Re: Will Karmic Koala automatically have trim in effect ...
Quote:
Originally Posted by
idigchess
Which one? OCZ vertex 120gb? I was looking at that but I thought that the used performance didn't look great compared to the intel drives and we still don't know if trim will really work all that great. Is the trim support in thos drives stable? I need 100gb, so I can't really go with anything smaller.
Read the threads at OCZ forum (two SSD sections). Lots of people give benchmarks.
http://www.ocztechnologyforum.com/forum/index.php
I don't think used performance is a big problem, maybe intel used is better I dunno. Pretty sure recent anandtech article did benchmarks. TRIM and garbage collection is supposed to take care of that. You should end up with 90% new speed. Anandtech recommends 20% drive free space for optimal speed (and probably wear leveling).
With OCZ you can use TRIM or Garbage collection firmware. Since TRIM only works on win7, if running linux you'd want garbage collection firmware until linux support improves.
TRIM on OCZ is stable as far as I know. But linux doesn't support TRIM fully yet, so you'd be better off with GC firmare until ubuntu 10.04 release (unless TRIM for linux stabilizes a lot).
Demand is high, supply low for SSD, so they are still expensive (similar prices as March 2009). You pretty much have to wait for a big sale for a decent price.
Check out OCZ vertex 120gb price history at newegg
http://diskcompare.com/disk/666/ (graph on left side)
The new firmwares (TRIM) out for SSD have caused demand increase.
Intel prices are no better. Still ~$60 above suggested price from 3 months ago...
Re: Will Karmic Koala automatically have trim in effect ...
trim works on linux, use wiper.sh every day. Where is the problem?
Make a daily cronjob for wiper.sh (and comment out line 511-517 in wiper-2.5 for the "do you really want..")
as /etc/cron.daily/wiper:
Code:
#!/bin/sh
test -x /usr/local/bin/wiper.sh || exit 0
date >> /var/log/wiper.log
/usr/local/bin/wiper.sh --commit / >>/var/log/wiper.log
/usr/local/bin/wiper.sh --commit /home >>/var/log/wiper.log
w7/vista is not needed for partitioning. Use "fdisk -H32 -S32 /dev/sda" for 512k alignment or use "fdisk -u" to set the sector numbers like men, like real men!!!
used performance after wiper with Supertalent Ultradrive 128GB Firmware 1819
Code:
root@locutus:~# iozone -I -a -s 256M -r 4k -r 64k -r 512k -i 0 -i 1 -i 2
Iozone: Performance Test of File I/O
Version $Revision: 3.308 $
Compiled for 64 bit mode.
Build: linux
Contributors:William Norcott, Don Capps, Isom Crawford, Kirby Collins
Al Slater, Scott Rhine, Mike Wisner, Ken Goss
Steve Landherr, Brad Smith, Mark Kelly, Dr. Alain CYR,
Randy Dunlap, Mark Montague, Dan Million, Gavin Brebner,
Jean-Marc Zucconi, Jeff Blomberg, Benny Halevy,
Erik Habbinga, Kris Strecker, Walter Wong, Joshua Root.
Run began: Thu Oct 29 00:42:25 2009
O_DIRECT feature enabled
Auto Mode
File size set to 262144 KB
Record Size 4 KB
Record Size 64 KB
Record Size 512 KB
Command line used: iozone -I -a -s 256M -r 4k -r 64k -r 512k -i 0 -i 1 -i 2
Output is in Kbytes/sec
Time Resolution = 0.000001 seconds.
Processor cache size set to 1024 Kbytes.
Processor cache line size set to 32 bytes.
File stride size set to 17 * record size.
random random bkwd record stride
KB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
262144 4 52588 59931 49144 49201 29894 15370
262144 64 115506 122981 229070 229314 133219 129767
262144 512 145350 140268 259870 260296 216764 146239
iozone test complete.
Re: Will Karmic Koala automatically have trim in effect ...
Quote:
Originally Posted by
antiram
trim works on linux, use wiper.sh every day. Where is the problem?
Make a daily cronjob for wiper.sh (and comment out line 511-517 in wiper-2.5 for the "do you really want..")
as /etc/cron.daily/wiper:
Code:
#!/bin/sh
test -x /usr/local/bin/wiper.sh || exit 0
date >> /var/log/wiper.log
/usr/local/bin/wiper.sh --commit / >>/var/log/wiper.log
/usr/local/bin/wiper.sh --commit /home >>/var/log/wiper.log
w7/vista is not needed for partitioning. Use "fdisk -H32 -S32 /dev/sda" for 512k alignment or use "fdisk -u" to set the sector numbers like men, like real men!!!
used performance after wiper with Supertalent Ultradrive 128GB Firmware 1819
Code:
root@locutus:~# iozone -I -a -s 256M -r 4k -r 64k -r 512k -i 0 -i 1 -i 2
Iozone: Performance Test of File I/O
Version $Revision: 3.308 $
Compiled for 64 bit mode.
Build: linux
Contributors:William Norcott, Don Capps, Isom Crawford, Kirby Collins
Al Slater, Scott Rhine, Mike Wisner, Ken Goss
Steve Landherr, Brad Smith, Mark Kelly, Dr. Alain CYR,
Randy Dunlap, Mark Montague, Dan Million, Gavin Brebner,
Jean-Marc Zucconi, Jeff Blomberg, Benny Halevy,
Erik Habbinga, Kris Strecker, Walter Wong, Joshua Root.
Run began: Thu Oct 29 00:42:25 2009
O_DIRECT feature enabled
Auto Mode
File size set to 262144 KB
Record Size 4 KB
Record Size 64 KB
Record Size 512 KB
Command line used: iozone -I -a -s 256M -r 4k -r 64k -r 512k -i 0 -i 1 -i 2
Output is in Kbytes/sec
Time Resolution = 0.000001 seconds.
Processor cache size set to 1024 Kbytes.
Processor cache line size set to 32 bytes.
File stride size set to 17 * record size.
random random bkwd record stride
KB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
262144 4 52588 59931 49144 49201 29894 15370
262144 64 115506 122981 229070 229314 133219 129767
262144 512 145350 140268 259870 260296 216764 146239
iozone test complete.
Real trim would be automatic, OS based, without needing a script to manually do it. And if the tool works, it works. It doesn't matter how it's done.
Re: Will Karmic Koala automatically have trim in effect ...
@antiram
great performance indeed!
Re: Will Karmic Koala automatically have trim in effect ...
Quote:
Originally Posted by
andrewabc
Read the threads at OCZ forum (two SSD sections). Lots of people give benchmarks.
http://www.ocztechnologyforum.com/forum/index.php
I don't think used performance is a big problem, maybe intel used is better I dunno. Pretty sure recent anandtech article did benchmarks. TRIM and garbage collection is supposed to take care of that. You should end up with 90% new speed. Anandtech recommends 20% drive free space for optimal speed (and probably wear leveling).
With OCZ you can use TRIM or Garbage collection firmware. Since TRIM only works on win7, if running linux you'd want garbage collection firmware until linux support improves.
TRIM on OCZ is stable as far as I know. But linux doesn't support TRIM fully yet, so you'd be better off with GC firmare until ubuntu 10.04 release (unless TRIM for linux stabilizes a lot).
Demand is high, supply low for SSD, so they are still expensive (similar prices as March 2009). You pretty much have to wait for a big sale for a decent price.
Check out OCZ vertex 120gb price history at newegg
http://diskcompare.com/disk/666/ (graph on left side)
The new firmwares (TRIM) out for SSD have caused demand increase.
Intel prices are no better. Still ~$60 above suggested price from 3 months ago...
correct me if I'm wrong, but I didn't think that the OCZ garbage collecting firmware worked on non-NTFS partitions.
Re: Will Karmic Koala automatically have trim in effect ...
garbage collection has nothing to do with ntfs. gc is needed on every ssd with write combining = log structured (eg. indilinx, intel).
http://lwn.net/Articles/353411/
it is only the question how much garbage must be cleaned. Most garbage cleaning means heavy writing and wear out for good performance without trim. The ssd lives longer with trim.
Re: Will Karmic Koala automatically have trim in effect ...
Quote:
Originally Posted by
antiram
garbage collection has nothing to do with ntfs. gc is needed on every ssd with write combining = log structured (eg. indilinx, intel).
http://lwn.net/Articles/353411/
it is only the question how much garbage must be cleaned. Most garbage cleaning means heavy writing and wear out for good performance without trim. The ssd lives longer with trim.
Okay, I think I might not get this. I thought that garbage collecting was an indilinx-specific firmware algorithm that allowed the ssd to figure out which files had been deleted and remove them, similar to trim, when the computer was idle for 2+ hours, completely independent of the OS. I was pretty sure that Intel had nothing of the like.
Yeah, based on this article:
http://www.anandtech.com/storage/sho...px?i=3631&p=14
it appears that only samsung and indilinx have this right now. So I don't think that it's needed on every ssd with write combining. But anyway, I still am pretty sure that the indilinx one can't work on an ext4 fs.
Re: Will Karmic Koala automatically have trim in effect ...
the ntfs feature has only samsung. anand talks about a "cleaning lady"
http://www.anandtech.com/storage/sho...spx?i=3631&p=4
which is the garbage collection as basic technical necessity.
Re: Will Karmic Koala automatically have trim in effect ...
Quote:
Originally Posted by
antiram
Okay, we're talking about different things. What I was talking about was the idle garbage collection, not the cleaning lady. As here, it is not necessary because only Indilinx and samsung have it right now, but not Intel. And, at least for samsung (probably Indi) it only works on NTFS.
Quote:
Impact of Idle Garbage Collection
The other option that Indilinx provides its users to improve used performance is something called idle or background garbage collection. The idea is that without any effort on your or the OS’ part your drive, while idle, will defragment itself.
The feature was actually first introduced by Samsung for its RBB based drives, but I’ll get to the issues with Samsung’s drives momentarily.
It either works by looking at the data on the drive and organizing it into a less fragmented state, or by looking at the file system on the drive and attempting to TRIM based on what it finds. Both Indilinx and Samsung have attempted to implement this sort of idle garbage collection and it appears they do it in different ways. While the end result is the same, how they get there determines the usefulness of this feature.
In the first scenario, this is not simply TRIMing the contents of the drive, the drive doesn’t know what to TRIM; it must still keep track of all data. Instead, the drive is re-organizing its data to maximize performance.
The second scenario requires a compatible file system (allegedly NTFS for the Samsung drives) and then the data is actually TRIMed as it would be with the TRIM instruction.
Details are slim, but the idle garbage collection does work in improving performance:
PCMark Vantage HDD Score New "Used" After TRIM/Idle GC % of New Perf
Corsair P256 (Samsung MLC) 26607 18786 24317 91%
Presumably this isn’t without some impact to battery life in a notebook. Furthermore, it’s impossible to tell what impact this has on the lifespan of the drive. If a drive is simply reorganizing data on the fly into a better (higher performing) state, that’s a lot of reads and writes when you’re doing nothing at all. And unfortunately, there’s no way to switch it off.
While Indilinx is following in Samsung's footsteps with enabling idle garbage collection, I believe it's a mistake. Personally, real TRIM support (or at least the wiper tool) is the way to go and it sounds like we’ll be getting it for most if not all of these SSDs in the next couple of months. Idle garbage collection worries me.
Citations: http://www.anandtech.com/storage/sho...px?i=3631&p=14
Re: Will Karmic Koala automatically have trim in effect ...
I'm 100% certain that Garbage collection will work on non NTFS. It will work on ext4. GC has nothing to do with filesystem or operating system.
Also garbage collection starts working within minutes. You no longer have to put computer idle for several hours (although it will allow GC to work better and more thorough eg. you fill, then empty the drive several times).
What settings do I need to make garbage collection work?
There are several threads at OCZ forum about this.
GC works automatically, and on any filesystem or OS.
GC goes to 95% performance in 1 hour
So either the people working at OCZ are complete liars and no one bothered to refute their claims (there is a kernel developer there...), or it does what they say.