Upload
camron-owen
View
214
Download
0
Tags:
Embed Size (px)
Citation preview
A High-Throughput Path Metric for Multi-Hop Wireless Routing
(ETX)
Douglas S. J. De Couto, Daniel Aguayo, John Bicket, Robert Morris
Minimum Hop Count
• Assumes links either work or don’t work
• Minimize hop count -> Maximize the distance traveled by each hop– Minimizes signal strength -> Maximizes the loss ratio– Uses a higher Tx power -> Interference
• Arbitrarily chooses among same length paths
Is there a better metric?
• Cut-off threshold– Disconnected network
• Product of link delivery ratio along path– Does not account for inter-hop interference
• Bottleneck link (highest-loss-ratio link)– Same as above
• End-to-end delay– Depends on interface queue lengths
ETX metricDesign goals
• Find high throughput paths
• Account for lossy links
• Account for asymmetric links
• Account for inter-link interference
• Independent of network load (don’t incorporate congestion)
ETX metricDefinition
• ETX – predicted # of data tx required to successfully send a packet over link/path including retransmissions
• ETX (link) = 1 / df x dr• ETX (path) = ∑ ETX(link)• ETX (link) measured by broadcasting
periodic probe packets• Reverse-delivery ratio piggybacked in
forward probe packets
ETX caveats
• ETX estimates are based on measurements of a single link probe size (134 bytes) i.e. Probe size ≠ Data/Ack size– Under-estimates data loss ratios, over-
estimates ACK loss ratios
• ETX assumes all links run at one bit-rate• “Broadcast” has lower priority.• ETX assumes that radios have a fixed
transmit power level.
Take aways
• Pros– ETX performs better or comparable to Hop Count
Metric• Accounts for bi-directional loss rates
– Can easily be incorporated into routing protocols as detailed experiments on a real test bed show it
• Cons– May not be best metric for all networks
• Mobility, Power-limited, Adaptive Rate (multi-rate)– Predications of loss ratios not always accurate as
seen in experiments sometimes.– Experiments (30 sec transfer of small packets) may
not complement real-world scenarios
Comparison of Routing Metrics for Static Multi-Hop Wireless Networks
Richard Draves, Jitendra Padhye and Brian Zill
Routing in Multi-hop Wireless Networks
• Mobile Networks– Minimum-hop routing (“shortest path”)– DSR, AODV, TORA (covered previously)
• Static Networks– HOP based routing chooses short but lossy wireless
links thereby reducing throughput– Taking more hops on better quality links can improve
throughput
Contribution of the paper
• Design and Implementation of a routing protocol based on notion of link quality– LQSR (Link Quality Source Routing)
• Experimental comparison of three link quality metrics– Per-hop Round Trip Time (RTT)– Per-hop Packet Pair Delay (PktPair)– Expected Transmission (ETX)
Summary of Results
• ETX Provides best performance for static wireless network
• Performance of RTT and PktPair suffer due to self-interference
• HOP performs well over ETX in mobile wireless networks
LQSR Architecture
• Implemented in a shim layer between Layer 2 and 3.
• The shim layer acts as a virtual Ethernet adapter– Virtual Ethernet addresses– Multiplexes heterogeneous
physical links
• Advantages:– Supports multiple link
technologies– Supports IPv4, IPv6 etc unmodified
– Preserves the link abstraction– Can support any routing
protocol
• Architecture:
• Header Format:
Ethernet 802.11 802.16
Mesh connectivity Layer with LQSR
IPv4 IPv6 IPX
Ethernet MCL
Payload:TCP/IP,
ARP,IPv6…
LQSR
• Source Routed, link state protocol– Derived from DSR
• Each node measures quality of its link to its neighbor
• The info regarding link quality propagates through the mesh
• Source selects route with best cumulative metric
• Packets are source-routed using this route
Link Quality Metrics
• Per-hop Round Trip Time (RTT)– Routing based on minimizing total RTT
• Per-hop Packet Pair Delay (PktPair)– Routing based on minimizing PktPair
• Expected Transmission (ETX)– Routing based on maximizing ETX
• Minimum hop routing (HOP)– Routing based on minimizing HOP
Metric 1: Per-hop RTT
• Advantages– Easy to implement– Accounts for link load and bandwidth– Also accounts for link loss rate
• 802.11 retransmits lost packets up to 7 times• Lossy links will have higher RTT
• Disadvantages– Expensive – Self-interference due to queuing
Metric 2: Per-hop Packet-Pair
• Advantages– Self-interference due to queuing is not a
problem– Implicitly takes load, bandwidth and loss rate
into account
• Disadvantages– More expensive than RTT
Metric 3: Expected Transmissions• Advantages
– Low overhead– Explicitly takes loss rate into account
• Disadvantages– Loss rate of broadcast probe packets is not the
same as loss rate of data packets• Probe packets are smaller than data packets• Broadcast packets are sent at lower data rate
– Does not take data rate or link load into account
Impact of TCP flows (one at a time)
• ETX performs better by avoiding low-throughput paths• RTT suffers heavily from self-interference
Experimental results for mobile wireless networks
• Shortest path routing is best in mobile scenarios– Why?
Contributions
• This paper contributes the first complete design and implementation of a link/network-layer diversity routing technique that uses standard radio hardware.
• It demonstrates a substantial throughput improvement and provides insight into the sources of that improvement.
Why ExOR promises high throughput? - 1
S
S
SD
S
S
25%
25%
25%
25%
100%
100%
100%
100%
Reception at different node is independent, no interference Traditional Routing: 1/ 0.25 + 1 = 5tx ExOR: 1/ (1-(1-0.25) ) + 1 = 2.5tx4
Why ExOR promises high throughput? - 2
N1 N3 N5 N7N6N2 N4 N8S D
Gradual falloff of probability with distance (80%, 40%, 20%..) Lucky longer path can reduce transmission count Shorter path ensures some forward progress
Traditional Path
Design Challenges
• The nodes must agree on which subset of them received each packet – Protocol ?
• A metric to measure the probable cost of moving packet from any node to destination
• Choosing most useful participants
• Avoid simultaneous transmission to minimize collisions
Evaluation Setup
• 65 node pairs from a physical layout of 38 Roofnet nodes participated
• No ExOR + Traditional routing, hence the ExOR run was asked to transfer 10% more.
• One hop at a time for fair comparison in traditional routing.
Take aways
• Pros– ExOR achieves 2x to 4x throughput improvement for
more distant pairs– ExOR implemented on Roofnet and evaluated in
detail– Exploits radio properties, instead of hiding them– Does not require changes in the MAC layer
• Cons– Not scalable to large network as traditional routing– Overhead in packet header (batch info)– Batches affect the TCP performance– What if not enough packets to make the batch?
Extra –related work• Opportunistic Channel Protocols
– Use channel reservation to avoid collisions– Cons: require channel stability, use signal strength to predict reception,
does not use intermediate nodes to relay• Opportunistic Forwarding
– Select forwarding nodes based on channel conditions– Cons: use channel measurements or distance to predict the delivery
success rate• Multiple Path Routing
– Maintain multiple routes to use as alternative routes or split the traffic among them
– Cons: Ensure the paths are disjoint, need to identify specific paths in advance
• Cooperative Diversity Routing– Exploit nearby nodes which overhear the transmission– Cons: duplicate transmissions
A Rate-Adaptive MAC Protocol for Multi-Hop Wireless Networks
By Gavin Holland, Nitin Vaidya and Paramvir Bahl
Introduction• Rate Adaption
– Rate adaption is the process of dynamically switching data rates to match the channel conditions. There are two aspects to rate adaption:
– Channel quality estimation• By Sender• By receiver-> RBAR(Receiver Based Auto rate)
– Rate Selection• By Sender ->ARF(Auto rate Fallback)• By Receiver -> RBAR(Receiver Based Auto rate)
• Why receiver based rate adaption– The goal of rate adaption is to provide optimum throughput.
• Rate selection can be improved by proving more timely and more complete channel quality.
• Channel quality information is best acquired at the receiver.
RBAR modified DCF Protocol
DRTS: Reservation time (IEEE 802.11)
DCTS: Reservation time
DRSH: Final reservation Time
DCF: To coordinate the transfer of data packet.
NAV: To announce the duration of packet.
DRTS: Tentative reservation time (RBAR)
RBAR EVENT FLOW• S choose a data rate r1, using some heuristic,
and sends r1 and the size of the data packet n in the RTS to R.
• A, overhearing the RTS, uses r1 and n to calculate the duration of the reservation, marking it as tentative.
• R, having received the RTS, uses some channel quality estimation and rate selection technique to select the best rate r2 for the channel conditions, and sends r2 and n in the CTS to S.
• B, overhearing the CTS, calculates the reservation using r2 and n.
• S responds to the CTS by placing r2 into the header of the data packet and transmitting the packet at the selected rate. If r1≠r2, S uses a unique header signaling the rate change.
• A, overhearing the data packet, looks for the unique header. If it exists, it recalculates the reservation to replace the tentative reservation it calculated earlier.
S R Br1, nr1, n
r2, n
r2, n r2, n
r2, n
ACK
A
RBAR MAC Header
Framl
control
Duration Dest.
Address
Source
Address
BSSID Sequnce
control
Body FCS
Framl
control
Duration Dest.
Address
Source
Address
BSSID Sequnce
control
HCS Body FCS
IEEE 802.11 MAC Header
RBAR MAC Header
RBAR Reservation SubHeader
RBAR RTS/CTS Implementation
Frame
control
Duration Dest.
Address
Source
Address
FCS
IEEE 802.11 RTSRBAR RTS
Rate &
Length
Frame
control
Duration Dest.
Address
FCS
IEEE 802.11 CTSRBAR CTS
Rate &
Length
In RBAR, instead of carrying the duration of the reservation , the packets carry the modulation rate and the size of the data packet.
If there is rate mismatch between sender and receiver DRTS refer to as tentative reservation.
Final reservations are confirmed by the presence or absence of Reservation SubHeader (RSH).
RBAR PLCP Header
Sync SFD Signal Service Length CRC
802.11 PLCP header
Data
Rate
RSH
Rate
RBAR PLCP header
In standard 802.11, the PLCP header contains an 8 bit signal field.
In RBAR, the PLCP header has been divided into two 4 bit rate subfields.
Thus, the PLCP transmission protocol is modified as follows: when the MAC passes a packet down to the physical layer, it specifies two rates, one for the subheader and one for the remainder of the packet.