A Cross Layer Approach for Power Heterogeneous Ad hoc Networks Vasudev Shah and Srikanth Krishnamurthy ICDCS 2005.

A Cross Layer Approach for Power Heterogeneous Ad hoc Networks

Vasudev Shah and Srikanth Krishnamurthy

ICDCS 2005

Motivation• In the future multifarious devices together will

form an ad hoc network. e.g laptops, PDAs, low power sensor nodes etc.

• Power Control Schemes deploy multiple transmission power levels.

• These artifacts will result in link level asymmetry.

• Performance of Legacy IEEE 802.11 MAC has been shown to degrade in a network having nodes with varying transmission power.

• Current MAC layer protocols cannot handle inherent link asymmetry.

Road Map

• Problem Description

• Contribution

• Assumptions and Definitions

• Our Approach

• Simulations Models/Results

• Conclusions

Problem Description• In the presence of power heterogeneity, the IEEE 802.11

MAC signaling mechanism is inefficient: Number of hidden terminals increases. Increase in the number of False Link Failures reported to

the Routing Layer. Consequently, low power nodes suffer in terms of

throughput.

• The inefficiencies at the MAC layer affect the performance of higher layers ROUTING (e.g. AODV, DSR)Increase in the number of Route Discoveries.

Unidirectional routing schemes largely ignore MAC dependencies.

TRANSPORT TCP re-transmissions.

Inefficiency of IEEE 802.11

B

C

A

H

RTS

CTS

CTS

Range of A

Range of B

Range of H

H3

H2

Quantifying the problem with IEEE 802.11

• MAC layer Metrics: 1. Throughput Efficiency (%) 2. Data Success Rate (%)

• Simulation Models: Simulator ns2, IEEE 802.11 MAC, 40 nodes (50% nodes at 0.14W & 50% at 0.56 W), Poisson Traffic at 1000 packets/sec, Random Waypoint movement with randomly distributed nodes, speed varied from 5-10 m/s.

• To avoid Transport and Routing layer artifacts a packet generating agent is used just above the MAC layer.

• The agent randomly chooses a neighbor to communicate.

Evaluation of IEEE 802.11

Inability to identify unidirectional links

• H can reach L1 but not vice versa.

• L thinks link exists.

• May send RTS messages to H that fails.

• Repeated attempts -- can lead to wasted route discovery attempts.

• While unidirectional routing schemes attempt to route around the link they do not examine effects at MAC layer.

What we have done ..

A framework that spans the MAC and routing layers to address the problem of link asymmetry.Alleviates the hidden terminal problems at

MAC layer.Enables the identification and effective usage

of unidirectional links at the routing layer. Improvement by as much as 25 % in

terms of throughput over traditional layered approaches.

Definitions and Assumptions• In this work we consider an ad hoc network in which the

different nodes differ in terms of their maximum achievable transmission range.

• For simplicity we only consider two types of nodes, we refer to these two types as lower power nodes and higher power nodes respectively.– We assume that the transmit power of the high power nodes is

such that the transmission range is doubled.

• We use the terms homogeneous and heterogeneous to refer to networks in which all nodes have, respectively, identical or non-identical power capabilities in terms of the maximum transmission range.

Some simple MAC layer solutions

• We considered standard flooding and GPS based schemes [ICC 2001] to propagate CTS messages. – The proposed schemes further degraded the

performance of IEEE 802.11 MAC in terms of throughput of the low power nodes due to the overhead incurred due to flooding.

• In our subsequent work we attempted to use (i) smart broadcasting and (ii) using a single reservation for sending multiple DATA messages [ICC 2004].– Performance improved marginally over the traditional

IEEE 802.11 protocol.

Related Work : Unidirectional Routing

• A plurality of methods for unidirectional routing exist.

• They do not consider MAC layer dependencies.

• Some of the previous work correctly identifies the need for a MAC layer scheme that can handle link asymmetry.

Key Ideas

• Propagate the CTS message (a variant thereof called the BW_RES message is used) to high power nodes.

• Enlist the support of the routing layer at the MAC layer to “route” the BW_RES message to relevant nodes (as opposed to simply using broadcasts).

• Use this structure to identify and effectively tunnel packets to span unidirectional links.

Topology Aware CTS Propagation (TACP)

• Nodes exchange HELLO messages

• Each message includes a list of the transmitters inbound n-hop neighbors (as this information becomes available).

• Using these HELLO messages, a node constructs a localized graph

• The graph represents the node’s local neighborhood.

• The node then constructs a minimum cost tree to reach all of the high power nodes on this graph.

• It is essential to restrict “n” to small values to keep the size of these HELLO messages small.

• High power nodes are used to the extent feasible for data distribution (tree construction) so as to minimize the number of data broadcasts.

Benefits of TACP

• TACP helps in reducing the overhead of disseminating propagated BW_RES messages.

• We can use (and do use) the multi-reservation technique for making a single reservation for multiple data packets, in conjunction with TACP.

• The use of high power nodes to the extent possible minimizes the latency incurred in BW_RES messages.

• And ... TACP inherently facilitates the use of unidirectional links.

Identifying reverse tunnels• Using the localized graph, L1 can construct a reverse path to H1.

• A tunnel is established using this reverse path.

• The tunnel is used for disseminating both control and data messages.

• Unidirectional links are therefore transparent to the routing protocol.

MAC layer Simulations -- Set up and Metrics

• Throughput Efficiency: Fraction of total time that is used for sending successful data.

• Data Success Rate: Percentage of successful Data transmissions after a successful RTS/CTS exchange.

Comparisons

• We compare the results from four cases:

1. Case A: The legacy IEEE 802.11 Protocol.

2. Case B: With TACP

3. Case C : With Multi-reservations

4. Case D: With TACP + Multi-reservations.

• All nodes use TACP irrespective of whether they are high power or low power nodes -- CTS propagation was seen to help high power nodes as well.

Improvement in MAC layer Data Success Rate.

Reduction in BW_RES overhead

Improved MAC layer Throughput Efficiency

Simulation Settings: Higher Layers Included

Reduction in Route Discovery Attempts

Increase in Packet Delivery Ratio at Higher Layers

Penalty: Marginal Increase in Delay

Open Issues

• We have examined a network wherein nodes use static but fixed transmission power levels.

• What if nodes could vary the transmission powers ?

• Goal in a power controlled setting is much more lofty -- achieving an even higher capacity (in terms of throughput) than, in a homogeneous setting.

• Current approaches consider only one layer -- energy efficient routing/ MAC layer solutions.

• How can one integrate the two to achieve application layer end to end performance ?