View Single Post
Old 03-07-2012, 04:33 PM   #21
ilovejedd
hopeless n00b
ilovejedd ought to be getting tired of karma fortunes by now.ilovejedd ought to be getting tired of karma fortunes by now.ilovejedd ought to be getting tired of karma fortunes by now.ilovejedd ought to be getting tired of karma fortunes by now.ilovejedd ought to be getting tired of karma fortunes by now.ilovejedd ought to be getting tired of karma fortunes by now.ilovejedd ought to be getting tired of karma fortunes by now.ilovejedd ought to be getting tired of karma fortunes by now.ilovejedd ought to be getting tired of karma fortunes by now.ilovejedd ought to be getting tired of karma fortunes by now.ilovejedd ought to be getting tired of karma fortunes by now.
 
ilovejedd's Avatar
 
Posts: 5,110
Karma: 19597086
Join Date: Jan 2009
Location: in the middle of nowhere
Device: PW4, PW3, Libra H2O, iPad 10.5, iPad 11, iPad 12.9
Quote:
Originally Posted by kacir View Post
I have two observations:
1. You would need to install special program that would trap and record every single write to some log. I suspect that programs write to the disk more often than you think. Once I was tracing complicated problem with a Visual Foxpro application that was opening some tables. So I installed a special program that has recorded every single access to the disk to a log file. I was *astonished* how many files how many files get accessed before a complex application, such as Foxpro program even starts thinking about actually opening tables with data.
2. What you write is correct, in terms of math, but you would need to have a perfect wear-leveling mechanism, and by that time you would exhaust the last writeable byte ;-)
You can easily view the amount of Host Writes on Intel SSDs (wear-leveling included) via CrystalDiskMark or Intel SSD Toolbox. No need for a special program to log it for you. As for other brands, you don't really need to monitor every little disk access. You just need to see the writes. For SandForce drives, write amplification can be even lower depending on how compressible the data is.

I don't have access to my main OS SSD right now but one of the SSDs I use for my virtual disk images is currently at 545GB host writes at 1151 hours of operation. That equates to just 11.4GB/day. That drive contains 3 Windows XP virtual machines (all 3 with 512MB or 1GB pagefiles) which get defragged and compacted every 2 months or so (lotsa, lotsa writes).

Yes, there's a lot of disk access but I reckon most of those are just small 4KB or so writes. Besides, it's all those small disk accesses which make moving to an SSD such an upgrade.

Here's a thread from folks trying to kill their SSDs:
http://www.xtremesystems.org/forums/...e-25nm-Vs-34nm

Several have succeeded in rendering their SSDs unusable (at upwards of 200 TiB). Surprisingly, the 2nd gen 34nm 40GB Intel (Kingston SSDNow 40GB) is still trudging along (albeit with mapped sectors).

Sure, trying to minimize the amount of data on the SSD makes sense if you don't have a lot of space on it (e.g. you only have a 40GB SSD - barely enough for Windows 7 after you factor in OS rot). However, moving everything off the SSD to "prolong" its life? Kind of a non-issue since the SSD is likely to die first due to something else (controllers seem to be particularly delicate, moreso than the NAND anyway).
ilovejedd is offline   Reply With Quote