Changes between Initial Version and Version 1 of Internal/WinlabMadwifi/MultiMACPaper


Ignore:
Timestamp:
Nov 16, 2005, 2:18:28 AM (19 years ago)
Author:
Surya Satyavolu
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • Internal/WinlabMadwifi/MultiMACPaper

    v1 v1  
     1WinlabMadwifi -> MultiMAC
     2
     3Paper Title:
     4= MultiMAC - An Adaptive MAC Framework for Dynamic Radio Networking  =
     5
     6
     7''Christian Doerr, Michael Neufeld, Jeff Fifield, Troy Weingart, Douglas C. Sicker, and Dirk Grunwald[[BR]]
     8''University of Colorado, Boulder''
     9
     10== Motivation ==
     11 * Bridging the gap between firmware based MAC and software radios
     12 * Capability to use dynamic MAC plug-in based on scenario.
     13
     14== Primary Achievements ==
     15This has been done with Madwifi driver on Atheros chipset.
     16
     17 * Override 802.11 MPDU frame format
     18   * Necessary to send and receive custom frames
     19   * In each mode (Master, managed, monitor) card will accept certain types of frames
     20   * Monitor mode accepts all frames (even with CRC errors)
     21   * '''Approach: Use monitor mode to send and receive custom frames (transmit while monitor allowed by chipset)'''
     22   * Hardware automatically stamps packets with sequence numbers
     23   * Use retry = 1 bit to prevent this
     24
     25
     26 * Eliminate automatic ACK and retransmission
     27   * Again use monitor mode
     28   * Set multicast bit in dest MAC address using previous packet injection mechanism
     29   * Need to handle timeouts at sender appropriately
     30
     31 * Eliminate virtual carrier sense (NAV)
     32   * By operating in monitor mode, NAV is disregarded by default
     33
     34 * Controlling Physical Carrier Sense
     35   * Tricky! No easy way to do this
     36   * Introduce desired bias using pre-calibration
     37     * Card periodically calibrated noise floor
     38     * Raise the noise floor and bias the calibration
     39     * Disable periodic calibration after that
     40     * Cards assume “higher/lower noise” based on calibrated level
     41   * May be used for immediate send (TDMA based MAC)
     42   * '''Does not necessarily mean improved throughput, since Xmission may still interfere (in CSMA based MAC)'''
     43
     44 * Controlling BAckoff Duration
     45   * Setting CWMin = CWMax in the driver
     46
     47 * Timer management
     48   * Atheros chipset has a 64 bit 1-μsec timer, can be read/set by driver       
     49   * Tag packet with delay value before inserting in Tx queue
     50   * Reset timer to zero
     51   * Driver spin-waits until the NOW = basetime + delay
     52   * By fixing CW, the additional defer can also be adjusted for