| Author |
Message |
LP-SolidRaven
Evil Belgian Waffle
Joined: 06 Jun 2004
Posts: 8144
Location: Belgium
|
 Posted: Mon Apr 27, 2009 3:07 pm Post subject: |
 |
|
The result would greatly depend on the NTFS driver that's being used though...
_________________
Dilly dally, shilly shally. |
|
| Back to top |
|
| |
ClickFanatic
Est. 2005
Joined: 18 Jan 2005
Posts: 4685
Location: 37°45'18.24"N 14°59'42.9"E
|
| Posted: Mon Apr 27, 2009 5:51 pm Post subject: |
|
|
HFS is from Apple, NTFS is from Microsoft. The tests concern filesystems on Linux. Just pointing out that the benchmark (which is supposed to be just an example of one anyway) is not relevant to your discussion.
Of course the driver that is used greatly influences the outcome, because that's what ultimately manages the filesystem.
| Xtreme $niper wrote: | | Well yeah I understand that, but in the end if the discussion is going to be about the file system and which operating system it is running under, then those factors need to be considered anyway. |
Then I must have interpreted your previous post wrong, because I thought you were talking about subjectivity when you said "experiences" (like prefering one filesystem because you have been using it for ten years without a problem).
Of course many things influence performance of a filesystem and even though they can be kept a constant factor in benchmarking, different people have different setups. However, a benchmark can reveal a filesystem's strong points and weak spots, which allows users who are concerned with such decisions to make a better choice for the situation in question.
_________________
If you can read this, my post is on an alternating background. |
|
| Back to top |
|
| |
Xtreme $niper
Lifeless Person
Joined: 29 Mar 2005
Posts: 1870
Location: Canada
|
| Posted: Tue Apr 28, 2009 10:11 am Post subject: |
|
|
Well HFS is from Apple, sure, but since OS X is technically UNIX based, it was included in the benchmarks, sooo I didn't think it was too far off from our discussion since it was included in the testing.
_________________
Come visit Shattered Abstracts! (Photoblog!) |
|
| Back to top |
|
| |
ClickFanatic
Est. 2005
Joined: 18 Jan 2005
Posts: 4685
Location: 37°45'18.24"N 14°59'42.9"E
|
| Posted: Tue Apr 28, 2009 12:43 pm Post subject: |
|
|
The version they tested was probably the ported version for Linux, which doesn't support journaling in read/write mode by the way. They could have tested the ported NTFS as well, but it wouldn't be very representative.
In my opinion it doesn't really make sense to use a filesystem which has no public specification or implementation on a system where it is not supported, because you'll have to rely on reverse engineered drivers. Unless you really need the compatibility (eg. in a dual-boot environment).
_________________
If you can read this, my post is on an alternating background. |
|
| Back to top |
|
| |
linuxdoctor
Infallible Persona
Joined: 23 Apr 2005
Posts: 1530
Location: Ottawa, Canada
|
| Posted: Tue Apr 28, 2009 8:49 pm Post subject: |
|
|
Continuing on the topic of file systems and why so many. A choice of file system is also determined by the nature of the hardware that media for that device is designed for.
Take, for example, the tape drive. This is a sequential access device which requires that the tape be positioned over the read or write head of the device before it can be read from or written to, Any file system designed for a tape needs to take that into account. One would be foolish to apply a file system designed for (say) a disk drive to an old fashion 9-track tape.
Actually, for fun we did that back when I was in university. We placed the standard Unix file system onto a 9-track tape and then started reading files. It was certainly fun to watch but very frustrating trying to get the data off of it.
A Compact Disc device was originally a very slow device designed for sequentially accessing music. It did not need to be fast because audio range sounds are relatively slow signals and they didn't have to be as large as LPs to provide greater playing times. Old 12-inch LPs can play for maybe 50 minutes or an hour whereas the 12-centimetre CD can play for 70 or 80 minutes.
Because the data speed from the device was relatively slow, a high performance random access file system would not be an efficient way of storing and retrieving information from the disc. The ISO 9660 file system used on CDs and DVDs was designed for the relatively slow speed of these devices. Originally, CD players (and their CD-ROM counterparts) were slower than floppy discs biut modern devices are significantly faster permitting higher performance file systems on the media. I am, however, unaware of any tests done in this area comparing different file systems on Compact Disc media.
It might be interesting to try.
_________________
Misanthrope: someone who realizes that humans really are as stupid as they appear.
If you think I'm 'politically' incorrect you have the wrong politics.
Big business is a disease we will need to cure before we will ever achieve real prosperity. |
|
| Back to top |
|
| |
LP-SolidRaven
Evil Belgian Waffle
Joined: 06 Jun 2004
Posts: 8144
Location: Belgium
|
| Posted: Wed Apr 29, 2009 12:23 pm Post subject: |
|
|
The issue would be that once again the drivers would have a serious impact on the performance. You'd have to do the test on several operating systems.
_________________
Dilly dally, shilly shally. |
|
| Back to top |
|
| |
ClickFanatic
Est. 2005
Joined: 18 Jan 2005
Posts: 4685
Location: 37°45'18.24"N 14°59'42.9"E
|
| Posted: Wed Apr 29, 2009 12:45 pm Post subject: |
|
|
The mechanical aspect of reading CDs is very similar to that of a hard disk, right?
Also, a lot of factors that are relevant for hard disk file systems do not really apply on a CD. Unless we take into account rewritable CDs, we are mainly interested in fast reads. Linking, unlinking and write performance may be sacrificed for this purpose.
_________________
If you can read this, my post is on an alternating background. |
|
| Back to top |
|
| |
linuxdoctor
Infallible Persona
Joined: 23 Apr 2005
Posts: 1530
Location: Ottawa, Canada
|
| Posted: Wed Apr 29, 2009 11:14 pm Post subject: |
|
|
| LP-SolidRaven wrote: | | The issue would be that once again the drivers would have a serious impact on the performance. |
Only if the drivers are so inefficient that it ends up being slower than the device. CD device speed for a 1x speed device is about 153KB/s. Even for a 52x drive that is under 8MB/s. Very slow compared to hard discs Modern hard disks can sustain 125MB/s transfer rates or better. As long as the driver is faster than the device all is well. Another factor is the seek time, the time it takes the device to arrive at a given block on the device and become ready to transfer the data. Hard disks have very small platters compared to CDs. The read/write read on a CD device takes longer to arrive at the appropriate place than a hard disk. On hard-disks, the data on the platter can be interleaved to increase efficiency; data on CDs are recorded sequentially.
| ClickFanatic wrote: | | The mechanical aspect of reading CDs is very similar to that of a hard disk, right? |
Yes, they are similar, in the broadest sense. That isn't really as important as the speed of the device itself. In Unix-like operating systems, all the operating cares about is that a CD and a hard disk are both random access block devices. The driver for the hardware takes care of interpreting the unique hardware characteristics for the operating system.
_________________
Misanthrope: someone who realizes that humans really are as stupid as they appear.
If you think I'm 'politically' incorrect you have the wrong politics.
Big business is a disease we will need to cure before we will ever achieve real prosperity. |
|
| Back to top |
|
| |
|
|
|
Forum FAQs 
Knowledgebase 
Memberlist 
Usergroups 
Hosting
Register
Donate 
Profile 
Log in 
