB that I haven't used yet that came from the factory with FW1.5. Don't know if i'll torture test it or the older drive.

 

I just had an interesting coversation with SonyStyle. Granted, Sony's representatives are pretty "hit or miss" on the knowledge scale. Robert took his time and appeared to really be researching with Tech Support what the correct answer was. Here's the transcript:

Interpretations?

 

Nice grilling, counselor!

He deflected, IMO. The burning out shows that the was referring to wear leveling. Your description of the difference was dead-on. Not sure he had an answer to give you.

The guy gets brownie points for trying. Not often you get this level of research.

 

That screen capture and a couple of bucks will get you a cup of coffee.

 

Was just thinking...

The latest version of the Intel RAID driver I can find is 8.​9.​0.​1023 with a date of 7/17/2009. The iaStor.sys is from 6/4/2009 as far as I can tell.

The version of iaStor.sys on the Sony is 9.5.4.001 with a date of 11/20/2009.

Anybody know information about this RAID driver?

 

Intertering
But I don't think Trim will work at raid 0 setting.

 

Actually, I think the latest (currently in beta, IIRC) RST drivers from Intel are 9.5.7.1002 (see here), and there is no confirmation that those support TRIM.

The first post in this thread seems to indicate that 9.6.+ drivers will support TRIM. Time will tell...

(bolding mine)

One doubt is that the info about leaving unpartitioned space is not supported by real world tests given later in that thread.


-Peter

 

Option 1: If I run the drives in RAID and do experience significant degradations, I am screwed because there is no way to secure erase the SSD. However, I can use the Tony-TRIM method weekly to maintain the drives speed (although not confirmed).

Option 2: If I run the drives in JBOD, I loose some performance but the chances of performance degradation is greatly reduced because they will support TRIM.

What will your choice be?

 

Sure wish there was a definitive way to tell if TRIMs are being passed to the drive.

We are at Intel's mercy.

 

Duplicate Post

 

TRIM implementations aside,

Based on these articles, and if the FW revision is above 18C1Q, can we safely assume these drives have GC?

Mentions the FW #
http://www.anandtech.com/storage/showdoc.aspx?i=3631&p=19

More on Sammy garbage collection:
http://www.pcper.com/article.php?aid=766&type=expert&pid=11

 

Perhaps you can test the drives in RAID and see if performance takes a nosedive?

 

Being on the cautious side, how abt these options?

1)image the ssd in RAID0 as soon as you get the Z. make it JBOD. when ever you get secure erase of the SSD or intel TRIM support for the RAID0 (either of the 2 options which ever occurs first), convert the ssd back into RAID0 again.

2)keep the RAID0. Image as soon as you get the Z. Just trust the sony Garbage collection and try TONYTRIM on a regular basis.

want to hear opinions

 

I would like to see if zoinks can test the drives in RAID and check for performance degradation and then test the Tony-TRIM method to see if it works ..

 

I remember someone said Z is using software raid. Then I find a post seems to say software raid does support Trim, is it true?
http://forum.corsair.com/v3/showthread.php?t=85391

 

I'm leaning towards #2

 

I have a few more things to do in JBOD, including the space consolidation and FF overwrite (TonyTrim).

I'll do it in RAID as well.

True, but no RAID, hard or soft, supports trim right now.

 

The drives support TRIM, so I would assume they have some sort of GC available.

 

Hmm...there may be an easy way to detect if NTFS-only garbage collection is at work. You could try booting up a Linux liveCD and format the SSD with ext3 or ext4 or the like and then use the Linux equivalent of the tools you are using now to season the drive and test before and after performance. Assuming Linux versions of these tools exist, you may be able to find out whether a non-NTFS partition would suffer worse degradation.

 

Sorry to jump in on the conversation. I am thinking of buying a Vaio Z and some people in the other forum (what notebook should I buy) raised concerns about the degradation of SSD in the Z. I know nothing about the technical terms but I just wanted to know whether this is something I should be worried about or something that would change my decision of buying this laptop? I'm thinking of getting one with 192GB SSD.

 

I might be showing my lack of knowledge here, but aren't TRIM and GC exclusive?

They both accomplish the same task of clearing unused blocks, but GC is a process of finding these blocks when the system is idle (amongst other things). TRIM prevents the condition by clearing the blocks as data is deleted by the OS.

 

The FW on the drives is VBM-43S1Q

 

There might be different interpretations, but I think TRIM just allows the drive to mark the space as free so that it can be erased later by some sort of "GC" process when the drive is idle - that way no erase operation is incurred during the actual IO. When a drive only supports GC, the GC must be able to find the blocks to erase. Samsung drives did this by having knowledge of NTFS' file structure on disk, and using that to determine whether blocks are free or not.

-Peter

 

You are correct. I wasn't sure about this difference .. But because I thought TRIM is more advanced/refined than GC I assumed that if the drives have TRIM they must support GC at a lower level ..

 

ok all this talk about SSD/TRIM/GC is hurting my head now .. lol .. time to get some sleep! .. you guys should take a break from this too ..

P.S. Why can't we get SONY to answer these questions? After all they built this thing.

 

Not to call the phone support guy nuts, but as of yet I have not read one review regarding TRIM nor Garbage Collector from any major source on RAID unless I missed them. Once we do, then maybe we can settle this, but for now, RAID does not look as if it will have TRIM.

And, ya, where the heck is Sony on this important subject anyway ?

 

ZoinksS2k, thanks for the info. I appreciate it.

 

Many people here argue that based solely on their own past experience coupled with their knowledge of the current state of the art, that Sony could not possibly have gotten this right.

My preference is to take a more measured approach by actually ordering one and putting it through its paces. If I'm not convinced that there's a workable solution here, then I'll return it.

(Mine is supposed to arrive within the next 2 hours!)

 

Sooo basically, Secure Erase can only work in IDE mode???? Which is apparently not selectable on our BIOS. Do I have that right so far?

What about JBOD? Do we have a problem totally erasing back to normal performance with that set up? Or is it just RAID? Will we feel the need to secure erase to factory set up with JBOD? Or will trim keep it in shape enough? Woo this is alot to keep up with.

So I guess a regular erase (as opposed to a secure erase) just erases the data but still keep the crappy performance of your used drive as if you never erased anything? Ok ill stop talking now, im just confusing myself.

Its gonna suck if we can never ever completely erase our drives 100%. Not to be captain obvious or anything !

 

It's (unfortunately) slightly more complicated.

TRIM works in AHCI mode if running individual drives, i.e. not RAID
BUT: Sony has disabled the ability to choose AHCI mode in the BIOS

The DOS HDDerase tool requires the drive to be in IDE mode using any of the above means
BUT: Sony has disabled the ability to choose IDE mode in the BIOS.

The Linux hdparm tool has a --security-erase flag which does not require the drive to be in IDE mode
BUT: Sony freezes the security settings on the drive as part of the BIOS startup, which prevents hdparm from doing a security erase

SSD drives can also be reset from a different PC
BUT: Sony has put a non-standard connector on the drives

In other words, Sony has gone to extreme measures to prevent users from being able to use TRIM or be able to wipe the drives if they ever become slow.


What MIGHT work, but this has not yet been tested, is to do the following from linux:

hdparm --dco-identify /dev/sdXXX
(note the number of sectors)
hdparm --trim-sectors 0 (number-from-above)

 

Oops dont know why this posted again. Delete.

 

So with everything you stated, what are the reasons Sony went to extreme measures in disabling those options ? So when problems occur Sony hopes a new SSD will be bought and due to their propritery hardware, they know it's going to be an expensive replacement which enables them to make a lot of money ? Or, a user will just go ahead and buy a new computer, which perhaps is what Sony wants anyways, which means even more money to Sony ? Or, any other reasons ?

 

Usually when something is disabled like this, it's usually for one of two reasons:

1) Simplify support - the less the user can do, the less support has to worry about

2) Distinguish products within a product line. If a cheaper product can mimic a more expensive product through tweaks, sales are lost.

My guess is 1. Sony might be going the route of forcing users to replace expensive hardware, but that seems too conspiratorial to me, even for Sony.

-Peter

 

The other thing is from what people are saying SSD can degrade in speed as soon as 6 months for the smaller drives. Why would Sony make something that could break within the warrenty. Now they aren't getting your money for a new drive or a new PC. Some country warrenties are 2 years and it is for sure that most people would run into this issue within 2 years. I just can't see Sony building something that will break for everyone, especially a large portion within their warrenty period.

 

I can't speak for Sony's reasons. My personal speculation is that they cripple their machines as much as they can in order to prevent users from screwing them up and require tech support; the decision makers in the New Sony are MBAs, and not from the floor like in Old Sony, and simply aren't tech savvy enough to predict in which cases this approach might backfire. I could be wrong.

 

hot plugging works on old Z serise.. in HDPARM for intel ssd..
Advanced option menu will possibly be available once it is hacked

 

read above

 

excuted Tony-trim on my machine in RAID confguration

Before/After shots

 

It seems like the two tests show totally opposite results?

 

Could it because of the different file size? HD Tune test uses 64MByte length...cache may play an important part here.

 

Maybe I'm missing it but I only see small differences here, on the order of 10% at best.

The main issue of concern was whether there would be occasional delays on the order of tens of milli-seconds to seconds when doing small block writes, after all blocks of the disk had been 'visited' at least once.

Does it still seem that this will be a problem?
Would these benchmarks be able to show that?
Can we cross this one off or is the jury still out?