Slashdot Mirror


User: BodeNGE

BodeNGE's activity in the archive.

Stories
0
Comments
28
First seen
Last seen
Profile
(view on slashdot.org)

Comments · 28

  1. Re:Zero Packet Loss on A Possible Cause of AT&T's Wireless Clog — Configuration Errors · · Score: 1

    Yes, the radio uplink to the tower. A mobile device will 3 to 10 times faster in the downlink than the uplink. The tower to the SGSN is symetric too. The GGSN to the Internet connection is a regular symetric.

  2. Re:Software Robustness on A Possible Cause of AT&T's Wireless Clog — Configuration Errors · · Score: 1

    On Windows Mobile you can reduce the MTU to help on long-thin networks (high delay, low bandwidth). Sadly, it is about all you can do...

  3. Re:Zero Packet Loss on A Possible Cause of AT&T's Wireless Clog — Configuration Errors · · Score: 2, Interesting

    OP here. It could possibly be done between the IP stack of the device and the GGSN. The six layer stack diagram in the article does show the IP layer going between the two, but in reality it doesn't. On the mobile side there are usually two IP stacks. The GSM one to the GGSN access point (APN), and the stack presented to the client AP, or the the PC connected to it (calleed Terminal Equipment in GSM speak). If you tracert from the device you will usually see the device IP as a loopback address. The GPRS/3G part then takes over and spits our the rest of the IP connection on the GGSN side. The GGSN side may give you a public IP if you pay for it, but usually it is a pooled address, and not externally addressable. You could change the "inner" IP link between the GGSN and MS if it was implemented in the MS IP stack. You wouldn't need to implement it in the outer bearer servce between the PC and the destination IP. Other than that there really isn't much you can change/tweak in the network. Smaller buffers will cause packets to be discarded, then TCP takes over as normal. Setting the MTU size on the GGSN also helps. Optimum from the MS perspective is around 1400.