Presentation on theme: "Distributed Assignment of Encoded MAC Addresses in Sensor Networks By Curt Schcurgers Gautam Kulkarni Mani Srivastava Presented By Charuka Silva."— Presentation transcript:
Distributed Assignment of Encoded MAC Addresses in Sensor Networks By Curt Schcurgers Gautam Kulkarni Mani Srivastava Presented By Charuka Silva
Every bit transmitted takes a bite out of the node’s lifetime
Majority of wide scale traffic consists of few bytes. That include all network and application layer Ids MAC address in shared medium is major overhead when chosen network-wide unique
Propose MAC Address Reduces the size of the MAC address –In the factor of 3-6 –Energy saving for data transmission Propose distributed addressing scheme –Based on spatial reuse and encoded address representation Scales well with the network size –Varying or unknown network sizes and densities.
Background Some attributes, location recognize the destination Routing protocols directly uses those Unique network address is rarely used. Nodes process data locally, acquired from neighbors and forward aggregation Network of nodes require 14 bits for address
Why Spatial? Multi-hop transmission Only required to identify intended receiver at each hop Functionality restricted to direct neighborhood. Spatial re-use reduces the number of distinct addresses Thus need small number of bits
Why Distributed? Centralized algorithm may too energy costly Network topology is not perfectly constant Address assignment algorithm should quickly acquire correct topology. Centralized algorithm may need costly global update.
B and C are bi-neighbors –Establishes their bi-directional neighbors thru discovery protocols D and E out-neighbors G and F in-neighbors –A has no direct means to gather information on addresses of G and F –Not even known their existence
Condition 1 When the normal-mode data communication of a node A is restricted to its bi-neighbors, a valid assignment of addresses is such that all bi-neighbors have distinct addresses and that all in-neighbors have addresses different from those of the bi- neighbors, for any node A.
First: –B and C have different addresses since they are A’s intended receivers Second: –F and G have addresses different from B and C –F and G themselves need not to be distinct
Case 1: –One cycle after node boot up, selects address satisfying condition 1 –Constraints include its on-hop, two-hop neighbors information –At time out invalidate the entry for neighbor, if not heard for a while –Periodically broadcast packets broadcast_pkt, containing its own address and those of its neighbors.
Case 2: –Can make explicit request() to broadcast broadcast_pkt from neighbors, when entering or on topology change. –If conflict is detected from broadcast information, orders one of the bi-neighbors to change Case 3: –On conflict packet receive, node choose new address and broadcast.
Address Representation Codeword is used instead of address Use Huffman coding – Gives the minimum ave. codeword length In data packets include the codeword instead address
Address Selection Use lower addresses more Incremental shows choosing lowest non- conflicting addresses Based on the Incremental curve, Huffman code is derived Here, for 6 bit fixed representation needs 4.41 bits for codeword A range uses the same address. Can choose randomly one from the possible range
Frequency Vs. Adress
Any encoding, optimal or not, is always able to represent all possible addresses In fixed address size, if the field is estimated to be small, serious overflow may occur.