Pegasus2 Performance

jasonvp

macrumors 6502a
Original poster
Jun 29, 2007
604
0
Northern VA
Originally posted in another thread in the Peripherals section:

I've been seeing what I'd call "interesting" performance issues with the Pegasus2 enclosures that I have connected to the new Mac Pro. Each is on its own Thunderbolt channel. One has 4 drives in it, with 2 RAID0 volumes. The other has 2 drives in it, also in a RAID0 volume. These RAID volumes were created using the Promise utility, not OS X.

The drives in question are 7200RPM SATA2 or SATA3; they're certainly not the fastest things available.

What I've found with Blackmagic is inconsistent read and write results. Sometimes a volume will get 200+MB/s reads and writes. Other times it'll stumble with the test first starts, and stay around 10MB/s (not a miss-print). It applies to both enclosures.

As a test, I nuked two of the volumes, reset the drives to "Pass-Thru" mode, and recreated the RAID0 volumes in OS X software. The results? Consistent performance numbers of ~250MB/s on reads and writes. Not necessarily better, but reliable.

At this point, I can't recommend the Pegasus2's internal RAID controller if performance is needed. Certainly if you intend on using RAID5, which OS X doesn't support internally, then the enclosure's controller would work. But if you're just doing RAID0 or 1, there's really no reason to not use software RAID. The CPU in the Mac is certainly significantly more powerful than what's in the enclosure, and handling RAID/storage requests won't make it break a sweat at all.

What have others found?
 

snouter

macrumors 6502a
May 26, 2009
767
0
Out of the box RAID 5 with the Toshiba drives I'm seeing 450-600MB/s write and 400MB/s reads.
 

RoastingPig

macrumors 68000
Jul 23, 2012
1,605
62
SoCal
try some real world test by copying lets say like a 50gb folder and time it on a pegasus raid0 configuration and then the same but with the osx software raid...to see if theirs any difference
 

jasonvp

macrumors 6502a
Original poster
Jun 29, 2007
604
0
Northern VA
try some real world test by copying lets say like a 50gb folder and time it on a pegasus raid0 configuration and then the same but with the osx software raid...to see if theirs any difference
The problem with doing that is the disk cache intrinsic with OS X. While real world tests are useful in some cases, synthetic benchmarks are a better way to do direct comparisons of the disks like this.

VirtualRain said:
BlackMagic is not that reliable.
Based on what? It's used in every benchmark I've come across.
 

jasonvp

macrumors 6502a
Original poster
Jun 29, 2007
604
0
Northern VA
i personally use aja hard drive test.
AJA results for the same pair of disks using a 4GB file size:
Software RAID writes: 220MB/s
Software RAID reads: 230MB/s

Hardware RAID writes: 290MB/s
Hardware RAID reads: 180MB/s

The results are interesting, and given the faster reads, I think I'll stick with the software RAID0.
 

VirtualRain

macrumors 603
Aug 1, 2008
6,304
114
Vancouver, BC
Back it up and I'll be happy to believe you. Otherwise you're contributing precisely nothing to this thread, and you're completely free to: move along.
Sheesh, why so harsh? I'm trying to help you get consistent benchmarks rather than waste your time troubleshooting a problem with the Pegasus2 that may not exist.

May I present to you, your own experience as evidence?

What I've found with Blackmagic is inconsistent read and write results. Sometimes a volume will get 200+MB/s reads and writes. Other times it'll stumble with the test first starts, and stay around 10MB/s (not a miss-print). It applies to both enclosures.
Which happens to mirror my own experience... it was very unreliable on my own external WD Duo setup whereas QuickBench and timing file transfer tests worked fine.

At any rate, I'm moving along now. Seeya.