I have recently added a new Toshiba P11 laptop to my wireless home network, running WinXP SP2, and with an onboard IEEE 802.11n wireless adapter (Intel® Centrino® Advanced-N 6200 AGN to be specific).
I use a TP-Link TL-WR841N wireless router (in Access Point only mode, since I use another home PC as my routing "gateway" to the outside world, more specifically my Cable Modem). But I digress... the real issue is what's going on, on the _inside_ of my home network.
I noticed that my new laptop performed atrociously when I used IPX/SPX Windows file sharing (i.e. around 5-10 Mbit/sec at best in excellent signal conditions, in 802.11n mixed mode), and yet terrifically when the Windows file sharing used TCP/IP as the underlying protocol (i.e. regularly within the 45-40 Mbit/sec range). The way I first spotted this problem was to note that anything TCP/IP-based to destinations outside of the home network, were going at lightning fast rates, and yet the normal drive-mappings, and streaming of media from locally served content was terrible and unusable (i.e. I originally had it using IPX/SPX).
If this was the same story with my existing Toshiba A8 laptop (running WinXP SP2 and with an onboard IEEE 802.11g wireless adapter, Intel® PRO/Wireless 3945ABG ), then I probably wouldn't be posting this. I'd merely write it off as TCP being superior to SPX in my particular networking environment (i.e. things which come to mind are, perhaps, better error recovery due to its true sliding-window nature over SPX's burst mode, which for all I know could be set to just "1" in my case – making it a stop-and-wait protocol...). The problem is, for this older laptop with the IEEE 802.11g adapter, the situation is completely reversed!!! In most ad-hoc tests I've done (using both Windows Explorer and FTP server file up/downloads), it performs about 3-4 times better with IPX/SPX as the underlying protocol in support of Windows file sharing, as compared to TCP/IP.
So... go figure!? Why is this the case? Why should there be protocol sensitivity to the underlying wireless technology, in the case described above? i.e. seemingly, in my "radio propagation environment", for some reason the IPX/SPX and 802.11g prefer each other on one computer, while the TCP/IP and 802.11n prefer each other on the other computer. And the interesting this is, it appears to be very reproducible.
The only possible theory I have (although I admit I'm not wholly satisfied with it), is this: for some reason the 802.11n, albeit running at a far faster speed, also has more "burst errors" at the physical layer; these are then more easily dealt with in the context of a sliding window protocol, especially if running enhancements like TCP SACK.. compared a stop-and-wait behaviour of SPX (if as I suspect it is behaving in this way). Oh and the other thing which I noticed was that on the 802.11g laptop, the PING times to my gateway are reported as <1ms, while on the 802.11n laptop, they are =1ms. So the longer latency and more bursty error profile of 802.11n, would make TCP more suitable than SPX in this context.
Any thoughts/comments/alternative theories? I'd LOVE to hear from you.
Thanks for your time all,
Milo