How do i do this SSD provisioning?

[StarTech]

Can you (SetiroN) point to a link for why one would want to leave fully 10% of an SSD completely unused and unusable to the OS? It isn't as if the SSD can sense which part of the drive is formatted for C: and which isn't and do something intelligent there, so why are you doing this?

While it is not absolutely necessary in today's SSDs, the OP concept has been key in the SSD technology. The fact that you ask for links just indicates how recently you have come to the SSD world. No offense intended. You may want to read the excellent early work posted by Anand years back at the SSD infancy. As far as I know all SSDs ship with built in over provisioning. It is up to you to reserve more or not. Last one I installed, an Intel 256GB, I did add some because my use hits the drive above normal.

And not, the hard drive does not care where the empty space is.

 

[dclive]

While it is not absolutely necessary in today's SSDs, the OP concept has been key in the SSD technology. The fact that you ask for links just indicates how recently you have come to the SSD world. No offense intended. You may want to read the excellent early work posted by Anand years back at the SSD infancy. As far as I know all SSDs ship with built in over provisioning. It is up to you to reserve more or not. Last one I installed, an Intel 256GB, I did add some because my use hits the drive above normal.

And not, the hard drive does not care where the empty space is.

So you don't have a link proving a word of this. C'mon - I've done quite a bit of searching, and all I've been able to find talks about the manufacturer setting aside space (ie an 80GB SSD becomes a 64GB SSD), with better SSDs having more set aside. No one is talking about how to set aside user-level space...

 

[Cerb]

1. How does the SSD know that your OS has marked an unused partition/empty space as free? Is there some driver somewhere that no one has mentioned?

That's what TRIM is. It is a command to tell the drive that certain LBAs have been discarded.

As far as partitioning for OP is concerned, it's a feature of many drives. The OP from the factory should be sufficient for notebook/desktop/workstation loads (I'm of the strong opinion that when not, the drives should be avoided, not have OP added), but you can have a bit of an out, if you're doing DB development, or trying to use cheaper drives for DB work, or other heavy stuff.

Here's a link for that:
http://www.storagereview.com/intel_ssd_520_enterprise_review

Partitioning with MBR, with the drive in a fresh state (just secure erased), don't put any partitions on it that go to the end of the address space. Since this space never gets addressed by the OS, and thus those LBAs are never touched, those LBAs remain free for the drive to do whatever with.

The performance improvements really need little to no idle time (low-QD random writing), and/or high QDs for significant periods, to be there. FI, that's a QD=32 test, which is unrealistic for most of us--outside of software installs, or version control synchronizing on a fast LAN, we will rarely get above maybe QD=4. High-QD tests are easy to game, but may not reflect usage. Usage typically goes in fits and starts, with random write batches tending to be small (PF update here, log here, a few files added to browser cache there, etc.).

Writes in flight can be benefited by the kind of optimizations that a write cache would be used for. Given that writes will commonly complete within a few milliseconds, it's going to be pretty rare that such optimizations can be put to good use. This is one of the reasons AT performs tests with low QDs, as well as high, and why several major sites try to have tests that mimic hard usage by applications, or actually perform scripted application usage.

More logical is that the manufacturer sets aside XXX% of the drive as 'extra' and uses that for wear-leveling and other GC tasks, which is what the link I posted says....Help!

They do that, too. All have at least ~7% (the difference between 1000^3 and 1024^3 GBs), some more. If the unused space at the end trick will work, you effectively get a total spare area of:
[manufacturer OP] + [TRIM free space] + [user-defined OP]

 

[Cerb]

Not in a modern Windows....

http://malwaretips.com/Thread-How-to-Resize-Partition-in-Windows-Server-2008-and-2008-R2-SP1

Has it come down to non-server Windows versions? In Windows 7, FI, system restore and the PF get in the way, and maybe the hibernate file. I just tried, and it wouldn't let me.

 

[Yellowbeard]

Just a lil nugget to add here, AFAIK, manual overprovisioning accomplishes the same goals as factory overprovisioning with 1 exception. According to (either LSI/SandForce or LAMD/Hynix, I can't recall which) only factory set/firmware controlled OP can use the OP space for bad block replacement. Manual OP space cannot be used for bad block replacement. Otherwise, they both perform exactly the same.

 

[kmmatney]

Over provisioning works and does make a difference: you don't need to wait three years, filling up the drive can take two weeks. Don't forget that over-provisioning also improves the lifespan of your NAND by a fair amount. It doesn't make sense not to do it on a 256GB drive, just don't overdo it, 15-20GB are enough, 30 is ideal, more than that is pretty much a waste of space for normal usage.

Instead, having two partitions seems pointless to me, it's not a mechanical hard drive where the first sectors perform faster than the last ones.

Well - I don't buy this. I've had SSDs since they first came out, and have never over-provisioned beyond what is already there, and have never experienced slow downs - at least any that I notice in every day work. Unless you are constantly filling up the drive, then I don't see the point in over-provisioning. The OP already has a large mechanical drive, which I guess will be used for large media files. The built-in over-provisioning should be good enough for the life-span of the SSD.

My Samsung 830s have 7% of their capacity set aside for provisioning (hidden from view). I would think that your drive has the same, which should be fine for normal desktop use.

 

[CakeMonster]

I'd love some specific input from someone with the knowledge about whether it makes any difference if you just make sure to never fill up the drive beyond a certain limit, or if you actually set aside unpartitioned space. I have a 256GB drive (nevermind about the controller, I want general advice here) and I don't think I've ever had less than 90GB free space on it). Is there any point whatsoever in setting aside unpartitioned space in that case if I know I won't fill it up anyway? Do the controllers change behavior depending on whether the unused space is partitioned or unpartitioned?

 

[mrpiggy]

I'd love some specific input from someone with the knowledge about whether it makes any difference if you just make sure to never fill up the drive beyond a certain limit, or if you actually set aside unpartitioned space. I have a 256GB drive (nevermind about the controller, I want general advice here) and I don't think I've ever had less than 90GB free space on it). Is there any point whatsoever in setting aside unpartitioned space in that case if I know I won't fill it up anyway? Do the controllers change behavior depending on whether the unused space is partitioned or unpartitioned?

Short answer is that for "most" people, it's a non-issue. The majority of good drives have at least 7% overprovision built in. Sure if you fill the drive to 100% of formattable area and start running benchmarks, it's going to get bad numbers. In the real world where most of the data is pretty static (for desktop usage) with very little writing, and a lot of reading, it will have little "human-perceptual" impact as the 7% built-in overprovision in a "good" SSD will handle that kind of load without problems. MSWord is going to open just as fast. World of Warcraft will load just as fast. Netflix will play movies, just as fast. You can download the family pics from the USB connected camera, just as fast. Sending and receiving email will be just as fast. For most real-world normal usage, just don't worry about it. While filling the drive to near 100% formattable area is not ideal, a good drive with trim will keep the built-in 7% overprovision ready and available for this type of drive-is-full real-world normal usage, so the "human" sitting at the chair sees little impact in how they interact with the PC.

On the other hand, "benchmarks" might show your drive dog slow when at 99% full and hammering away at the built-in overprovisioning with bajillions of write/read cycles and no time for any trimming; but that isn't how desktop PC's with real human interaction normally operates. Benchmarks don't give SSD's time to recover, they just keep electronically hammering away. Real-world "desktop" (i.e. not acting as the server for other computers) usage always gives plenty of time for the drive to recover a bit and while it may not be at peak efficiency compared to an near empty SSD, it's still far faster than the human is mouse clicking at the desk jumping between player zones in WoW to kill frozen players zoning in.

For someone like you who has some knowledge, cares/maintains their PC, doesn't let it fill even close to 100%, cleans out temp files, and performs other similar basic maintenance, it's even less of an issue to "worry" about. Not saying to not care, just not something to "worry" about if you accidentally download a bigger porn movie than you were bargaining for and you fill the drive. Just watch the giant-sized HD skin flick, then delete it later. Your SSD isn't going make your PC's performance turn to mush or go belly up because you ran it a while at 99.99% full.

 

[StarTech]

So you don't have a link proving a word of this. C'mon - I've done quite a bit of searching, and all I've been able to find talks about the manufacturer setting aside space (ie an 80GB SSD becomes a 64GB SSD), with better SSDs having more set aside. No one is talking about how to set aside user-level space...

No interest in looking for link. Years back, on my second SSD, an Intel 160GB, I used a document from Intel to OP the drive for a Lenovo Thinkpad with a heavy I/O DB2 application. The document had tables with life expectancy vs. OP.

You do not have to believe it.

 

[kmmatney]

I'd love some specific input from someone with the knowledge about whether it makes any difference if you just make sure to never fill up the drive beyond a certain limit, or if you actually set aside unpartitioned space. Is there any point whatsoever in setting aside unpartitioned space in that case if I know I won't fill it up anyway? Do the controllers change behavior depending on whether the unused space is partitioned or unpartitioned?

I'd like to know this as well. I think most power users keep plenty of free disk space to handle large downloads, DVD ripping, etc.. I always keep at least a 30GB+ cushion, but it's always formatted space.

I have a 40GB Intel SSD in my daughters machine, running Windows 7, and ran into problems when the hard drive ran out of space. The problems were more to do with Windows not liking the disk being full, rather than slowdown of the actual SSD. I was going to buy a new SSD, but decided to just try compressing the entire drive. Compression gave back about 10GB, and it's been running fine ever since. ATO be hoinest I really can't tell any difference wit teh compressed drive - it's still feels "SSD-like".