|
U320 SCSI HDD ROUNDUP
Performance Summary - Software Raid, RAID_0
Now, we were curious to discover if the Adaptec
Host Raid showed any advantage or disadvantage over
the built-in software RAID_0 or "disk striping" of
Windows XP, 64-bit edition. We decided to use
the Seagate 15K.5s for this test.
Interestingly, this opened up a new possibility as
well - the ability to stripe across two different
SCSI channels, something not possible with Adaptec
Host Raid.
The % difference gain or loss of using different
channels over same channels are shown below.
Most notably, there is a spike in the 16k row,
where a 169% increase in write speed and 15%
increase in read speed was measured when
different SCSI channels were used.
| |
% Difference in
Writes
(Diff. Channel - Same Channel) |
% Difference in
Reads
(Diff. Channel - Same
Channel) |
|
0.5 KB |
-1.23% |
0.47% |
|
1.0 KB |
-2.18% |
-36.0% |
|
2.0 KB |
-18.22% |
9.48% |
|
4.0 KB |
-45.87% |
2.06% |
|
8.0 KB |
-0.74% |
11.57% |
|
16 KB |
168.99% |
14.83% |
|
32 KB |
39.5% |
6.67% |
|
64 KB |
16.7% |
-4.62% |
|
128 KB |
19.2% |
1.82% |
|
256 KB |
-0.97% |
-0.81% |
|
512 KB |
1.98% |
3.33% |
|
1024 KB |
5.16% |
3.48% |
Not surprisingly, striping across the two
different SCSI channels showed a small yet
measurable increase in performance percentage
wise. This may be a good argument to
purchase a controller (regardless of type,
whether it be SCSI, SATA or PATA) with more
ports than one with built-in hardware RAID, if
you are simply looking for pure RAID_0 I/O
throughput. Of course, hardware
controllers offload CPU workload, and most
importantly, the parity calculations when
redundant RAID levels are implemented, such as
RAID_5 or RAID_6.
|
