Path Capacity in Multirate and Multihop Wireless Networks Wireless Information Networking Group (WING) Path Capacity in Multirate and Multihop Wireless Networks Yuguang “Michael” Fang, Professor University of Florida Research Foundation Professor Wireless Networks Laboratory (WINET) Wireless Information Networking Group (WING) Department of Electrical & Computer Engineering University of Florida In Collaboration with Dr. Hongqiang Zhai Wireless Communications & Networking Department Philips Research North America
Outline Introduction Path capacity and routing metrics Wireless Information Networking Group (WING) Outline Introduction Path capacity and routing metrics Performance evaluation Conclusion
Wireless Multihop Ad Hoc Networks Wireless Information Networking Group (WING) Wireless Multihop Ad Hoc Networks
Interesting Applications Wireless Information Networking Group (WING) Interesting Applications Tactical ad hoc networks First deployment in Iraq campaign (learn from Chip Elliot at BBN ) Public safety Hurricane Katrina (mesh networking) Surveillance and tracking (sensor networking) Many applications demand some kinds of QoS!
Wireless Information Networking Group (WING) Design Challenges Can we support the desired QoS? Particularly E2E QoS? E.g., minimum e2e throughput? Perceived to be a failed proposition! … Wireless links are different! Broadcast in nature, interference, mobility, time-varying channel Low reliability, channel errors, packet loss rate Multiple rates 802.11b: 1, 2, 5.5, and 11 Mbps 802.11a/g: 6, 9, 18, 24, 36, and 54 Mbps Current MANETs have not taken advantage of “plenty of resource”! Need a better understanding of networking capability!
IEEE 802.11 MAC Problems in MANETs Wireless Information Networking Group (WING) IEEE 802.11 MAC Problems in MANETs Low throughput Hidden terminal problem Exposed terminal problem Limitations on NAV procedure Receiver blocking problem Intra-flow contention Inter-flow contention Limitations on the use of single channel
Intra-Flow Contention Wireless Information Networking Group (WING) Intra-Flow Contention Intra-flow contention is the contention from the transmissions of packets at upstream and downstream nodes along the path of the same flow. Packets may continuously accumulate at the first few hops of the path because the transmission at the first few hops encounters less contention than that at subsequent nodes. Nodes separated by fixed length less than the maximum transmission range. Transmission from the first node is interfered with by three subsequent nodes. The number is four for the second one and five for the third node. The first node could inject more packets into the chain than the subsequent nodes could forward, leading to congestion!
Inter-Flow Contention Wireless Information Networking Group (WING) Inter-Flow Contention Inter-flow contention occurs when two or more flows pass through the same region. This region becomes the bottleneck and could make it severer to accumulate packets at some nodes. Node 4 and 11 encounter the most frequent contentions and have little chance to successfully transmit packets to their downstream nodes. Packets will accumulate at and be dropped by node 1, 2, 3, 4, 8, 9, 10, and 11.
Impact of Multiple Rates on Routing Metrics Wireless Information Networking Group (WING) Impact of Multiple Rates on Routing Metrics Be careful with your intuition! E.g., a multi-rate network should use the highest supported data rate for each link A higher rate for a good link where the hop distance is short A lower rate for a poor link where the hop distance is long However, if hop count is used as the routing metric, A multi-rate network may perform worse than a single-rate network. (V. Kawadia and P. R. Kumar, IEEE Wireless Communications, Feb 2005) minimum hop algorithm often choose long hops with low data rates A single rate network may have one best single rate To maintain connectivity To get higher end-to-end throughput A good routing metric is necessary to utilize the advantages of multiple rates. 11Mbps
Routing Metrics and Path Capacity Wireless Information Networking Group (WING) Routing Metrics and Path Capacity Routing metrics (some arising from cross-layer design) Hop count Rate: data rates over wireless links ETX: expected transmission count (Mobicom’03) ETD: end-to-end transmission delay WCETT: weighted cumulative expected transmission time (Mobicom’04) MTM: medium transmission time metric (MONET’06) BDiP: bandwidth distance product, or link rate times hop distance (Infocom’06) It has been shown to have 27% and 12% higher throughput than routing metrics hop count and rate in a random chain topology, respectively. Path capacity The maximum throughput of the path (e2e QoS)
Example 2Mbps 54Mbps Path 1 2 Throughput (Mbps) 2/3 54/12=4.5 3 4 2/3 54/12=4.5 S1 to D1 Path 1: hop count, ETX Path 2: rate, ETD, BDiP S2 to D2 Path 3: hop count, ETX, ETD Path 4: rate, BDiP To determine which routing metric is the best among them or even find a better one, we need to have a comprehensive understanding of factors which affect the path capacity. Routing metrics do have significant impacts on the path capacity
Path Capacity and Routing Metrics Wireless Information Networking Group (WING) Path Capacity and Routing Metrics Impacts of multirate capability on path selection Extended link conflict graph and a new routing metric Path capacity Using routing metrics in path selection
Multiple Rates over Wireless Links Wireless Information Networking Group (WING) Multiple Rates over Wireless Links Requirements for correct reception Receiver sensitivity Received signal strength is higher than a certain threshold SINR (signal-to-interference-plus-noise ratio) SINR has to be larger than a certain threshold An IEEE 802.11a system
Multiple Rates over Wireless Links Wireless Information Networking Group (WING) Multiple Rates over Wireless Links Tradeoff between rate and distance Carrier sensing range and spatial reuse Protocol overhead for different rates Effective data rate rd
Extended Link Conflict Graph Wireless Information Networking Group (WING) Extended Link Conflict Graph Conflict between links wij describes the impact of link i on link j. (Maximum) Independent set S All links in S can successfully transmit at the same time For all j in S, if (Maximum) Interference clique C Any two links in C cannot successfully transmit at the same time
Clique Transmission Time and Path Capacity Wireless Information Networking Group (WING) Clique Transmission Time and Path Capacity Interference clique transmission time (CTT) TC Tl: transmission time for a packet over link l S: the set of all the maximum interference cliques Lp: packet size CP: path capacity
A New Routing Metric: CTT Wireless Information Networking Group (WING) A New Routing Metric: CTT One optimum scheduling Time slot 1 2 3 4 5 6 7 Scheduled links Upper bound of path capacity: Local interference clique transmission time (LCTT): Local interference clique: an interference clique with a series of adjacent links along the path New routing metrics: CTT and LCTT
Path Capacity Problem Notation P: the set of links in a given path Wireless Information Networking Group (WING) Path Capacity Problem Notation P: the set of links in a given path Eα (1≤α ≤M): independent sets, Eα is a subset of P λα: time share scheduled to Eα Rα: a row vector of effective data rates for Eα f: demand vector , feasible: CP: path capacity
Data Rates over Links For each independent set Eα Wireless Information Networking Group (WING) Data Rates over Links For each independent set Eα Calculate the received power level Pr and SINR Select the highest rate rd satisfying the requirements of receiver sensitivity and SINR Consider the packet error rate per r=rd × (1-per) We assume each device uses a predefined transmission power
Optimization Problem for Path Capacity Wireless Information Networking Group (WING) Optimization Problem for Path Capacity Linear programming problem
A Feasible Scheduling from the Solution Wireless Information Networking Group (WING) A Feasible Scheduling from the Solution Schedule S S={λα, 1≤α ≤M} Time is divided into slots of duration τ seconds τ is partitioned into M subslots, each lasts λατ seconds Links in Eα are scheduled to transmit in the αth subslot In each slot τ, throughput fe over link e is Path capacity CP = fe
Extension to Multiple Paths Wireless Information Networking Group (WING) Extension to Multiple Paths Notation Pk (1≤ k ≤K): the set of links in path k fk: end-to-end throughput of path k P: I(Pk): a row indicator vector Maximize aggregate end-to-end throughput
Optimum Path in Multirate Wireless Networks From source s to destination t Max-flow problem Multiple paths Linear programming Unicast and single-path routing Mixed integer-linear programming
Routing Metrics and Algorithms Wireless Information Networking Group (WING) Routing Metrics and Algorithms Routing algorithms min-hop: the smallest hop count or ETX min-delay: the shortest (expected) end-to-end transmission delay max-rate/maxmin-rate: the widest bottleneck link rate max-BDiP/maxmmin-BDiP: the widest bottleneck link BDiP min-CTT/minmax-CTT: the smallest bottleneck CTT min-LCTT/minmax-LCTT: the smallest bottleneck LCTT Generalized Bellman-Ford algorithm A routing algorithm for additive routing metrics min-hop, min-delay Easy to be generalized to a widest routing algorithm max-rate, max-BDiP, min-CTT, min-LCTT
Performance Evaluation Wireless Information Networking Group (WING) Performance Evaluation Simulation setup N nodes: randomly distributed in the network Channel rates: 54, 18, 11, 6, 1 Mbps 802.11b rates: 11, 1 Mbps 802.11g rates: 54, 18, 11, 6 Mbps Transmission radii: 76, 183, 304, 196, 610 m Interference range: 900 m Packet size: 1000 bytes Two-way handshake: DATA/ACK Source: the node nearest to the upper left corner Destinations: N-1 choices Performance metric: path capacity
Comparison with Optimal Routing Wireless Information Networking Group (WING) Comparison with Optimal Routing N = 25 nodes, 200m X 2500m
Comparison of Six Metrics in a Larger Topology Wireless Information Networking Group (WING) Comparison of Six Metrics in a Larger Topology N = 400 nodes, 1500m X 300m
Path Length and Computational Time Wireless Information Networking Group (WING) Path Length and Computational Time
Multi-Rate Network vs. Single-Rate Network Wireless Information Networking Group (WING) Multi-Rate Network vs. Single-Rate Network
How Can We Utilize the “Available” Resources? Wireless Information Networking Group (WING) How Can We Utilize the “Available” Resources? Current MANETs Single-channel, single rate, single power level…. “Available” resources Multiple rates Multiple channels Multiple power levels Multiple antennas Multiple radios (cognitive radios) Need to find efficient ways to utilize them!
Current Research Path capacity analysis Wireless Information Networking Group (WING) Current Research Path capacity analysis Multi-channel interference analysis Conduct experiments on mesh network testbed Theoretical modeling Flow/link scheduling on mesh networks Channel assignment for multi-radio and multi-channel systems (diversity) Directional wireless networks Connectivity and capacity Flow/link scheduling
Wireless Information Networking Group (WING) Conclusion Routing metrics have significant impacts on path capacity in multirae and multihop wireless networks An upper bound of path capacity is found A new routing metrics is found Interference clique transmission time (CTT) Finds paths with up to 10% higher throughput than e2e min-delay, and even much higher than others An optimization problem is formulated to find the path capacity in multirae and multihop wireless networks Multi-rate networks are better than single-rate networks However, we should be careful to choose an appropriate routing metric