Promise Pegasus R6 & Mac Thunderbolt Review
by Anand Lal Shimpi on July 8, 2011 2:01 AM ESTThe Pegasus: Quirks
I did encounter occasional glitches with the Promise Utility. No show stoppers, but annoying nonetheless. Sometimes when deleting a logical drive I'd get an error telling me that the delete operation failed (even though it didn't):
Refreshing the logical drive page revealed that the drive had been deleted, despite the warning.
Sometimes various fields will be populated with not a number (NaN) instead of the actual data itself. Just as before, refreshing the page in question usually cleared the problem:
The Pegasus itself was most problematic when operating at or near its capacity. In one case I tried filled a 10TB RAID-5 array with 10TB of data. Rather than copy a bunch of large files over and over again, I used Iometer to generate a single 10TB file on the drive. Somewhere around the 9TB marker Iometer stopped writing to the drive. Mac OS X reported a single 10TB file on disk but the actual file was under 9TB in size leaving over 1TB of free space on the drive. I force quit Iometer and tried writing to the drive manually. At this point the drive became incredibly slow to respond. I tried canceling the copy but Finder crashed under OS X. The Pegasus itself actually hung and refused to power down, I had to pull the plug on the device in order for it to power cycle. When I plugged it back in the R6 wouldn't appear under OS X. I had to pull two drives to break the RAID-5 array then delete/recreate the array to get it working again. Of course I lost all of the data I wrote to the drive, thankfully it was just a bunch of repeating bytes created by Iometer. I attempted the same thing again (twice) and couldn't duplicate the issue. I'm going to assume this was an Iometer related issue (or a problem with creating a single ~9TB file on the array), but it's worth disclosing regardless.
The only other time I had an array go bad was when I swapped in four SandForce SSDs and created a giant RAID-0 array. One of the drives simply dropped out of the array, forcing me to delete and recreate the array. As I mentioned earlier, I can't be entirely sure if this is a Promise issue, SandForce issue or a little of both. I never had a drive mysteriously disappear when using the Hitachi drives that came with the Pegasus however.
Other than the issues I've mentioned here, I didn't encounter any problems during my testing of the Pegasus R6.
Facebook
Twitter
RSS
88 Comments
View All Comments
Conner_36 - Friday, July 8, 2011 - link
Or even in the office, to able to take your entire project and move between the rooms carrying ALL of the data? That's unheard of!From what I understand with HD movie editing I/O is the bottleneck.
All we need now is an SVN hardware device with thunderbolt to sync across multiple thunderbolt RAIDs. Once thats out you could have a production studio with some real mobile capabilities.
Exodite - Friday, July 8, 2011 - link
I wager pretty much any usage scenario can come up with a high-performance 12TB storage solution for significantly less than 2000 USD.You're right though, it's definitely not the solution for me.
Or anyone I know, or am likely to ever know. *shrug*
Zandros - Friday, July 8, 2011 - link
What happens if you try the Macbook Pro -> Pegasus -> iMac in Target Display Mode -> Cinema Display connection chain?Focher - Saturday, July 9, 2011 - link
Pretty sure the DP monitor has to be the last device in the chain. Maybe that is just a current limitation because there are no Thunderbolt displays.Zandros - Monday, July 11, 2011 - link
AFAICT, the iMac is a Thunderbolt display, since it does not support Target Display Mode from Display Port sources with Display Port cables.tipoo - Friday, July 8, 2011 - link
Is there a way to make it shut off the drives after idling for a while?piroroadkill - Friday, July 8, 2011 - link
But when you saw the file creation maxed out at 9TB, on 10TB array..Since.. uh, Snow Leopard, Apple changed file and drive sizes to display decimal bytes as used by the manufacturers, which is the same as the 10TB array.
However every other thing ever reports in binary bytes, such as windows describing "gigabytes" even though it means gibibytes in reality.
Ugh, anyway, what I'm trying to get at is that maybe you did infact fill the array. That said, the thing shouldn't have fucked up..
CharonPDX - Friday, July 8, 2011 - link
If I had way too much money, my usage model for Target Display Mode would be to use the iMac as a Virtual Machine host/server, connected to either a second iMac or a MacBook Pro as a dual-screen workstation.With the minimum 27" iMac, you're basically buying a 27" Cinema Display plus a $700 Mac mini-on-steroids. If you want a second Apple display for your iMac or MacBook Pro, and want a Mac Mini to use as a server, that is an excellent value to instead just get a second iMac. (That value may drop depending on the next Mac Mini update, of course.)
etamin - Friday, July 8, 2011 - link
in the block diagram on the first page, why is the Thunderbolt Controller connected to the PCH thru PCIe rather than to the processor? I thought PCIe connections came off the processor/NB?repoman27 - Sunday, July 10, 2011 - link
The lanes that come off the processor/NB are usually used for dGPU. On the new MacBook Pros, Apple borrowed four of them for the Thunderbolt controller. Apparently on the new iMacs, however, they decided to give all 16 lanes from the CPU to the graphics card and pulled four from the PCH instead.