AirPort’s Interference Robustness and Transfer Speed
Posted September 23rd, 2006 @ 03:31pm by Erik J. Barzeski
I'm sitting downstairs on my 12" PowerBook. I've mounted via file sharing a disk on my G5 upstairs. The transfer of 400 MB of pictures was going to take about 40 minutes (after a minute of waiting). I unchecked "Use Interference Robustness" on my PowerBook and the estimate dropped to seven minutes. I waited a minute and turned it back on again - the rate at which the transfer came through slowed once again. Turn it off, it sped up.
Substantially.
Have there been any studies done on how much AirPort's Interference Robustness slows down file transfers?
P.S. Turning it off on the G5 resulted in a similarly substantial gain. I ran another transfer just for the heck of it. I'll test this theory with just one (the PowerBook) using interference robustness versus none using it and post times in the comments.
Donate Life
Posted 23 Sep 2006 at 4:10pm #
File was my FSS.dmg file, 390 MB.
No Interference Robusness
Sat Sep 23 15:35:30 EDT 2006
Sat Sep 23 15:43:18 EDT 2006
Total Time: 7:48
Using Interference Robustness
Sat Sep 23 15:44:42 EDT 2006
Sat Sep 23 15:52:57 EDT 2006
Total Time: 8:15
I ran them in that order, which may or may not have had some sort of affect, and using the same file. The G5's robustness remained "off" the entire time.
The experience I talked about in the actual post above was with 400 MB of JPEGs, and the robustness seemed to result in a 2x slowdown, not a 30 second slowdown over eight minutes. So, I'd call these quick results inconclusive.
Posted 23 Sep 2006 at 4:36pm #
I tried copying a 2.3MB file from my ibook G4 (airport) to intel imac (wired)
with Interference Robustness - 15 seconds
without Interference Robustness - 20 seconds
- vineet
Posted 23 Sep 2006 at 11:45pm #
From what I have read, turning on "Interference Robustness" instructs the card to use smaller payloads and not fail-over to the lower speeds. The probability of bit-errors (caused by phones/microwaves) is reduced with smaller packets. The cost is that smaller packets are less efficient so the throughput is reduced significantly.
For reference, 802.11 has 34 bytes (I think?) of headers and the payload can vary from 0 to 2312 bytes. Add on the ACK for each fragment. I don't know what size payload Apple's card uses with Interference Robustness turned on, but I imagine it's probably around 128 or 256 bytes. So yea, throughput goes to heck (that is, if you weren't experiencing problems with bit-errors with large packets).