RAID 0 and SSDs

[Tom Sawyer]

FWIW, I have 2 120gb OCZ Vertex 2's and have set them up single drive and raid 0. Surprisingly, the Raid 0 setup with the drives boots significantly slower(!) than a single drive. As in, the 'activity circle' below the apple at boot will revolve barely 2 times with a single drive... and hits 13-16 with raid 0. This is after using the OCZ Toolbox to do a full secure wipe (which restores the drive to full factory defaults, resets cells etc). The array certainly benched higher sequential, but the boot time was strange. I ended up using 1 drive for OS X and the other for 7. Not sure where I'll land storage wise... I've thought about flashing my Areca 1220x card to EFI, but decided to just sell that and the two 120's to fund the purchase of the next gen drives just around the corner.

 

[VirtualRain]

Hmmm strange... I would be interested to know what's going on with boot times here.

 

[walshlink]

I've used striped OWC REs in my MBP...it's nice for bragging rights, but little else (and bragging rights last for a day before no one gives a *****)

There is a sucky part though; I have yet to load Windows into Bootcamp from either an external firewire or usb dvd-rom...for whatever reason Bootcamp and Windows does not like this setup.

If you stripe the drives, sequential transfers are sick, but for day-to-day ops, you won't notice a difference. One drive setup boots faster than a striped drive setup...striped drives array take time to initialize.

I am currently not striping my drives as I am using one drive for the Lion Preview (and the Lion Preview won't run on a striped array...yet.)

 

[nanofrog]

FWIW, I have 2 120gb OCZ Vertex 2's and have set them up single drive and raid 0. Surprisingly, the Raid 0 setup with the drives boots significantly slower(!) than a single drive. As in, the 'activity circle' below the apple at boot will revolve barely 2 times with a single drive... and hits 13-16 with raid 0. This is after using the OCZ Toolbox to do a full secure wipe (which restores the drive to full factory defaults, resets cells etc). The array certainly benched higher sequential, but the boot time was strange. I ended up using 1 drive for OS X and the other for 7. Not sure where I'll land storage wise... I've thought about flashing my Areca 1220x card to EFI, but decided to just sell that and the two 120's to fund the purchase of the next gen drives just around the corner.

Odd.

Granted the random access performance won't benefit from being in a stripe set, it shouldn't be that slow (I've heard of it slowing down a bit from a single disk, but not this much).

BTW, have you checked to see if there's a newer firmware revision for those drives?
If there is, you'd very likely need a PC to flash them, as flash utilities tend not to work in Macs (Intel is the only exception I can think of for SSD's, though OWC may have something as well, given their specialization in the Mac segment).

Hmmm strange... I would be interested to know what's going on with boot times here.

As would I.

I've a suspicion, but nothing more (interaction between either the firmware or actual controllers <Sandforce 1200> and the ICH10).

I've used striped OWC REs in my MBP...it's nice for bragging rights, but little else (and bragging rights last for a day before no one gives a *****)

There is a sucky part though; I have yet to load Windows into Bootcamp from either an external firewire or usb dvd-rom...for whatever reason Bootcamp and Windows does not like this setup.

Bootcamp doesn't work on arrays (requires a single disk).

As per initialization, there's no RAID card, so there's no additional firmware to load (RAID cards can add some time; say 25 seconds to 1.5 minutes, including disk spin up). Nor are the disks mechanical, which means no need to spin up (can eat time, particularly in larger arrays due to staggered spin up).

 

[VirtualRain]

Odd.

Granted the random access performance won't benefit from being in a stripe set, it shouldn't be that slow (I've heard of it slowing down a bit from a single disk, but not this much).

BTW, have you checked to see if there's a newer firmware revision for those drives?
If there is, you'd very likely need a PC to flash them, as flash utilities tend not to work in Macs (Intel is the only exception I can think of for SSD's, though OWC may have something as well, given their specialization in the Mac segment).


As would I.

I've a suspicion, but nothing more (interaction between either the firmware or actual controllers <Sandforce 1200> and the ICH10).

I saw a number of posts in the recent TRIM patch thread here that noted boot times increasing after applying the TRIM patch. There's definitely some oddities with SSD's and booting OSX. It may go beyond striped arrays. Who knows. There's too little information to go on. It would be nice if a review site spent some time on this issue since people here likely don't have the time and resources to keep rebuilding arrays in different conditions and benchmarking them. I know I certainly don't

 

[DeeEss]

Do you have them installed in the ODD bays?

 

[Tom Sawyer]

Do you have them installed in the ODD bays?

Nope, they are in bay 1 and 2. It really was strange. I did update both drives to the latest firmware using the OCZ toolbox (had to be in 32 bit windows 7), both drives were initialized at that time as well.

Nanofrog, I think you're right... something with the firmware and the controller, mixing in OS X software raid and it seems to fall down. As I mentioned, benchmarks are excellent... in the 500 mb/sec range where you would expect 2 SF-1200 drives to perform.

 

[nanofrog]

I saw a number of posts in the recent TRIM patch thread here that noted boot times increasing after applying the TRIM patch. There's definitely some oddities with SSD's and booting OSX. It may go beyond striped arrays. Who knows. There's too little information to go on. It would be nice if a review site spent some time on this issue since people here likely don't have the time and resources to keep rebuilding arrays in different conditions and benchmarking them. I know I certainly don't

I figured this wasn't the case here, given the description of a single SSD working faster than the pair in a stripe set.

But you may be on to something (if it's higher than the firmware level, it's possible driver issues are the cause; TRIM support, other changes to the SATA drivers, ...).

Nope, they are in bay 1 and 2. It really was strange. I did update both drives to the latest firmware using the OCZ toolbox (had to be in 32 bit windows 7), both drives were initialized at that time as well.

Nanofrog, I think you're right... something with the firmware and the controller, mixing in OS X software raid and it seems to fall down. As I mentioned, benchmarks are excellent... in the 500 mb/sec range where you would expect 2 SF-1200 drives to perform.

Like VirtualRain indicated, proper testing would be needed to rule things out (i.e verify that the new TRIM support isn't causing any issues, or even rolling OS X back, and see if that helps - OS updates can cause problems as well).

10.6.4 was hell with RAID, and users had to roll back to 10.6.3 to get their RAID's back to working order (hardware and software implementations).

BTW, the Windows 7 OS used to flash the SSD's - was that run on the MP or a PC?

 

[walshlink]

I wasn't striping my SSDs when I tried it. Apparantly Windows 7 does not like to load from an external DVD-ROM...

Odd.

Bootcamp doesn't work on arrays (requires a single disk).

As per initialization, there's no RAID card, so there's no additional firmware to load (RAID cards can add some time; say 25 seconds to 1.5 minutes, including disk spin up). Nor are the disks mechanical, which means no need to spin up (can eat time, particularly in larger arrays due to staggered spin up).

 

[Tom Sawyer]

I figured this wasn't the case here, given the description of a single SSD working faster than the pair in a stripe set.

But you may be on to something (if it's higher than the firmware level, it's possible driver issues are the cause; TRIM support, other changes to the SATA drivers, ...).

Like VirtualRain indicated, proper testing would be needed to rule things out (i.e verify that the new TRIM support isn't causing any issues, or even rolling OS X back, and see if that helps - OS updates can cause problems as well).

10.6.4 was hell with RAID, and users had to roll back to 10.6.3 to get their RAID's back to working order (hardware and software implementations).

BTW, the Windows 7 OS used to flash the SSD's - was that run on the MP or a PC?

That's true... I'm running latest firmware, latest OS X patches and something within either of those variables could be causing the issue.

Windows 7 used to flash the SSD's was running on the MP.

 

[nanofrog]

I wasn't striping my SSDs when I tried it. Apparantly Windows 7 does not like to load from an external DVD-ROM...

I misunderstood you then.

What model of MP do you have?

I ask, as those with EFI32 (2006/7 Mac Pros) don't handle versioned Windows disks well. There is a way to get around this however, assuming this is the problem (USB is able to boot; Windows has dropped Firewire support in Windows 7).

That's true... I'm running latest firmware, latest OS X patches and something within either of those variables could be causing the issue.

Windows 7 used to flash the SSD's was running on the MP.

Are you sure that they flashed properly?

If possible, you could try re-performing it on a PC, and see if that changes matters (Macs tend to have problems passing low level information such as surface scans of drives, and flashing tools for firmware).

 

[ecstacy2012]

Two 240gb owc mercury pro in raid 0

I have two 240gb Mercury extreme Pros (non re) in bay 1 & 2
with raid 0 configured stripping. It's as my main boot. As far as loading differences I really can't notice any difference. Read/Write values are not double but average was close it was around 133mbs difference which is pretty damn close. I'm on 10.6.7 with macpro 2010 quad incase anyone is wondering. I'm finding them very reliable

 

[Tom Sawyer]

I misunderstood you then.

What model of MP do you have?

I ask, as those with EFI32 (2006/7 Mac Pros) don't handle versioned Windows disks well. There is a way to get around this however, assuming this is the problem (USB is able to boot; Windows has dropped Firewire support in Windows 7).


Are you sure that they flashed properly?

If possible, you could try re-performing it on a PC, and see if that changes matters (Macs tend to have problems passing low level information such as surface scans of drives, and flashing tools for firmware).

I have access to plenty of PC's at the office so it's possible. Unless the firmware flash only incremented the version # and didn't actually update the code, it seems to have flashed fine as the version number did update in Profiler. I would think the OCZ toolbox would have honked if it didn't successfully flash. The 'secure erase' however is another story, it said it performed the task but who knows really.

If I'm able to get some spare time together and flash and secure erase these on a PC I'll do that and post back if there is any change.

 

[nanofrog]

I have access to plenty of PC's at the office so it's possible. Unless the firmware flash only incremented the version # and didn't actually update the code, it seems to have flashed fine as the version number did update in Profiler. I would think the OCZ toolbox would have honked if it didn't successfully flash. The 'secure erase' however is another story, it said it performed the task but who knows really.

Given the issues with passing low level information (flash and HDD surface scans are major examples), it's possible that the version number is all that was changed.

So giving it a run on a PC isn't a bad idea IMO, assuming the OCZ Toolbox does more than just scan the version number (most flash tools are old, and were written well before EFI was created). Then there's Apple's proprietary implementation to deal with... (it's not 100% compliant with the EFI 1.10 standard from issues posted about the Intel based systems).

If I'm able to get some spare time together and flash and secure erase these on a PC I'll do that and post back if there is any change.

Good luck.