110 likes | 256 Views
Limited Contribution of Self-Management in Chaotic Wireless Deployments. Presented by Karl Deng. Limited Contribution. No. Measurement Devise self-managing algorithms: Channel assignment algorithm Rate adaptation algorithm Power control algorithm. No. No. Yes (but problematic).
E N D
Limited Contribution of Self-Management in Chaotic Wireless Deployments Presented by Karl Deng
Limited Contribution No • Measurement • Devise self-managing algorithms: • Channel assignment algorithm • Rate adaptation algorithm • Power control algorithm No No Yes (but problematic)
Channel assignment algorithm • Effect of Optimal Static Channel Allocation (Section 4.3.1) • Improves performance significantly In combination with careful channel assignment, our power control algorithms attempt to minimize the interference between neighboring APs by ... Where is the channel assignment algorithm?
Rate adaptation algorithm • Bear little relationship with minimizing interference. • Fixed power Rate Selection Algorithms (Section 7.2) • ARF (Auto Rate Fallback) • ERF (Estimated Rate Fallback) • Two algorithms that combine power and rate control to try to minimize interference (Section 7.3) • PARF (Power-controlled Auto Rate Fallback) • PERF (Power-controlled Estimated Rate Fallback)
Rate adaptation algorithm • How do you combine power and rate control? • ARF - Rate selection via in-band probing • PARF - Power selection via in-band probing ERF - Rate selection via SNR measurement PERF - Power selection via SNR measurement The only “combine”: • “If failures continue, the transmit power is raised until the maximum transmit power is reached after which rate fallback begins.” Rate control only happens when using maximum transmit power. Such rate control does not help to reduce interference. Such relation is not “combine”, but only some “similarity”
Rate adaptation algorithm In Section 9 (“Discussion”): LPERF (Load-sensitive, Power-controlled Estimated Rate Fallback) Reduce the power as long as the resulting transmission rate is sufficient to support the actual traffic demands. Conclusion: Using the LPERF technique is very difficult.
Power control algorithm • Problematic: • Socially responsible power control algorithms • Individual access points and clients behave in an altruistic manner • Not simply relying on the altruism of end users • Equipment vendors implement the algorithms (From their point-of-view, reducing interference is beneficial.) • Regulatory mandates • New technology is quickly adopted in chaotic networks. • Many users in chaotic networks do not change factory default settings
Power control algorithm • (Continue) • Backward compatibility • Heterogeneous transmit power levels cause problems Example: “asymmetric carrier sense” (Section 8.1, Figure 13) “When using power control, care must be taken …, an AP-client link should not reduce power so much that it becomes overwhelmed by neighboring uncooperative high power nodes.” But how? • Barrier for future technology (forward compatibility) E.g., signal measurement could be even harder. (Recall why we use CSMA/CA not CSMA/CD.)
Power control algorithm • (Continue) • Backward compatibility • Heterogeneous transmit power levels cause problems Compatibility with old APs • Compatibility with clients PERF requires receiver side cooperation: • SNR is measured at receiver (client) • Receiver feeds back power decrease decision • Client side power control is ignored • Assumption: Amount of upload traffic is small. • What about p2p traffic? • What about multiple clients, variant range?
Limited Contribution ? • Measurement • Devise self-managing algorithms: • Channel assignment algorithm • Rate adaptation algorithm • Power control algorithm No No Yes (but problematic)
Measurement Contribution • Qualitative results are straightforward: • Lower transmission power lower interference • Lower client or AP density lower interference • Optimize channel allocation lower interference • Quantitative results are not representative: • Only one sample – a 20-node topology derived from the Metropolis Wardrive data set.