SSDs - State of the Product?

[Santilli]

The rejected article about the slow down points out that how windows deals with data can be part of the solution.

In particular, using the Windows utility that writes everything to cache, sequentially, then deletes it all on shutdown. So, as the drives become more common place, perhaps MSFT can figure out a way to write data that will limit, or get rid of any possible slowdown.

It would be nice to know if the Mac OS writes data in a different way then XP or Vista.
What about LinuX?

 

[Bozo]

I believe I saw an article on Slashdot that told how to tune Linux to work better with SSD drives.
Aparently Windows and Linux both need some driver tuning to use SSDs.

 

[LunarMist]

Probably by Windows 7, SP2...

 

[blakerwry]

Something that seems obvious to me... but perhaps not a layman, is that HDD's have a cache and firmware that allow it to behave optimally for a range of applications/operating systems/file systems.

Why the complete reversal in position? Why should operating systems now have to tune themselves to work with the limitations of common SSDs instead of making the SSDs more application neutral?

It seems to me that all of the performance and limitations of SSD technology could be mitigated by placing a few MB of cache and some intelligence at the head end of a write pipeline.

This could, first of all, speed up slower write sequential speeds. And second, and most importantly, would allow the write combining (sequential writing) of small writes/updates that seem to be the bane of SSD speed and longevity.

I would think the cost of this would be marginal as all disk manufacturers already have the technology and software integrated into existing magnetic based hard drives. The cost of providing a small amount of memory and perhaps a battery to allow cached writes to be written in case of power failure would likely be minimal (a few dollars) compared to the price of the device as a whole.

 

[Mercutio]

I would think the cost of this would be marginal as all disk manufacturers already have the technology and software integrated into existing magnetic based hard drives. The cost of providing a small amount of memory and perhaps a battery to allow

The companies that are presently offering successful SSDs generally aren't magnetic storage companies. Samsung might be an exception, but it's so huge that it's anyone's guess how much one group communicates with any of the others.

 

[timwhit]

I decided to install XP with SteadyState to see if it would have a large enough performance benefit to downgrade from 2003 Server.

This took several days, because I found out that XP doesn't support GPT partitions. So, after installing XP I had to restore my old 2003 installation, copy all the data off the GPT disks and reformat them. Then I restored the XP install I had done.

Anyways, SteadyState does seem to make a pretty good difference in speed. I haven't noticed any slowdowns at all since I installed it. I opened up 20 Firefox tabs in a row all heavy in images and didn't notice any stuttering.

There are some downsides to this software. For example, if the system crashes or BSODs you will most likely lose any data that was saved on your boot partition since the last reboot. I don't store any data on my boot partition, so this isn't a huge deal for me. I'll see how it works over the next couple days. I can always restore my 2003 install if it doesn't work as planned. Though, my Patriot 32GB SSD is becoming unbearable to me with the stuttering. I guess none of this would be necessary if I would just drop the money for one of the Intel disks. I just can't rationalize $350+ for a disk right now.

If anyone else has a chance to try this, let me know how it goes. DD, I'm looking your way.

 

[ddrueding]

I'll try to set up some systems today if I have time. Is there a good guide to get Steady State going?

 

[timwhit]

I'll try to set up some systems today if I have time. Is there a good guide to get Steady State going?

This is what I followed. I did the "fast" way, as I don't store my data on the boot drive.

One thing I haven't been able to figure out is whether SteadyState is caching writes to the other drives in my system or only writes to the boot drive. I wouldn't think it would do that, but I don't know how to tell.

 

[LunarMist]

SteadyState doesn't work in my computer. The file system is not compatible.

 

[timwhit]

SteadyState doesn't work in my computer. The file system is not compatible.

What OS and filesystem are you running?

 

[LunarMist]

XP SP3. I think it wants the boot partition to be NTFS rather than FAT32.

 

[timwhit]

That's correct, SteadyState requires NTFS.

 

[LunarMist]

I tried the conversion, but many programs no longer worked, so I reverted to a previous image. It is not worth the hassle, since X25-E is working fine.

 

[timwhit]

Anyone had a chance to try out SteadyState? It's still working great for me. This SSD is now performing like it should. I'm quite happy with the outcome.

 

[ddrueding]

I'm building a pair of identical systems at the office, I should be able to do some benchmarking tomorrow.

 

[Stereodude]

So, is the 60GB OCZ SSD for $102 after rebate from Amazon going to drive me nuts if I put it in my HTPC that's currently running a 40GB 4200RPM Toshiba IDE 2.5" drive, or am I unlikely to notice any stuttering? I'm not looking for a speed boost as much as I am in removing one more noise source from the system.

 

[timwhit]

So, is the 60GB OCZ SSD for $102 after rebate from Amazon going to drive me nuts if I put it in my HTPC that's currently running a 40GB 4200RPM Toshiba IDE 2.5" drive, or am I unlikely to notice any stuttering? I'm not looking for a speed boost as much as I am in removing one more noise source from the system.

You will probably notice some stuttering. If you install SteadyState, you will notice no stuttering.

 

[Stereodude]

You will probably notice some stuttering. If you install SteadyState, you will notice no stuttering.

I'm running XP Pro on my HTPC. I'm not interested to mess with anything other than cloning the install to the new drive. I use the HTPC basically only for media playback (DVD ISO's over the network, .mkv files over the network, and Blu-Ray and HD-DVD from the optical drive in the system) so I figure it shouldn't be tagging the HD very much.

 

[timwhit]

I'm running XP Pro on my HTPC. I'm not interested to mess with anything other than cloning the install to the new drive. I use the HTPC basically only for media playback (DVD ISO's over the network, .mkv files over the network, and Blu-Ray and HD-DVD from the optical drive in the system) so I figure it shouldn't be tagging the HD very much.

Just so you know SteadyState runs fine on XP and installs in about 10 minutes including the reboot.

 

[ddrueding]

I'm building a pair of identical systems at the office, I should be able to do some benchmarking tomorrow.

I still haven't gotten to this, but they are waiting on my workbench.

 

[Stereodude]

Are SLC drives immune to the stuttering, or is that a function of the Jmicron controller?

My HTPC has a very small install of XP on it, and I don't need a large drive. I didn't realize the one I linked to at Amazon was an older OCZ model, so I will be passing on it.

 

[timwhit]

I believe that most or all of the SLC drives are immune to the stuttering. I think DD would have a better idea of this.

 

[ddrueding]

All my SLC drives don't have the stuttering issue.

 

[Stereodude]

All my SLC drives don't have the stuttering issue.

But do you have any SLC drives that aren't Intel?

 

[Handruin]

Actually he does if his system link is up to date. He has three of the MTRON MSD-6000 SLC.

 

[ddrueding]

My system link is not entirely up to date. But I still have those MTRON SSDs sitting here. They do not have the stuttering issue. I also have an MTRON MSD-PATAA3018-032-ZIF2 that I haven't tried yet.

 

[LunarMist]

Blech. One should not need SteadyState or other kludges. The SSD needs refinement.

 

[timwhit]

Blech. One should not need SteadyState or other kludges. The SSD needs refinement.

Maybe not, but the kludge turns a sub $100 drive into an insanely fast piece of hardware.

 

[Mercutio]

I don't think I'd mind Steadystate on my windows machines, if it worked on Windows Server.

 

[Stereodude]

Has anyone tried the whole partition alignment thing talked about here?

 

[Handruin]

That has some good info. I didn't see results of before/after when using a typical hard drive. I wonder if there are performance gains under XP with an aligned partition? Given some of the documentation suggested increased performance with SQL databases, it might make sense.

 

[Stereodude]

I guess there could be some advantage even for a spinning disk. Maybe I'll try it with my new 500GB 2.5" drive.

 

[Handruin]

I'll see if I can give it a try in a VM doing a before and after comparison on WinXP. I know it's not exactly the same as bare metal, but if could show some differences. I don't have any spare drives floating around to try this with.

 

[Stereodude]

Well, I ordered a 64GB PATA SSD to play with in either my HTPC, or my older Dell laptop, or maybe I'll try it in both.

 

[Stereodude]

And, not content to stop there, I just ordered a 30GB OCZ Vertex Series SSD.

I'm out of control!

 

[P5-133XL]

6 TB 24 drive SSD raid

 

[Stereodude]

Does RAID-1 make any sense for a SSD? Obviously a SSD can fail, but I would think that more likely it would be the cycles on the drive that would cause it to fail. In which case RAID-1 doesn't buy you much of anything against that type of failure.

 

[Fushigi]

Depends on the intelligence of your controller. Smart controllers will spread reads across the mirrors for added performance; dumb controllers use one drive for everything and merely mirror writes.

 

[Stereodude]

Smart controllers will spread reads across the mirrors for added performance; dumb controllers use one drive for everything and merely mirror writes.

I don't think there are any RAID controllers that actually spread the reads across the drives for improved performance.

 

[Fushigi]

IBM white paper from 2001
Check the RAID-1 section: "The read performance of RAID 1 is superior to RAID 0 or RAID 5, because read operations are distributed over a pair of drives instead of being concentrated on a single drive."