SSD MacBook Air to LaCie Thunderbolt HD Slow Transfers

[tcsdoc]

I have a 2011 MacBook Air with the Samsung 256GB SSD drive. I purchased the LaCie 2TB Thunderbolt external hard drive to use for backups. I'm using SuperDuper 2.64 to backup my internal hard drive. I am backing up to a Read Only image to the Thunderbolt drive. Effective copy speed is only registering at ~30MB/s. I can transfers a 1GB file from my SSD to the LaCie in 38 seconds. Why are things much slower using SuperDuper?

[dnanian]

Probably because you're doing twice the I/O, to the image's virtual volume and then the image to the drive. Have you tried writing directly to the drive instead?

[tcsdoc]

I know this is mixing apples and oranges but I used Finder to copy my Documents folder (containing 2066 files in 28.86GBs) to my LaCie drive in just over 3 minutes, yielding around 160MBs per second transfer rate.
I erased the LaCie drive and used SuperDuper to copy my Macintosh HD to the LaCie drive (SmartUpdate). I did not check permissions. SuperDuper reported copying 449206 files with an effective transfer rate of 59 MBs per second (varied thru copying process starting out reporting 20MBs then reaching 63MBs at the end).

Any idea where the overhead is coming from for such a wide disparage in results?

[dnanian]

You guys are making me not want to print any statistics at all...it all depends on the mix of files, metadata, etc...

[tcsdoc]

Don't stay up all night worrying about it. I'm not a techie and was curious. I'll find a different program with a helpful support staff that offers assistance instead of wisecracks.

[dnanian]

Wisecracks? Seriously, I wasn't wisecracking. The problem with showing a statistic is that people compare "real world" results - with a mix of different sized files, I/O patterns, disk placement, metadata updates, etc - with "bus speed" results, unproductively. It's intended to help you determine whether, say, a USB drive is running at low-speed (~.1MB/s) vs. a comparison with box specs.

In general, those specs are "ideal" rates. So, with an infinitely fast destination accepting a single stream writing to a blank disk, you get fast speeds (maybe). On top of that, when people then "test" the drive using Finder, they'll tend to grab a big file or two and copy that - which is, again, a different kind of I/O (two large streams).

Finally, remember that the specs for the drive are almost always in mb (megabits) not MB (megabytes) per second - an 8-times larger number.

We're feeding the files to the system as fast as it will accept them. We then have to update the metadata (which involves additional I/O). But we're not CPU bound, and the backup is proceeding very close to whatever the calls have as their maximum rate.

[dnanian]

One more thing to mention - we ensure that all the metadata for the files copied (ACLs, EAs, etc) are all updated properly, something Finder (and OSX's own file calls) don't usually do. This means that some of those attributes are copied twice (bad version replaced with good version), which does some additional I/O, as you might expect.