RAID Striped increase speeds with regular USB HD?

[michaelwithe21]

I'm quite new to the OS (Level) Raid concept provided by the Disk Utility within Leopard OS X 10.5.6 and other OS X OS's, but I saw a great video http://cnettv.cnet.com/2001-1_53-50004485.html showing 2 USB External HD's hooked up to a macbook running a raid to tie them together (striped or mirrored). This concept is simple! But heres my situation:

I have an iMac 22" 2.4 GHz Intel Core 2 Duo, 4 GB 667 MHz DDR2 SDRAM, Running off of a "298.1 GB WDC WD3200AAJS-40VWA0 Media Hard Drive". I also have an External Drive "465.8 GB WD 5000AAV External Media".

What I did: Booted to leopard disk, used disk utility to "Merge" a "STRIPED RAID" from my Local (built in 298 GB) HD with a PARTITION of my External (same size 298) as to not waste any left over space on the External (Only as good as the smallest partition). I used 32K block or whatever was default... Everything went very smoothly!

QUESTIONS:
1) Does anyone think this could possibly speed up my system in any aspect?
2) Or slow down my system in any aspect? (because of the USB 2.0 limiter)
3) Was it "BAD" that I used a partition instead of the whole External hard drive like the video shows?
4) Minus the "loosing everything if the OS fails", is it a "BAD" idea to use a "Soft" RAID with the HD that your OS runs off of?
5) Do you know of any "Speed Tests" that would prove this one way or another?

I herd that you are required to have at least 3 Drives to have a "STRIPED" RAID... but obviously the OS layer of RAID in OS X allows you to use 2.
I also understand that if my OS fails, than my data on the External (or the Internal for that matter) may not be recoverable. IM OK WITH THIS.

**Can Striped Raid With Local And External USB Drive Increase System Speeds?**

 

[Amdahl]

Only a benchmark can tell for sure. I would expect diminished speed on small files because your internal drive supports SATA NCQ, which lets multiple read requests be outstanding at one time. USB-IDE bridging does not support this. I would expect that sequential reads might be same or faster due to drive read-ahead, but issuing multiple random reads would probably be slower, because the reads that land on the USB would not get to the drive at once, and instead be one at a time.

If you can't tell the difference, then it isn't worth the trouble unless you want the merged size of both volumes in one.

 

[michaelwithe21]

Thank you for your insight Amdahl...
Unfortunately I didn't run any tests before I made the switch to this Striped RAID... so as far as benchmark is concerned... I'm screwed...
But I THINK I have noticed a bit of speed increase in boot up time (but this could be caused by the clean install) and individual File transfers to and from other HD's, but then again... my mind tends to lean in directions I hope for =).

Please if you can... clarify what you meant by "If you can't tell the difference, then it isn't worth the trouble unless you want the merged size of both volumes in one". I mean... I cant really notice a difference unless I'm running allot of transfers at once... or trying to open multiple applications while file transfers are in process. I hear my External running allot during curtain processes (cant really specify which ones)... As far as the "...unless you want the merged size of both volumes in one." isn't that what I already did?

Questions:
1) Does the OS decide (by speed) which Slice (HD) gets used for Files (or how often)?
2)Is there any way to see how much of each Slice is using (space or CPU)?
3)Or any way to "prefer" the Sata HD so as not to cause lag when not needed?

 

[michaelwithe21]

Udpate

Here is my XBench Results after my Striped RAID "See above for computer and more info" with a Partitioned USB External.. Unfortunately I did not know about this app until after I made the switch.
BUT: does anyone see anything in the Disk Test or other that might show Increased or Decreased Speeds caused by the Striped RAID?

Results 161.76
System Info
Xbench Version 1.3
System Version 10.5.6 (9G55)
Physical RAM 4096 MB
Model iMac7,1
Drive Type iMacHD
CPU Test 151.47
GCD Loop 282.25 14.88 Mops/sec
Floating Point Basic 136.52 3.24 Gflop/sec
vecLib FFT 112.46 3.71 Gflop/sec
Floating Point Library 150.42 26.19 Mops/sec
Thread Test 220.51
Computation 212.74 4.31 Mops/sec, 4 threads
Lock Contention 228.87 9.85 Mlocks/sec, 4 threads
Memory Test 159.64
System 163.87
Allocate 239.01 877.74 Kalloc/sec
Fill 137.51 6685.84 MB/sec
Copy 145.97 3014.98 MB/sec
Stream 155.62
Copy 142.82 2949.84 MB/sec
Scale 149.84 3095.64 MB/sec
Add 166.42 3545.11 MB/sec
Triad 166.14 3554.24 MB/sec
Quartz Graphics Test 194.67
Line 177.93 11.85 Klines/sec [50% alpha]
Rectangle 229.93 68.65 Krects/sec [50% alpha]
Circle 187.85 15.31 Kcircles/sec [50% alpha]
Bezier 187.08 4.72 Kbeziers/sec [50% alpha]
Text 198.13 12.39 Kchars/sec
OpenGL Graphics Test 182.99
Spinning Squares 182.99 232.13 frames/sec
User Interface Test 352.37
Elements 352.37 1.62 Krefresh/sec
Disk Test 80.42
Sequential 76.62
Uncached Write 89.60 55.01 MB/sec [4K blocks]
Uncached Write 91.29 51.65 MB/sec [256K blocks]
Uncached Read 43.90 12.85 MB/sec [4K blocks]
Uncached Read 136.87 68.79 MB/sec [256K blocks]
Random 84.62
Uncached Write 45.11 4.78 MB/sec [4K blocks]
Uncached Write 159.36 51.02 MB/sec [256K blocks]
Uncached Read 89.40 0.63 MB/sec [4K blocks]
Uncached Read 130.86 24.28 MB/sec [256K blocks]

Anyone See anything in the Disk Test or other that might show Increased or Decreased Speeds caused by the Striped RAID?

 

[Amdahl]

Questions:
1) Does the OS decide (by speed) which Slice (HD) gets used for Files (or how often)?
2)Is there any way to see how much of each Slice is using (space or CPU)?
3)Or any way to "prefer" the Sata HD so as not to cause lag when not needed?

I don't know the answer to any of those, although I suspect #1 is a No. I doubt speed is taken into consideration, since RAID-type applications presume equality in the drive elements. If it is a stripe, it would alternate between drives in an equal fashion. A merge without striping, could be doing anything.