I work at a reality tv post house and we’ve been using the G-RAID removable drives with (2) 4 TB drives inside them configured to raid 0. We’ve used these drives for 3 seasons of shows so far, filling them up to probably around 90% capacity before reallocating to different shows.
I wasn’t the one who reallocated them so data was copied to them before they could be fully reformatted. Now when data from multiple sources is copied to the drives, the write speed slows down from 90-130MB/s (depending on source drive) to 15-17 MBs per drive. Obviously this isn’t due to the usual bottle neck of multiple write processes (usually see a drop from 90 to 70ish with 5400rpm lacie ruggeds) so wondering what the issue is. I backed up the contents that are currently on there and tried an erase within disk utility to no avail. Also tried reconfiguring the raid array in raid configurator. Attempting to zero it out right now but would prefer to avoid the 8 hr process if possible.
Any ideas? Certainly cannot deal with 15MB/s write speeds as I consistently need to copy from multiple sources.
If you are getting that kind of performance and the drives are set to RAID0 then one of the drives might actually be on the way out and it is effecting the performance of the unit.
I would suggest switching the RAID to JBOD and then formatting each disk separately and try writing to each one and see if it allows full speed to the drives on their own. If one is going very slow then you know the culprit.
I’d be surprised if any of the drives are defective though as they are well under a year old and have only been filled to 90-95% once or twice. Would this be considered premature?
After running a bunch of tests, the slow write speed is occurring with 4 of our RAID0 arrays, which could mean 4 of our individual drives could be bad? Again these are maybe 6-8 months old so this would seem unusual at the least.
We use the rsync command within Terminal on Macs as a check sum for transfers and that gives a live readout of MB/s. The exact command is "rsync -avP "followed by the source file path and then the destination file path. Destination being the GRAIDs in question.
I’ve also used Black Magic Disk Speed test and had similar results to what I previously described. Though some drives appear to be underperforming in general. One in particular started with read/write speeds approaching 400MB/s, and now that one is around 130-150MB/s. As soon as I start an independent copy while the disk speed test is running, write performance plummets. I don’t have an exact number but will report back once I’ve tested.
Obviously multiple transfers will cause some bottle necking but in the past I’ve transcoded raw material from these GRAIDs while copying from two Field drives (that write at 90MB/s max) and some times copying back material to the field drives and only seen 15-30% decreases in speed. Maybe we’ve been over working these drives and not known it, but it’s my understanding that the advantages of Thunderbolt 2 is the ability to handle multitasking from multiple devices simultaneously.
Moving forward I’m going to do what you suggested yesterday and then try to get exact write speeds of each problem array with and with out another copy going simultaneously. I’ll also note the free capacity of each problem drive in case that is useful information at all.
Not 100% positive yet, but it appears formatting to JBOD, then formatting each separate disk, and then reformatting back to RAID0 seems to have alleviated the issue on at least one drive.
Will report back with more info when I have it. Preliminary results show same performance for each separate drive after formatting to JBOD and then formatting each drive.
It could have been a logical corruption in the overhead of the original RAID. Resetting it like that could alleviate as you’ve seen. Let us know how it goes from here.
Okay so this has become more complex than I imagined but it may be down to firmware / OS differences between computers, or maybe it’s the use of terminal.
On one computer the drives appear to transfer at normal speeds while multitasking. On two others they do something pretty bizarre. As I mentioned before, if I start one copy, it’ll transfer at it’s normal speed (90MB/s for the field drive I’m using), as soon as I start a 2nd transfer from a very similar drive through the same interface (thunderbolt 2, one daisy chained, one plugged directly into computer) the write speed for both drives plummets to 15MB/s. Now I just discovered that if I start a THIRD copy, the drives speed up closer to normal write speeds at roughly 50MB/s. I then didn’t see any speed drop off until I was copying 5 different folders from 3 different sources.
Going to do some more testing but really pretty baffled at this point!
Okay so, pretty much every variable addressed at this point and I’m pretty sure I’ve narrowed it down to using the rsync function within terminal. Dragging and dropping shows perfectly normal write speeds with 1, 2, 3 or more transfers. And running both AJA and Black magic disk speed test at the same time showed similar results. So I believe these drives are in perfectly fine health and maybe a couple of our systems just need some maintenance done to them, or (and I doubt this) maybe these drives don’t ‘like’ this kind of checksum verification.
The exact command we run within Terminal is: 'rsync -avP '.
