Jim Hauser said:
Greetings,
Some friends of mine have suggested that I take advantage of SATA RAID on
my Soyo CK-8 MB (AMD 64 3200+ CPU.) I'm already running a pair of IDE
drives; 120 Gig and 160 Gig.
My current main partition is 60 Gigs which is more than adequate at this
time. All else is archives and backups.
Any advice as far as RAID tutorials, drive models and all else would be
appreciated since this is new territory for me.
Would also appreciate any advice on the advantages/disadvantages of SATA
RAID.
If it matters in this discussion, I'll be sticking with the 32 bit OS for
a while longer (XP Pro.)
Hopefully, this link
http://www.rojakpot.com/default.aspx?location=3&var1=141&var2=0
will get you there. Adrian's Rojak Pot site uses frames so it is hard to
pin down the ULR. I tried it on two browsers and it worked for me.
If not, the
http://www.rojakpot.com is the site, select articles from the
left column and then select "RAID Optimisation Guide Rev. 4.1!" in the
"Tweaks" section.
It is aimed at hardware RAID but the same principals are used for software
RAID.
You can also set-up a software RAID if you are running WinXP Pro or Win2k
SP4 (I think). You need to set you disks to "dynamic" and from there you
can make your choice.
On my Abit NF7-S v2.0 I had a 160GB IDE HD as my boot disk and a 20GB HD
hooked up the same IDE port as a slave. The board came with a "SERILLEL"
adapter that I used to allow me to connect an 8GB HD via one of the SATA
ports. Each physical HD was used as a separate device, ie the 160GB was C:,
the 20GB was D: and the 8GB was G: - the two opticals using E: and F:.
All drives are Seagates, the 160GB HD is 7200.7 ATA100 the two others are
old ATA66 Seagate 5400 RPM ones with the plastic shroud. The RAID drives
were set up with a 16kB stripe.
Data on D: and G: was backed-up and they were both formatted with NTFS. I
used SiSoft Sandra to bench the two empty HDs and the 8GB HD got nearly
16MB/sec and the 20GB got a bit over 21MB/sec.
I converted both HDs to dynamic and set-up a striped RAID using all the 8GB
HD and the exact same sized partition on the 20GB HD. Using Sandra again, I
got a read of a little over 33MB/sec, not bad considering the 20GB HD was
slave on the first IDE port and the 8GB HD went through the adapter on the
SATA port.
After a reconfiguration of HDs with the 160GB boot HD moved the to SATA port
via the adapter and the 8GB and 20GB HDs sharing the same IDE port the
results using the same Sandra had the two RAID HDs getting a little over
35MB/sec.
To put into perspective, my nearly full 160GB HD got 43MB/sec when it was
using the SATA adapter and about 42MB/sec on the shared IDE channel with the
other HD in RAID configuration using the same Sandra.
CONCLUSIONS
My newish nearly full 160GB HD was faster than my empty old small slow
RAIDed HDs.
RAIDing the small old HDs through software gave an improvement approaching
the combined speeds of the individual disks.
A big late model HD is faster than 2 old small slow RAIDed HDs.
I got a slight speed increase when the boot 160GB HD was moved from a shared
IDE controller to a SATA port via an adapter.
OTHER CONSIDERATIONS
Not mentioned above, the two old small HDs in RAID on the same IDE port got
better burst speed test results than the combined total of each HD tested
separately. My 160GB HD had a higher burst speed that the two old HDs in
RAID because it has 2MB cache and each of the old HDs have 512kB cache.
During all my tests I used the same test proggy on the same OS. Other than
HD configuration, nothing else was changed except re-adding boot drivers for
SATA to make my 160 HD boot from SATA via the adapter.
My boot 160GB HD (NTFS) was not defragged before or during tests. The other
disks were tested after a complete format.
CRUCIAL POINTS
Modern HDs are great but not perfect. You must back-up crucial data on a
regular basis. If you have a DVD burner then you can use DVD-RWs to hold
your data and recycle them in order.
Dave
