On Notworking CS 268 Papers

On TCP and Link-Layer Wireless Notworking

September 30, 2008 · 1 Comment

Bharghavan et. al. first investigates media access protocols for “new” (circa 1994) mobile computing devices as applied to Xerox PARC’s radio technology. They evaluate these algorithms based on high throughput and fairness deferring to the latter when necessary. Based on the result of this evaluation, they introduce MACAW with an additional ACK to RTS-CTS-DATA (enhancement from Karn’s and Biba’s work) and a different backoff algorithm.  In MACAW, the backoff algorithm includes in the packet header a field for the backoff counter. This enables a station, upon every successful transmissions in the cell,  to have the same value for the counter as everyone else. They have showed that this slight modification results in fairness. Also to prevent oscillations (as seen in the original MACA design), they adopted a less aggressive adjustment algorithm called multiplicative increase and linear decrease (MILD). Another enhancement introduced in MACAW is the use of ACKs. The use of ACKs is to introduce link layer recovery rather than solely relying on the transport layer for this. This enables faster recovery at the link layer with minimal overhead. In the context of wireless networking, an interesting discussion point could be the concept of service differentiation in such environments. While MACAW improves on the link layer, how and where should QoS be provided in this more constrained environment can be discussed.

The next paper looks at wireless networking at the transport point of view. They first classify transport (TCP) schemes into three namely: end2end, link-layer protocols and split-connection protocols and then evaluate them based on throughput and goodput. Interestingly, a link-layer protocol that is TCP-aware  (as compared to TCP protocol that is aware of the link-layer protocol) is reported to have very good performance. Maybe a throwback and in relation to the first few papers in class, for this particular scenario, it is much better to push some intelligence to the lower layers to possibly improve efficiency and utilization.

Categories: Readings
Tagged:

1 response so far ↓

  • Randy Katz // September 30, 2008 at 8:33 pm

    Cute blog site title … took me a while to get it!

    The latter was VERY controversial at the time … many claimed it violated the end-to-end argument. What do you think?

Leave a Comment