A high-throughput path metric for multi-hop wireless routing
Summary (5 min read)
1. Introduction
- Much of the recent work in ad hoc routing protocols for wireless networks [25, 15, 26] has focused on coping with mobile nodes, rapidly changing topologies, and scalability.
- Retransmission does not make lossy links desirable for use in paths: the retransmissions reduce path throughput and interfere with other traffic.
- The metric predicts the number of retransmissions required using per-link measurements of packet loss ratios in both directions of each wireless link.
- This paper makes the following main contributions.
- Third, it describes a set of detailed design changes to the DSDV [25] and DSR [15] protocols (to which ETX is an extension), that enable them to more accurately choose routes with the best metric.
2. Performance of Minimum-Hop-Count Routing
- This section shows that minimum hop-count routing typically finds routes with significantly lower throughput than the best available.
- The evidence comes from measurements of DSDV on a testbed network.
- The authors explain why minimum hop-count does poorly by looking at the distribution of route throughputs and link loss ratios.
2.1 Experimental Test-Bed
- All the data in this paper are the result of measurements taken on a 29-node wireless test-bed.
- Each node consists of a stationary Linux PC with a Cisco/Aironet 340 PCI 802.11b card and an omnidirectional 2.2 dBi dipole antenna (a "rubber duck").
- RTS/CTS is turned off, and the cards are set to "ad hoc" (IBSS, DCF) mode.
- An 802.11b ACK packet takes 304 microseconds to transmit, the inter-frame gap is 60 microseconds, and the minimum expected mandatory back-off time is 310 microseconds, resulting in a total time of 2,218 microseconds per data packet.
- The DSDV implementation used in this paper is new, with modifications described in Section 4.
2.2 Path Throughputs
- Figure 2 compares the throughput of routes found with a minimum hop-count metric to the throughput of the best routes that could be found.
- Each curve shows the throughput CDF (in packets per second) for 100 node pairs; the pairs are randomly selected from the 29×28 = 812 total ordered pairs in the test-bed.
- Packets were only sent between one pair at a time.
- Potential best paths were identified by running an off-line routing algorithm, using as input measurements of per-link loss ratios, similar to those in Section 2. and with a penalty to reflect the reduction in throughput caused by interference between successive hops of multi-hop paths.
- The minimum hop-count routes are slow because they include links with high loss ratios, which cause bandwidth to be consumed by retransmissions.
2.3 Distribution of Path Throughputs
- Figure 3 illustrates a typical case in which minimum hop-count routing would not favor the highest-throughput route.
- The routes are the eight best which were tested in the experiments described above.
- The graph shows that the shortest path, a two-hop route through node 19, does not yield the highest throughput.
- The best route is three hops long, but there are a number of available three-hop routes which provide widely varying performance.
- A routing protocol that selects randomly from the shortest hopcount routes is unlikely to make the best choice, particularly as the network grows and the number of possible paths between a given pair increases.
2.4 Distribution of Link Loss Ratios
- Figure 4 helps explain why high-throughput paths are difficult to find.
- The authors use the term "ratio" instead of "rate" to avoid confusion with throughput delivery rates, which are expressed in packets per second.
- Third, many links have asymmetric delivery ratios.
- Of the 406 node pairs in Figure 4a (1 mW), there are 124 with links which delivered packets in at least one direction.
- Because 802.11b uses link-level ACKs to confirm delivery, both directions of a link must work well in order to avoid retransmissions.
3. ETX Metric Design
- This section describes the design of the ETX metric.
- A number of superficially attractive metrics are not suitable.
- Using the product of the per-link delivery ratios as the path metric, in an attempt to maximize the end-to-end delivery probability, fails to account for inter-hop interference; this metric would view a perfect two-hop route as better than a one-hop route with a 10% loss ratio, when in fact the latter would have almost twice the throughput.
- ETX, however, addresses each of these concerns.
- The authors goal is to design a metric that is independent of network load; load balancing can be performed with separate algorithms that use the information provided by ETX.
3.1 The Metric
- The ETX of a link is the predicted number of data transmissions required to send a packet over that link, including retransmissions.
- The ETX of a link is calculated using the forward and reverse delivery ratios of the link.
- The forward delivery ratio, df , is the measured probability that a data packet successfully arrives at the recipient; the reverse delivery ratio, dr, is the probability that the ACK packet is successfully received.
- ETX can use precise link loss ratio measurements to make fine-grained decisions between routes.
- Because the probes are broadcast, 802.11b does not acknowledge or retransmit them.
3.2 Discussion
- ETX makes at least two assumptions about the link layer.
- First, ETX only makes sense for networks with link-layer retransmission, such as 802.11b.
- As a result, a node might never be able to send its probes, causing its neighbors to believe that the reverse delivery ratio had become zero.
- If the highest-throughput path has three or fewer hops, ETX is likely to choose it: the throughput of such paths is determined by the total number of transmissions, since all of the hops interfere with each other [20] .
- There is a tradeoff between the accuracy of link measurements and the protocol's responsiveness to mobility.
4. Implementation
- The routing system in which ETX is implemented has four main parts: the Click toolkit [19] , and Click-based implementations of DSDV, DSR, and the ETX link measurement algorithms.
- The implementations can run in user-space, but running in the kernel allows use of the priority queuing described below, as well as easy access to transmission failure notification from the 802.11 MAC layer.
- The DSDV protocol is implemented following the description by Perkins and Bhagwat [25] , with ambiguities resolved by consulting Broch et al. [5] and the Rice/CMU implementation in the ns simulator [1, 27] .
- The DSR implementation follows the IETF Internet-Draft, version 9 [16] .
4.1 Operation of DSDV
- DSDV is a distance-vector protocol, which uses sequence numbers to ensure freshness, and a settling time mechanism to avoid unnecessarily propagating routes with inferior metrics.
- The node copies the sequence numbers for the other entries in the full dump from its routing table.
- Each route entry has an associated weighted settling time (WST).
- Whenever a node replaces a route entry with a newly received entry, it propagates the new route to its neighbors by sending a triggered update which contains only the changed information.
- In addition, regardless of each route entry's WST, triggered updates are sent at no more than a maximum specified rate.
4.2 Changes to DSDV
- The first change the authors made affects how the WST is used.
- The authors implementation still generates broken-route messages when routing table entries time out, but this rarely occurs during the experiments presented in this paper.
- Triggered updates contain only the changed routes, and full dumps are only sent at the full dump period.
- DSDV will use the best route with the previous sequence number until the WST has expired and the best route with the new sequence number has likely been heard.
- An ETX node broadcasts one probe per second, and remembers probes received from neighbors over the last ten seconds.
4.3 DSR Implementation
- The authors DSR implementation follows revision 9 of the IETF Internet-Draft specification [16] , following the requirements for networks which require bidirectional links to send unicast data.
- This section reviews DSR's basic operation as described in the draft, and describes their modifications to support ETX.
- The destination issues a route reply in response to every forwarded request it receives.
- If a transmission failure occurs when forwarding a route reply, the neighbor to which the node was trying to forward the reply is added to the blacklist, with an entry of unidirectionality probable.
- Entries are removed from the blacklist when the link is determined to be bidirectional, e.g. by a successful unicast transmission.
4.4 Router Configuration Details
- To mitigate this problem, the implementation maintains separate Click queues for data packets, protocol packets, and link probes.
- These queues all drain into a single queue in the wireless adapter's memory, managed by the driver, which has a capacity of three packets.
- The DSDV implementation looks up a packet's destination in the routing table after dequeuing the packet from the data queue, and just before handing the packet to the 802.11b card.
- This technique depends on the fact that the nodes have only one wireless interface.
- This ensures that the node experiencing the transmission failure does not spend additional time and spectrum retransmitting more packets over the broken hop.
5. Evaluation
- This section presents experimental results that show that ETX often finds higher-throughput paths than minimum hop-count, particularly between distant nodes.
- Unless otherwise stated, the experimental setup is as follows.
- At the start of each experiment, the routing software is reset (all tables are cleared), then the routing protocol and/or ETX probe algorithm is allowed to run long enough to stabilize (typically 90 seconds).
- Each graph below is labeled with the run from which it came.
- In DSR experiments with ETX, the source waits an additional 15 seconds before initiating the route request, to give the nodes time to accumulate link measurements.
5.1 Metric Performance with DSDV
- Figure 6 compares the throughput CDFs of paths found by DSDV using ETX and minimum hop-count, between 100 randomly chosen node pairs.
- The right half shows node pairs that could communicate directly, with loss ratios less than about 50% (i.e. with throughput greater than the maximum possible two-hop throughput of 225 packets per second).
- In these cases the minimum hop-count metric finds the one-hop route, which is the best route, and there is no opportunity for ETX to perform better.
- The points for two pairs in Figure 7 lie well below the y = x line; this is because of variations in link quality between the ETX and minimum hop-count tests for those pairs.
- When nodes send at the higher transmit power they have more links, as shown in Figure 4 .
5.1.1 Impact of Asymmetry
- Some fraction of ETX's gains comes from avoiding extremely asymmetric links.
- The problem of routing when there are asymmetric links has been addressed in previous work by Lundgren et al. [22] and by Chin et al. [8] .
- These authors propose a link handshaking scheme to detect and avoid asymmetric links.
- A node bootstraps the handshake by advertising provisional route entries, which in- dicate that the node has 'seen' another node, but not yet accepted routes from it.
- The authors implemented the handshaking scheme for DSDV with the minimum hop-count metric.
5.1.2 Effects of DSDV Modifications
- Section 4.2 described modifications to DSDV designed to increase its responsiveness to metrics.
- The delay-use modification causes DSDV to delay using a newly received route until it is permitted to advertise the route (i.e. 2×WST has passed).
- Figure 11 shows that the delay-use modification improves the performance of DSDV with ETX.
5.2 Metric Performance with DSR
- This section evaluates the performance of the DSR routing protocol with the ETX metric.
- As described in Section 4.3, DSR uses link-layer transmission failure feedback to help it avoid bad routes.
- To isolate the effects of using ETX with DSR, the authors evaluated DSR performance both with and without link-layer feedback enabled.
- This is consistent with the DSDV results in Section 5.1.
- ETX helps the source picks an initial route with high throughput.
5.3 Accuracy of Link Measurements
- For all the experiments in this paper, ETX used 134-byte packets to estimate link loss ratios.
- This is likely to lead to inaccurate metrics for packet sizes other than the size used for measurement probes.
- Furthermore, ETX uses the single packet size measurements to estimate the delivery ratio of link-layer ACK packets.
- For each experiment, the two nodes first take turns sending broadcast probes with a 104byte data payload.
- The y values on the graph represent the average actual transmission counts.
7. Conclusions
- This paper introduces a new metric for multi-hop wireless networks, called ETX.
- Route selection using ETX accounts for link loss ratios, the asymmetry of the loss ratios in the two directions of each link, and the reduction of throughput due to interference among the successive hops of a route.
- ACKs; its handling of networks with links that run at a variety of bit-rates; and the robustness of ETX probes when competing with high levels of data traffic.
- The protocol implementations described in this paper are available at http://www.pdos.lcs.mit.edu/grid.
Did you find this useful? Give us your feedback
Citations
28,685 citations
4,205 citations
2,633 citations
Cites background or methods from "A high-throughput path metric for m..."
...Although the ETX metric performs better than shortestpath routing [ 15 , 16], it will not necessarily select good routes in the two scenarios discussed earlier....
[...]
...The values of pf and pr can be approximated by using the broadcast packet technique described by De Couto et al. [ 15 ]....
[...]
...Following the methodology of [ 15 ], we selected 100 of these pairs at random....
[...]
...In this section, we will focus on the ETX (Expected Transmission Count) routing metric proposed by De Couto et al. [ 15 ]....
[...]
...This assumption is usually true for short paths, but the assumption is somewhat pessimistic for longer paths [ 15 ]....
[...]
2,190 citations
Cites background from "A high-throughput path metric for m..."
...The proposed solutions range from designing better routing metrics [ 31 ]–[33] to tweaking the TCP protocol [34], and include improved routing and MAC protocols [35], [36]....
[...]
...For e.g., the ETX metric [ 31 ] periodically computes the delivery probabilities and assigns each link a weight equal to 1/(delivery probability)....
[...]
1,575 citations
Cites background or methods from "A high-throughput path metric for m..."
...The graph indicates the distance that each transmission traveled, as measured in di.erence in ETX metric to N24 between the transmitter and receiver....
[...]
...The traditional route is chosen using the ETX metric, which has been shown to .nd the best routes [4, 5] when the link loss measurements are accurate....
[...]
...Each node s ETX value is the sum of the link ETX values along the lowest-ETX path to ETX = 2.28 ETX = 1.17 Figure 5: Estimated transmission count (ETX) to node E from each node in the sample network from Figure 4....
[...]
...Multi-hop wireless networks typically use routing techniques similar to those in wired networks [15, 16, 9, 4, 5]....
[...]
...7000 Packet Transmissions 1500 1000 500 0 Figure 12: Distance traveled towards N24 in ETX space by each transmission....
[...]
References
28,685 citations
"A high-throughput path metric for m..." refers background in this paper
...11b MAC can be unfair under high load [3]....
[...]
12,497 citations
11,412 citations
"A high-throughput path metric for m..." refers background in this paper
...With variable power radios, it might be preferable to maximize hop-count, thereby decreasing interference and minimizing the energy used by each packet [29, 12, 18]....
[...]
...Existing systems exploit this idea, often with a focus on minimizing the energy consumption required to successfully deliver data [12, 18, 29]....
[...]
11,360 citations
"A high-throughput path metric for m..." refers background in this paper
...Much of the recent work in ad hoc routing protocols for wireless networks [25, 15, 26] has focused on coping with mobile nodes, rapidly changing topologies, and scalability....
[...]
6,877 citations
Related Papers (5)
Frequently Asked Questions (16)
Q2. What have the authors stated for future works in "A high-throughput path metric for multi-hop wireless routing" ?
Several aspects of ETX could be improved in the future: its predictions of loss ratios for different packet sizes, particularly for 802.
Q3. What is the way to minimize the distance traveled by each hop?
Minimizing the hop-count maximizes the distance traveled by each hop, which is likely to minimize signal strength and maximize the loss ratio.
Q4. What does the node do in the event of a transmission failure?
The nodes do not perform packet salvage, in which forwarding nodes, in the event of a transmission failure or received route error, attempt to find alternate routes for queued packets.
Q5. What is the effect of a transmission failure on the neighbor?
If a transmission failure occurs when forwarding a route reply, the neighbor to which the node was trying to forward the reply is added to the blacklist, with an entry of unidirectionality probable.
Q6. What is the effect of delay-use modification on DSDV?
The delay-use modification causes DSDV to delay using a newly received route until it is permitted to advertise the route (i.e. 2×WST has passed).
Q7. Why is ETX’s gain more pronounced to the left of the graph?
the small packets are still useful for detecting very asymmetric links, which is why ETX’s gain over minimum is more pronounced to the left of the graph, where hop-count used very asymmetric links.
Q8. How long does the source wait before initiating the route request?
In DSR experiments with ETX, the source waits an additional 15 seconds before initiating the route request, to give the nodes time to accumulate link measurements.
Q9. How many ACK packets are used in the experiment?
ACK packets are only 38 bytes in total, including all 802.11b overhead, while the 134-byte data packets used in most of the experiments are 193 bytes with 802.11b overhead.
Q10. How many bytes of data payload is in the following measurements?
Each data packet in the following measurements consists of 24 bytes of 802.11b preamble, 31 bytes of 802.11b and Ethernet encapsulation header, 134 bytes of data payload, and 4 bytes of frame check sequence: 193 bytes in total.
Q11. What is the problem with a full queue?
If a node is sending large volumes of data, there is a danger that probe packets or routing protocol packets may be dropped or delayed due to a full queue.
Q12. What is the expected number of transmissions?
Because each attempt to transmit a packet can be considered a Bernoulli trial, the expected number of transmissions is:ETX = 1df × dr (1)ETX has several important characteristics:• ETX is based on delivery ratios, which directly affect throughput.•
Q13. Why do the points in Figure 7 lie below the y = x line?
The points for two pairs in Figure 7 lie well below the y = x line; this is because of variations in link quality between the ETX and minimum hop-count tests for those pairs.
Q14. How many packets per second can a single-hop route deliver?
A single-hop direct route can deliver up to about 450 packets per second, but the fastest two-hop route has only half that capacity.
Q15. What is the reason why the shortest route has a low loss ratio?
The minimum hop-count routes are slow because they include links with high loss ratios, which cause bandwidth to be consumed by retransmissions.
Q16. What is the definition of minimum hop-count?
Minimum hop-count is using links that deliver routing updates in one direction but deliver few or no data packets in the other, while ETX correctly avoids those links.