Medium Access Control of Wireless LANs for Mobile Computing

One-line summary: A dynamic multiple-access protocol will be the core of the design of a MAC for WLAN's. Background and goals for WLAN MAC standard IEEE-802.11 are described, and GRAP (Group Random Access Protocol) is presented as a candidate for the standard.

Overview/Main Points

• WLANs can be either infrastructured (like cel network) or ad-hoc (like WaveLAN)
• Downlink usually by broadcast
  • Need to build reliable multicast on top of it
  • When does a unit "wake up" from powersave mode in response to a broadcast?
• Uplink requires a MAC--more difficult than MAC design for a wired LAN:
  • Dynamic channel characteristics (multipath, shadows, etc.)
  • PHY layer implementation (i.e. carrier sense is harder for wireless than for wired)
  • Mobility, dynamic NW topology
  • Handoff
• Design Goals for 802.11 WLAN MAC standard: (there are 20; here are the less obvious ones)
  • Transparent to different PHY layers
  • Time-bounded (QoS) services
  • Fairness of access
  • Doesn't demand complex PHY interface
• Some problems with existing MAC schemes for WLANs:
  • Throughput of ALOHA and 1-persistent CSMA/CD (Ethernet) protocols has unstable peak
  • CSMA/CA "useless" for WLANs since fades last several symbol times and RTS/CTS have high overhead
  • Polling has nice properties but too much handshaking overhead for BS's to keep track of roaming
• Proposed solution: RAP (randomly addressed polling)
  • Upon receiving READY broadcast from BS, MH's each send a random number on orthogonal (e.g. CDMA) channels
  • Do a few rounds of "random number exchange" until BS encounters a round with a "large" number of distinct random numbers
  • Assign channels to the MH's that transmitted those random numbers

Relevance

Flaws

• GRAP Requires orthogonal signaling for transmission of the random numbers!