I'm not sure I'm so surprised with the results you are getting. I apologise if you know this stuff, but for info:
A single, sequential, giant file copy is the absolutely kindest thing you can do to a disk drive to get good performance. It's the equivalent of trying to find the top speed of a car but testing on a perfectly straight road, going down a steep hill and with a strong wind behind you.
If you think about it, ignoring any bottlenecks in the electronics, the speed you'll get for a sequential read depends on 2 basic factors:
1: How fast the disk platter is rotating under the read head and
2: How tightly packed together are the bits - i.e. bit density
The main reason speeds have been climbing over the last decade are bit density. If I can pack twice as many bits in a linear inch even if my rotation speed doesn't change I'll double my sequential performance.
I don't know your Green/Black model numbers, but have you checked to see if they are the same platter bit density? If the Greens are 333GB platters but the Blacks are 250GB then the bit density advantage to Green could offset the rotational speed benefit of Black - for large sequential reads or writes.
I can't speak for NAS environments in general or for your own needs at all, but in large scale enterprise systems where I have experience you tend to have dozens of jobs/databases/mail systems and hundreds of users all beating on the storage at the same time.
Ignoring any queues that build up, once it is the turn of your IO operation the effective speed of response of the disk is governed by 2 key points:
1: How long it takes to position the disk head over the part of the platter where your data lives and
2: How fast it reads once it gets there (the peak sequential speed we mentioned above)
Item 1, the positioning time has two components also. How long it takes to move the disk arm back and forth to position the right distance from the centre of the spindle, and how long it takes for the first bit of your data to rotate round the spindle to get under the disk head.
A 5400rpm disk rotates once every 1/5400 of a minute or every 11.1 milliseconds. On average it takes 1/2 a rotation to get your data in position (sometimes it's just gone past when you start looking, sometimes you get lucky, on average it is half a rotation). A 7200rpm disk rotates once every 8.3ms. If your sequential read speed is 100 MBytes/second you can read (or write) 100 Kbytes in a millisecond. [This gives the interesting observation that if your drive can do 100 MBytes per second = 100 KB per ms, and (for a Green) we know that it takes 11.1ms per rotation, then the disk must have about 100KB * 11.1 = 1.1MB around a full track rotation.]
Anyway, for non-video processing, normal enterprise workloads reading customer records and bank statements, or even wordprocessing etc average IO requests are probably in the few KB to few 10's of KB, i.e. well under 1ms to read at 100KB per ms, so overall response is goverened by the positioning speed and sequential transfer speed is almost irrelevant, hence IO performance goes in line with latency = rotational speed, hence 7200,10k, 15krpm drives.
For giant, sequential, one IO does gigabytes of transfer then positioning time is zero to all intents and purposes and the only thing that matters is the values you are measuring, but you are seeing minimal benefit of 7200rpm possibly because of bit density discussed above.
Finally, watch out that a peak speed of 100-110MB/sec tends to be on the outer tracks of the disk, down at the centre you'll probably see half of that so an average over the whole disk may be around 75-80 and that is well within the cababilities of a good gigabit ethernet controller.
Hope I didn't bore everyone to death and that was of some interest to someone somewhere.
Kind regards
Crash
