Since I happen to have a shielded RF test enclosure in front of me with an access point and a Squeezebox in it, here are the actual numbers. I have the computer wired directly to the squeezebox, and squeezebox connecting to the access point with bridging enabled.
First, ping the Squeezebox directly over ethernet: # ping -f -c 10000 -q 10.1.1.10 PING 10.1.1.10 (10.1.1.10): 56 data bytes --- 10.1.1.10 ping statistics --- 10000 packets transmitted, 10000 packets received, 0% packet loss round-trip min/avg/max/stddev = 0.165/0.214/1.760/0.038 ms Now ping the access point through the squeezebox: # ping -f -c 10000 -q 10.1.1.2 PING 10.1.1.2 (10.1.1.2): 56 data bytes --- 10.1.1.2 ping statistics --- 10000 packets transmitted, 10000 packets received, 0% packet loss round-trip min/avg/max/stddev = 1.266/1.945/10.717/0.456 ms And now with big packets to the squeezebox: # ping -f -c 10000 -q -s 1460 10.1.1.10 PING 10.1.1.10 (10.1.1.10): 1460 data bytes --- 10.1.1.10 ping statistics --- 10000 packets transmitted, 10000 packets received, 0% packet loss round-trip min/avg/max/stddev = 0.410/0.525/1.950/0.045 ms And to the access point: # ping -f -c 10000 -q -s 1460 10.1.1.2 PING 10.1.1.2 (10.1.1.2): 1460 data bytes --- 10.1.1.2 ping statistics --- 10000 packets transmitted, 10000 packets received, 0% packet loss round-trip min/avg/max/stddev = 2.129/2.857/12.592/0.396 ms And for good measure, the loopback: (31% packet loss?!? this is on macos) # ping -f -c 10000 -q -s 1460 127.0.0.1 PING 127.0.0.1 (127.0.0.1): 1460 data bytes --- 127.0.0.1 ping statistics --- 10000 packets transmitted, 6813 packets received, 31% packet loss round-trip min/avg/max/stddev = 0.017/0.022/0.104/0.005 ms Taking the averages and subtracting the loopback we get: SB 56B: 192 us AP 56B: 1923 us SB 1400B: 503 us AP 1400b: 2835 us And now you can work out all sorts of things... The difference in transmission time for 1400B vs 56B @ 100Mbps == (1400 - 56) * 8 * 10 ^ -8 == 108 us. Multiply by two for the rtt == 216 us. 503 - 192 - 216 == 95 us. That is the amount of extra work done somewhere, probably mostly on the computer's side for copying to/from the ethernet chip, to handle a large packet versus a small one. There is also some time spent verifying IP checksums on both sides. The transmission time for 56 bytes (plus preamble and frame headers == 56 + 28 == 84 bytes) would be 84 * 8 * 10 ^ -8 == 7 us. So we can see that the 192 us ping time for a small packet is dominated by os and application overhead: 192 - 7*2 = 178 us If we take the 95 us of overhead for response time difference between big packets and small packets through the ethernet link, and subtract that from the difference between big packets and small packets through the wireless+ethernet path, then we can work out the transmission speed of the wireless: 2835 - 1923 - 95 = 817 us. Divide by two since that's both directions == 409us. And to get the data rate: (1400-56)*8 bits per 409 us == 26.3 Mbps. Now test this by doing a file transfer through the SB+AP to another computer (both directions): # scp [EMAIL PROTECTED]:~/bigfile . bigfile 100% 10MB 3.3MB/s 00:03 # scp bigfile [EMAIL PROTECTED]:~/bigfile bigfile 100% 10MB 3.3MB/s 00:03 3.3*8 == 26.4 Mbps... nifty eh? -- seanadams ------------------------------------------------------------------------ seanadams's Profile: http://forums.slimdevices.com/member.php?userid=3 View this thread: http://forums.slimdevices.com/showthread.php?t=31850 _______________________________________________ discuss mailing list [email protected] http://lists.slimdevices.com/lists/listinfo/discuss
