Formatting a SSD and need advice for RAID 0 SSD

[Old Hippie]

You mean SSD right? Could you explain in more detail what you did?

I've seen this type of behaviour before.

Seems some MBs can't be fooled into unlocking the area of the HD that needs to be accessed for Secure Erase to do it's job.

Bandit1's post shows how he got around it and I have seen another method which envolved connecting the drive during POST.

I realize that creating the spare area after formatting isn't the best but this is why I'd like to see a little experimentation/documentation between the ways the spare area is created and the differences in performance.

HDDErase just doesn't seem to cooperate with everybody.

PS....I'm gonna add a little info for those who care...

Q: How can I bypass security freeze lock?

A: Three different ways can bypass a BIOS security freeze lock:
1. Most preferred method: If another computer is available, boot the drive
from another computer. Since the freeze lock is entirely BIOS dependent, another
computer's BIOS may not freeze lock the drive.

2. Second method: Switch the drive to another drive channel or another position
on the channel, e.g. Switch drive from secondary master S0 to secondary slave S1
or vice versa. Some BIOSs do not send the Freeze Lock command to all channel
master/slave positions.

3. Least preferred method: **There exists some danger to your drive in using
this method, use at your own risk** Shut down the computer system. Unplug the
four-wire power cable of the hard drive while leaving the signal cable plugged in.
To eliminate the danger of ESD, always ground yourself when removing the power
cord. Power on the system and boot into DOS with a DOS boot disk. Once DOS has
booted up and you are at a command line interface plug the power cord of the hard
drive back in. Run HDDerase.exe. The logic in this method is to prevent the
drives detection in BIOS, which is when the freeze lock command is issued.

Read the Secure Erase Q&A from here for more info.

 

[Old Hippie]

So, although Tweaktown appears to say it was possible, they were either caught up in the confusion, or they miss-wrote. I haven't seen recent news that it's now possible, but that can change at any time...

It was confusing enough that Intel changed the description to mention the fact that only drives not included in the array would be avaliable for TRIM.

It was definately confusing for awhile.

 

[FishAk]

If you have an uncooperative motherboard, couldn't the SSD be placed in another computer and be secure erased there? Correct me if I'm wrong, but isn't HDDerase on bootable media? Most people have access to multiple computers these days, so whats stopping you from doing the operation in a netbook, a laptop, or most any computer lying around with the right connections?

 

[sub.mesa]

Yes, no reason why you have to do it on your main computer.

 

[Old Hippie]

If you have an uncooperative motherboard, couldn't the SSD be placed in another computer and be secure erased there?

1. Most preferred method: If another computer is available, boot the drive
from another computer.

Looks like it to me.

 

[MerrickAggie97]

You mean SSD right? Could you explain in more detail what you did?

For my original set up I had a single SSD running on my SATA2 controller.

I tried as much as I could to perform a secure erase. Nothing would work. The program just wouldn't cooperate no matter what I did and nobody told me of some other programs to use aside from secure erase. Of course, I didn't really ask. I placed the SSD on different ports and everything, set up the motherboard to IDE mode and nothing would allow HHDerase to work.

After spending several hours (ya know I do need to get on with my life!) I finally gave up and just inserted the Windows 7 boot disc and removed all partitions and just formatted from there.
I then proceeded to put in my 2nd SSD (same brand, same model) and went into the Intel Controller and then setup RAID 0. Giving me about 477GB total space. When I set Windows 7. I partitioned about 380GB (just more than 80% of total space available) and left the remaining space unpartitioned. I installed Windows 7 and everything works well.

 

[MerrickAggie97]

If you have an uncooperative motherboard, couldn't the SSD be placed in another computer and be secure erased there? Correct me if I'm wrong, but isn't HDDerase on bootable media? Most people have access to multiple computers these days, so whats stopping you from doing the operation in a netbook, a laptop, or most any computer lying around with the right connections?

I don't have another computer around. Thats what!

 

[MerrickAggie97]

For right now, unless someone can offer a compelling reason otherwise (and please do both SSDs together cost me over $1000 wand obviously I want to get full benefit of them), I'm just going to consider this "done" and be happy with my current setup. Im not sure if using windows 7 format is the best way to format an SSD (I have been told it should be good enough) but for now it was really my only option.

I would ask if there is something I can do with the unallocated space. Could I perform a secure erase on that area in Windows 7?

 

[Old Hippie]

(ya know I do need to get on with my life!)

Heresy!

MerrickAggie97, you may given us the opportunity for a little long term comparision.

I have 2 Intel G2 80GB in RAID0 on an Asus P6T running w7. I used HDDErase to make a 30GB unallocated partition. These were installed on June 24th and this is my AS-SSD bench

If your setup to close to mine, over time we can track the differences if any, between these two ways of creating a free space.

After you get your system setup the way you want it, take a baseline AS-SSD test and we'll go from there.

 

[Old Hippie]

Could I perform a secure erase on that area in Windows 7?

Sorry, it's all or nothing.

 

[aigomorla]

according to tweaktown article he posted, the drivers DO pass TRIM to raid1 and raid0

NO

Intel fubarded the wording.

The Driver will pass trim on a SSD Thats NOT in a RAID ARRAY when you have your ICH10R set to RAID.

B4 if you had a raid array connected to your SSD you would not be able to pass trim.

The driver allows you to pass trim to the single SSD if you have a raid array attached to the other ports.

Intel reposted this:
http://www.intel.com/support/chipsets/imsm/sb/CS-031491.htm

"Intel® Rapid Storage Technology 9.6 supports TRIM in AHCI mode and in RAID mode for drives that are not part of a RAID volume.

A defect was filed to correct the information in the Help file that states that TRIM is supported on RAID volumes."

I have 2 Intel G2 80GB in RAID0 on an Asus P6T running w7.

.

3 x X25-V on Raid 0

2 x X25-V on Raid 0

I forget who told me its best to put 3 in raid... but im starting to believe the guy.

 

[Old Hippie]

I forget who told me its best to put 3 in raid... but im starting to believe the guy.

The more the marrier!

I'm sure I'd be faster if I connected one or two of my G1s but at these speeds it's pretty much of a moot point.

 

[aigomorla]

The more the marrier!

I'm sure I'd be faster if I connected one or two of my G1s but at these speeds it's pretty much of a moot point.

im sure as well since the 80gigs have 2x more channels.

Im just curious why my 4k's saw very little improvement.

 

[sub.mesa]

The 4K Read with single queue depth would NOT rise with use of RAID; this counts for both SSDs and HDDs. The 4K write and 4K-32/64 scores however, should see improvement.

Your 4K-32 read scores should be able to go higher. But keep in mind these speeds require a fast CPU too, particularly, a CPU with high clock frequency and IPC; since only one core would be used in the largely single-threaded Windows I/O backend. That could mean on a quadcore 25% cpu usage bottlenecks your SSD already, which could drag the 32/64-queue scores down.

So perhaps you should leave your taskmanager's CPU window open when doing the benchmarks and see how your CPU usage goes. A fast dual core would be better than a slow quadcore for this task, at least on Windows.

 

[Old Hippie]

Im just curious why my 4k's saw very little improvement.

I guess two 25-Vs are not the same as one X-25 80GB unit.

 

[Old Hippie]

I'm lookin' for MerrickAggie97's response.

I'm thinkin' this would be very interesting to compare our situations.

 

[sub.mesa]

By the way, i have four Intel X25-V myself and used a fifth for testing; when doing random reads with multiple queue depth, i got a whopping 1234MB/s of random reads [4K - 128K mixed]. Exceptional performance, but honestly the SSDs were still virgins at that point, with just a little data written to them. But at least the controller is capable of exceptional performance, and you get the same controller on X25-V as X25-M.

Interestingly, since X25-V has 5 channels as opposed to X25-M's 10 channels, it should get lower random reads as well. But it appears to reach 250MB/s which is about the same as X25-M in pristine condition. This leads me to believe the interface is holding Intel's SSDs back, which have superior read speeds. Therefore i think that around Christmas when the third generation of Intel SSDs get introduced with 6Gbps SATA; it would have little trouble filling that bandwidth with sequential and even random reads. But the writes would likely still lag behind. The new X25-E based on MLC memory instead of SLC would be interesting as well, as it would be using a different production technique increasing the write endurance.

Can't wait to try one at home.

But honestly, SSDs were already such a big step forward that differences between SSDs might not yield such a huge effect, depending on the workload ofcourse. It likely means the CPU is the usual bottleneck and never/unlikely the SSD. That didn't stop me from getting excited about the 1234MB/s random reads, though.

 

[MerrickAggie97]

I'm lookin' for MerrickAggie97's response.

I'm thinkin' this would be very interesting to compare our situations.

Im not really sure what you are asking of me.

 

[aigomorla]

Your 4K-32 read scores should be able to go higher. But keep in mind these speeds require a fast CPU too, particularly, a CPU with high clock frequency and IPC; since only one core would be used in the largely single-threaded Windows I/O backend. That could mean on a quadcore 25% cpu usage bottlenecks your SSD already, which could drag the 32/64-queue scores down.

i highly doubt im bottlenecked at the cpu speed.
If anything my windows 7 score says my HD4870X2 in Quadfire is the big bottleneck.

LOL...

 

[CKP]

Hi all

Question: When setting up new RAID array after erasing discs, I then have to choose how big I want it to be.. do I use all the available space, and then just not use it all when installing Win7 (i.e partion 160GB but only let windows partition/format 120GB and then leave rest as 40GB unpartitioned space-
or:
Should I only make array-size 120GB and then let windows use it all?

Right now I have done the last thing, so when looking in Disc-management it only shows win7 100mb partition and C:/ 120GB - The part I left out when making Array is not to be seen/available.

I´m bit confused on this part...

Regards
CKP

 

[vidsondak]

Hi Sub mesa,
Quaoting from your statement:
As far as i understand, only the Windows people do not get TRIM under RAID. New versions of both Linux and FreeBSD already allow SSDs to be RAIDed and still have TRIM capability, and that's not that hard or complicated at all.


Im new with linux and never making raid before. i bought 2 ssd already, now i want to do raid 0 on them on linux for my video storage.
Is it posible to have this raid 0 content be share with my windows 7 as well (install with the same pc)

Please teach me how to do properly.
Im on RHEL 5 update 5 - installed on a sata just for software/system.
my mobo x25a ud9. (it gets its own onboard raid controller - not sure if it matter)

cheers

 

[taltamir]

Hi Sub mesa,
Quaoting from your statement:
As far as i understand, only the Windows people do not get TRIM under RAID. New versions of both Linux and FreeBSD already allow SSDs to be RAIDed and still have TRIM capability, and that's not that hard or complicated at all.

um... this has absolutely NOTHING to do with the OS itself, and everything to do with your RAID controller firmware and driver. Unless you are doing pure OS RAID that is (which is heavily limited in windows).
AFAIK there are no drivers that can pass TRIM to raid members at the moment.
What you do have are drivers from intel which lets you pass TRIM to drives which are on a controller that is raiding other drives, but not that specific SSD.

eg. if you have 2 500GB spindle disks in RAID1 and an SSD not in raid all plugged to the same controller, the SSD will get TRIM... with the intel drivers (for windows). Any other controller, merely enabling raid mode will disable trim for all drives on the controller, even drives outside of arrays.

As for pure OS raid... I haven't heard of linux trimming drives in such a manner, although that would certainly be nice if they did. Anyone else knows more about it? link? anything?

 

[sub.mesa]

Linux has TRIM on Ext4 in recent kernels. Doesn't matter if you RAID them, as long as it is software RAID the TRIM will still work.

FreeBSD has TRIM on UFS only in FreeBSD 9 (development). But FreeBSD8 can send TRIM commands already, which works across software RAID arrays. The GEOM I/O framework is one of the most advanced I/O frameworks i've seen.

The RevoDrive with Silicon Image fakeRAID for example does not have TRIM under Windows and likely never will have, but under FreeBSD you get out-of-the-box TRIM (almost). You will see two disks, since the Revodrive has two RAID members. Since you can access the Sandforce controller directly, you also have TRIM capability.

For example, this command would secure erase your SSD:
newfs -E /dev/...

If you've got multiple SSDs in RAID configuration, and you enter the RAID device name, then all SSDs will be secure erased.

So TRIM on RAID already works, and its quite kewl. ;-)

 

[vidsondak]

So, partitioning with LVM to make LVM volume, then create XFS filesystem on it...
will still retain the TRIM function?

 

[LokutusofBorg]

I can't believe all the misinformation on the first couple pages of this thread. I hope you all have learned better by now.

AFAIK Crucial drives have no controller-level garbage collection so if you put them in a RAID you lose any and all deleted block reclamation. As soon as you've written the drive's capacity once you should hit a brick wall in performance unless you run a tool to manually reclaim space. Somebody correct me if I'm wrong there.

Intel and SF are the only controllers that are able to use partitioned-but-empty space the same as the over-provisioned space. All other controllers cannot, they are only able to play with the over-provisioned amount for wear-leveling.