5 th ACM MobiHoc – Tokyo, May 24, 2004 Istituto di Informatica e Telem atica More realistic energy models Open issues: 6/11  TC literature: emphasis is on reducing the transmit power, i.e., the power consumed by the RF amplifier  In practical settings, the transmit power contributes only in part to the power consumption of the wireless transceiver  Example: the sleep:idle:rx:tx min :tx max ratio is – 0.056 : 1 : 1 : 1.044: 1.418 in a CISCO Aironet 802.11 wireless card – 0.440 : 1 : 1.006 :0.872 : 1.114 in a Medusa II sensor node  Open issue: how do these energy models impact the performance of TC protocols? 5 th ACM MobiHoc – Tokyo, May 24, 2004 Istituto di Informatica e Telem atica More accurate analysis with mobility Open issues: 7/11  More work needs to be done to investigate the effect of mobility on TC: – Is mobility beneficial or detrimental? o On one hand, mobility increases message overhead o On the other hand, it balances the node energy consumption o Open issue: which is the overall effect on network lifetime? – Determination of the optimal frequency for reconfiguration o In general, there is a trade off between the frequency of re-execution of the TC protocol and “quality” of the topology o Open issue: given a certain constraint (on the network lifetime, or on the QoS), and given a TC protocol, which is the optimal frequency for reconfiguration? o Preliminary step in this direction: the determination of the information exchange period for the LMST protocol [Li et al.03b] 5 th ACM MobiHoc – Tokyo, May 24, 2004 Istituto di Informatica e Telem atica The effect of multi-hop traffic Open issues: 8/11  Implicit message in the current literature on TC: the sparser communication graph the better  Is this true, if the designer goal is to extend network lifetime?  The effect of multi-hop data traffic must be carefully investigated: – In a very sparse communication graph (e.g., the MST), the average path length between source/destination pairs is relatively long – Given the same quantity of packet delivered, the overall number of messages circulating in the network is relatively larger for relatively sparser graphs – In very sparse graphs, the data traffic might be very unbalanced – Open issue: which is the overall effect on network lifetime? 5 th ACM MobiHoc – Tokyo, May 24, 2004 Istituto di Informatica e Telem atica TC in the protocol stack Open issues: 9/11  Where should TC be positioned in the protocol stack?  No clear answer in the literature Our view: Routing Layer MAC Layer TC Layer 5 th ACM MobiHoc – Tokyo, May 24, 2004 Istituto di Informatica e Telem atica TC and Routing Open issues: 10/11 One possible view: Routing Protocol TC Protocol Trigger route updates Trigger TC execution 5 th ACM MobiHoc – Tokyo, May 24, 2004 Istituto di Informatica e Telem atica TC and MAC Open issues: 11/11 One possible view: TC Protocol MAC Layer Trigger TC execution (new neighbors) Set the power level Remark: a lot of work still to be done on this issue!!! 5 th ACM MobiHoc – Tokyo, May 24, 2004 Istituto di Informatica e Telem atica Towards an implementation of TC Level-based TC: 1/16  To end this tutorial, we present two protocols (CLUSTERPOW and KNeighLev) that explicitly take into account a feature of current wireless transceivers: the transmit power can be set only to relatively few (5-6) levels  For instance: – The CISCO Aironet 350 802.11 wireless card has the following transmit power levels: 1mW, 5mW, 20mW, 30mW, 50mW, 100mW – The transceiver of the Rockwell’s Wins sensor node has the following transmit power levels: 0.12mW, 0.30mW, 0.96mW, 2.51mW, 3.47mW, 13.8mW, 19.1mW, 36.3mW 5 th ACM MobiHoc – Tokyo, May 24, 2004 Istituto di Informatica e Telem atica The CLUSTERPOW protocol Level-based TC: 2/16  The protocol is an extension of the COMPOW protocol of [Narayanaswamy et al.02]  The goal of the CLUSTERPOW [KawadiaKumar03] is to overcome a problem of COMPOW: when the node distribution is not “uniform”, the protocol performs very poorly  Basic idea: every node u in the network maintains one routing table for each power level  The routing table for level i, RT i , is updated by a routing daemon (one for each level), and contains all the nodes that are reachable by u using power at most i 5 th ACM MobiHoc – Tokyo, May 24, 2004 Istituto di Informatica e Telem atica The CLUSTERPOW protocol (2) Level-based TC: 3/16  This way, CLUSTERPOW induces a node clustering: for every node u, several clusters are defined, with the cluster at level i formed by the nodes in RT i  When u needs to send a message to v, it sends the message with power level j, where j is the minimum level such that v ∈ RT j  Intermediate nodes relay the message according to the same rule, until v is reached 5 th ACM MobiHoc – Tokyo, May 24, 2004 Istituto di Informatica e Telem atica The CLUSTERPOW protocol (3) Level-based TC: 4/16 1mW cluster u 100mW 100mW v 10mW 1mW 10mW cluster 100mW cluster n 1 n 2 n 3 . reconfiguration? o Preliminary step in this direction: the determination of the information exchange period for the LMST protocol [Li et al.03b] 5 th ACM MobiHoc – Tokyo, May 24, 2004 Istituto di Informatica e. reconfiguration o In general, there is a trade off between the frequency of re-execution of the TC protocol and “quality” of the topology o Open issue: given a certain constraint (on the network. the RF amplifier  In practical settings, the transmit power contributes only in part to the power consumption of the wireless transceiver  Example: the sleep:idle:rx:tx min :tx max ratio is –