To all
I've been working on a project to help a non-technical friend set up an automatic Ghost backup system that requires as few as *one click* to create multiple backups of their system--I will be reporting on that in a separate thread--hopefully shortly--the system has been setup and my friend is testing to see if there are any glitches
before I report it here.
But, as a side project, I thought I would test how best to organize my HDD's on a given system for best performance. My first thought was that best performance would be to have one HDD as master on the primary HDD controller--the OS HDD basically--and then have the second HDD that would be the destination for Ghost images on the secondary controller, either as master or slave. My reasoning, one would likely get better performance when creating an image if the one controller was *reading* data while the other controller was *writing* data.
So, the test system is an older Asus CUV4X motherboard-which has a *Apollo Pro 133A* chipset (a VIA VT860 4/5/6 Pro Savage south bridge VIA Bus Master IDE controller)--vintage approximately 9/2000.
Test #1:
The source HDD has WinXP as the OS and is formatted NTFS--is hooked up on the primary channel as master. The destination HDD was formatted as FAT 32, and initially it was easiest to hook it up as slave on the secondary channel.
Ghost image creation:
245 MB/min, for 3250 MB of data--it took 13:12 minutes
Integrity check:
409 MB/min, for 3250 MB of data--it took 7:56 minutes
That was not very impressive speed.
Test #2:
Left the source OS HDD hooked up to the primary, as master, but now placed the destination HDD onto the primary as slave, i.e. both on the same channel.
Ghost image creation:
328 MB/min, for 3263 MB of data--it took 9:53 minutes (a 33.88% increase in speed!)
Integrity check:
696 MB/min, for 3263 MB of data--it took 4:40 minutes (a 70.17% increase in speed!)
Well, so much for my theory that putting the source on primary channel and the destination on the secondary channel would possibly increase performance!
Test #3:
Again, leaving the source OS HDD as primary, master--
I reformatted the destination HDD as NTFS
and placed it as master on the secondary channel
Ghost image creation:
511 MB/min, for 3263 MB of data--it took 6:23 minutes (a 108.57% increase in speed from the original!)
Integrity check:
1553 MB/min, for 3263 MB of data--it took 2:06 minutes (a 279.70% increase in speed from the original!)
Wow--the destination file system (NTFS vs FAT32) made a huge difference in performance!
(Christer--if you happen along to see this thread--I'm sure you will recall our thread a long time ago about VIA IDE HDD controllers and different performance in DOS depending on the file system (NTFS vs FAT32) that is the source and the destination HDD!)
Test #4:
Again keeping the source OS HDD as primary, master--now placed the destination HDD (still NTFS formatting) as primary, slave.
Ghost image creation:
512 MB/min, for 3263 MB of data--it took 6:22 minutes
Integrity check:
1201 MB/min, for 3263 MB of data--it took 2:43 minutes (a slight decrease in speed!)
Okay, so at least for the integrity check, being on a separate channel was faster vs both HDD's being on the same channel--but being as there is no actual *writing* during an integrity check--only *reading*--not sure what that proves--except it was faster if the two HDD were on separate channels on this system.
So, in summary--depending on your system's HDD controllers, and the chipset that you have--performance may depend on what file system you choose, and the position you choose for your HDD--only testing and recording the results will give you answers for your given system.
My friend's system is several years newer--and does not seem to have the same slower results as seen here when going from a source NTFS HDD to a FAT32 destination HDD--to create the *automated* backup routine for my friend's system, I had to use FAT32 for the destination HDD--I'll be reporting on that soon.