1. Implications of Power Control in Wireless Networks: A Quantitative Study Ioannis Broustis, Jakob Eriksson, Srikanth V. Krishnamurthy, Michalis Faloutsos Department of Computer Science and Engineering University of California, Riverside {broustis, jeriksson, krish, michalis}@cs.ucr.edu PAM 2007

2. The problem  Increasing the transmission power may: Increase throughput, due to increased signal quality Increase interference levels in neighboring links  Conflict in providing the best network-wide throughput Need to identify the possible interference scenarios For these scenarios, need to examine:  Whether power control can help alleviate interference  Whether fairness is affected by power control  Whether the traditional Virtual Carrier Sensing (RTS/CTS exchange) can co-exist with power control

3. In this work…  We conduct a set of experiments on an indoor wireless testbed We focus on the interference between pairs of links  We identify 3 interference scenarios: Overlapping  Neither power control nor RTS can improve performance Hidden-terminal  Power control essential for fairness  RTS with power control degrades performance! Potentially disjoint  Power control increases performance significantly  RTS with power control, however, results in lower throughput here as well

4. Experimental set-up  Indoor wireless testbed Nodes: 15 Soekris net4826 Wireless: EMP-8602-6G Atheros AR5006 MadWifi driver Linux kernel v2.6, mounted over NFS 802.11a for avoiding external interference

5. Methodology  In our experiments we activate two links at a time We observe the achieved throughput by each of the two links, for different power levels Fully-saturated UDP traffic  iperf measurement tool  30-sec back-to-back 1500-byte packets Power is maintained at a constant level during each 30-sec experiment  After the end of the experiment, we vary the powers and start a new 30-sec session  Power levels: sequentially from 1 dBm to 16 dBm Exhaustive search of different transmission power combinations (16 2 ) We conduct experiments with and without RTS/CTS exchange

6. Types of interference behavior  Overlapping case The two links always contend; power control not helpful  The channel between the two senders is typically better than the channel between sender and receiver

7. Types of interference behavior (ii)  Hidden-terminal case Most of the links under investigation belong to this category Senders cannot sense each others’ tranmissions: PCS fails  This is not a problem if the strength of the desired signal is significantly higher than that of the interfering signal  Competition between signal and interference  Fairness is consistently better along a diagonal (after a certain power level)

8. Types of interference behavior (iii)  Potentially disjoint case Here, power control can greatly improve performance  May enable simultaneous transmissions  15-16: 9 dBm  22-31: 6 dBm If power is different, the throughput is lower

9. Use of RTS/CTS  We repeat the experiments, with RTS enabled  Overlapping case with RTS Worsened, due to the RTS/CTS transmission overhead Without VCS With VCS

10. Use of RTS/CTS (ii)  Hidden-terminal case with RTS Consistently underperforms the plain vanilla CSMA in all considered scenarios!  Overall throughput is considerably lower  Fairness is also affected to a large degree ! Without VCS With VCS

11. Use of RTS/CTS (iii)  Hidden-terminal case with RTS A case where fairness is improved  However, RTS leads to large reduction in overall throughput Without VCS With VCS

12. Use of RTS/CTS (iv)  Potentially disjoint case with RTS RTS/CTS has a negative impact here as well  While the overall throughput is still higher than in the isolated link capacity, it is still lower than the achieved with the PCS alone  The regime of powers for achieving spatial reuse is also reduced More precise power control is now required Without VCS With VCS

13. Conclusion  Proper power control is beneficial in wireless deployments The topology determines if power control will help  Power control improves throughput in potentially disjoint case, and fairness in hidden-terminal case RTS/CTS cannot coexist with power control, while it degrades performance in indoor settings  Power control holds great promise for improving the performance of indoor wireless networks  Potential extensions of this work: Import higher complexity, by activating more than two links at a time Vary the sensing threshold as well

14. Questions? Thank you