16-08-2014, 01:26 PM
Project Report OnImproving TCP Performance in Ad Hoc Networks using Signal Strength based Link Management
[attachment=66667]
Abstract
Mobility in ad hoc networks causes frequent link failures, which in turn causes packet losses. TCP attributes
these packet losses to congestion. This incorrect inference results in frequent TCP retransmission time-outs and
therefore a degradation in TCP performance even at light loads. We propose mechanisms that are based on signal
strength measurements to alleviate such packet losses due to mobility. Our key ideas are (a) if the signal strength
measurements indicate that a link failure is most likely due to a neighbor moving out of range, in reaction, facilitate
the use of temporary higher transmission power to keep the link alive and, (b) if the signal strength measurements
indicate that a link is likely to fail, initiate a route re-discovery proactively before the link actually fails. We make
changes at the MAC and the routing layers to predict link failures and estimate if a link failure is due to mobility.
We also propose a simple mechanism at the MAC layer that can help alleviate false link failures, which occur due
to congestion when the IEEE 802.11 MAC protocol is used. We compare the above proactive and reactive schemes
and also demonstrate the benefits of using them together and along with our MAC layer extension. We show that,
INTRODUCTION
TCP performs poorly in wireless ad hoc networks as demonstrated in [1][2][3][4][5][9][10]. The main
reason for this poor performance is a high level of packet losses and a resulting high number of TCP
retransmission time-outs. First, a node drops a packet if it cannot forward the packet to the next hop
of the route on which the packet is to be relayed, as the next hop node has moved out of transmission
range. A second reason for packet loss is congestion in the shared medium. In the second case, a node
cannot reach the next hop node because there are too many nodes trying to access the channel at the same
time. The contention could even result in a single node capturing the medium, if the IEEE 802.11 MAC
protocol is used [5]. While congestion can degrade the observed performance of TCP even in wire-line
networks, mobility causes a degradation of performance of TCP in ad hoc networks even at very light
loads.
Our objective in this paper is to mainly stem the degradation in TCP performance due to mobility.
Towards this goal, we propose mechanisms to reduce the number of packet losses. These mechanisms are
based on signal strength measurements at the physical layer. Based on these signal strength measurements,
when a node fails to communicate with a neighbor, the MAC layer at the node estimate whether the failure
is due to congestion or due to the neighbor moving out of range. If the MAC layer deems that the neighbor
has just moved out of range, then, it stimulates the physical layer to increase the transmission power and
attempts to temporarily keep the link to the neighbor alive. It also prompts the routing layer to search
for a new route. The signal strength measurements can also be used to predict possible link failures to a
neighbor that is about to move out of range. Thus, if the measurements indicate that the signal strength
RELATED WORK
In [1][2][3] and [4], explicit link failure notifications are used to freeze TCP state upon the occurrence of
a route failure. Explicit route establishment notifications are used to resume TCP transmissions when a
new route is established. A fixed-RTO approach is proposed in [6] to deal with packet losses due to link
failures and route changes. In [7], a new transport layer protocol that is based on end-to-end rate control
is proposed. Various mechanisms have also been proposed to improve TCP performance at the routing
layer [8][9][10][11][12]. The COPAS protocol [8] uses node-disjoint paths for TCP-DATA packets in the
forward direction and the TCP-ACK packets in the reverse direction to eliminate interference between
TCP-DATA packets and TCP-ACK packets of the same TCP session. In contrast to COPAS, the authors
of [10] propose the use of the same route for both TCP-DATA and TCP-ACK packets in order to reduce
the total number of links that may stall the connection. A Route Failure Prediction (RFP) scheme is also
proposed in [10] to predict the occurrence of a link failure based on the trends observed in the signal
strengths of packet receptions from neighbors. In [9], the authors propose to split long TCP sessions into
multiple segments. By doing so, even if a link failure occurs in one of these segments, data flow can be
sustained on other segments. In [11], it is shown that the use of multiple paths, concurrently, does not
help in improving TCP performance. The authors of [11] propose using the shortest path as the primary
path and the shortest delay path as a backup path to improve TCP performance
REDUCING LINK FAILURES TO IMPROVE TCP PERFORMANCE
We propose mechanisms that help alleviate packet losses due to mobility. Our mechanisms are based
on measuring the signal strength at the physical layer. As pointed out in Section III, it is important to
first estimate whether a link failure is caused by mobility or by congestion. False link failures, which we
discussed earlier, cannot be overcome by tuning power levels. We propose a simple way to identify and
cope with false link failures. The methods we propose, however, only work when the level of congestion
in the network is not high and will have to be complemented by other techniques that can estimate the
level of congestion in the network. However, we justify the intuitions behind our approach via simulation
experiments (in Section V-C). The design of smart techniques to estimate the level of congestion in the
network is beyond the scope of this paper and is a topic for future research
Effects of Traffic Load
Figure 7 compares the packet losses with the Original scheme and the Combined scheme for one,
three and five TCP connections. With three and five TCP connections, the percentage of dropped packets
is higher than with one TCP connection. The reason for this increase in packet loss is increased link
layer contention, which leads to a higher percentage of false link failures (Figure 8). At higher levels
of congestion, the percentage of dropped packets increases. Thus, the proposed schemes are beneficial
primarily at light loads wherein mobility is predominantly responsible for link failures (see Figure 9).
Figure 9 shows the goodput improvement enjoyed by TCP with the Combined scheme with one, three
and five TCP connections. The total goodput improvement with three and five TCP connections is lower
than that observed with one TCP connection, except when node mobility is low. With increased network
contention, it is more difficult for the Proactive and Reactive LM schemes to salvage packets in transit
Effects of Node Mobility
Since Proactive LM and Reactive LM schemes are used to stem packet losses due to mobility, the
benefits that they provide are significant in scenarios of high mobility. Furthermore, the benefits due to
these two schemes increase with node mobility. In order to demonstrate the benefits of these schemes in
terms of helping TCP cope with mobility induced failures, we perform simulations with highly mobile
CONCLUSIONS AND FUTURE WORK
In this paper our objective is to reduce the packet losses due to mobility in ad hoc networks and
thereby improve the performance of TCP. Towards this, we propose a link management framework that
helps in salvaging TCP packets in transit upon the incidence of link failure. The framework consists of
three individual components. First, we induce a temporary increase in the transmit power level when a
node moves out of range to temporarily re-establish the failed link. This would enable the TCP packets
that are already in flight to traverse the link.
The use of the IEEE 802.11 MAC protocol causes false link failures due to congestion. We propose a
mechanism that allows us to distinguish between true link failures due to mobility and false link failures.This mechanism is based on the measurement of signal strength at the physical layer and is used to
determine if a node is still within range. We then increase power levels to temporarily reestablish a failed
link only if it is determined to be mobility induced. We include a proactive scheme, in which weak
links are identified based on these signal strength measurements and routes are proactively found prior to
failure. This scheme in turn helps in switching to the new route even before the failure occurs and thus
can stem packet losses. The proactive and reactive signal strength based schemes are unified with another
simple MAC layer extension. With our extension, the MAC layer, upon perceiving false link failures,
simply increases the number of RTS attempts in order to salvage transit TCP packets.
[attachment=66667]
Abstract
Mobility in ad hoc networks causes frequent link failures, which in turn causes packet losses. TCP attributes
these packet losses to congestion. This incorrect inference results in frequent TCP retransmission time-outs and
therefore a degradation in TCP performance even at light loads. We propose mechanisms that are based on signal
strength measurements to alleviate such packet losses due to mobility. Our key ideas are (a) if the signal strength
measurements indicate that a link failure is most likely due to a neighbor moving out of range, in reaction, facilitate
the use of temporary higher transmission power to keep the link alive and, (b) if the signal strength measurements
indicate that a link is likely to fail, initiate a route re-discovery proactively before the link actually fails. We make
changes at the MAC and the routing layers to predict link failures and estimate if a link failure is due to mobility.
We also propose a simple mechanism at the MAC layer that can help alleviate false link failures, which occur due
to congestion when the IEEE 802.11 MAC protocol is used. We compare the above proactive and reactive schemes
and also demonstrate the benefits of using them together and along with our MAC layer extension. We show that,
INTRODUCTION
TCP performs poorly in wireless ad hoc networks as demonstrated in [1][2][3][4][5][9][10]. The main
reason for this poor performance is a high level of packet losses and a resulting high number of TCP
retransmission time-outs. First, a node drops a packet if it cannot forward the packet to the next hop
of the route on which the packet is to be relayed, as the next hop node has moved out of transmission
range. A second reason for packet loss is congestion in the shared medium. In the second case, a node
cannot reach the next hop node because there are too many nodes trying to access the channel at the same
time. The contention could even result in a single node capturing the medium, if the IEEE 802.11 MAC
protocol is used [5]. While congestion can degrade the observed performance of TCP even in wire-line
networks, mobility causes a degradation of performance of TCP in ad hoc networks even at very light
loads.
Our objective in this paper is to mainly stem the degradation in TCP performance due to mobility.
Towards this goal, we propose mechanisms to reduce the number of packet losses. These mechanisms are
based on signal strength measurements at the physical layer. Based on these signal strength measurements,
when a node fails to communicate with a neighbor, the MAC layer at the node estimate whether the failure
is due to congestion or due to the neighbor moving out of range. If the MAC layer deems that the neighbor
has just moved out of range, then, it stimulates the physical layer to increase the transmission power and
attempts to temporarily keep the link to the neighbor alive. It also prompts the routing layer to search
for a new route. The signal strength measurements can also be used to predict possible link failures to a
neighbor that is about to move out of range. Thus, if the measurements indicate that the signal strength
RELATED WORK
In [1][2][3] and [4], explicit link failure notifications are used to freeze TCP state upon the occurrence of
a route failure. Explicit route establishment notifications are used to resume TCP transmissions when a
new route is established. A fixed-RTO approach is proposed in [6] to deal with packet losses due to link
failures and route changes. In [7], a new transport layer protocol that is based on end-to-end rate control
is proposed. Various mechanisms have also been proposed to improve TCP performance at the routing
layer [8][9][10][11][12]. The COPAS protocol [8] uses node-disjoint paths for TCP-DATA packets in the
forward direction and the TCP-ACK packets in the reverse direction to eliminate interference between
TCP-DATA packets and TCP-ACK packets of the same TCP session. In contrast to COPAS, the authors
of [10] propose the use of the same route for both TCP-DATA and TCP-ACK packets in order to reduce
the total number of links that may stall the connection. A Route Failure Prediction (RFP) scheme is also
proposed in [10] to predict the occurrence of a link failure based on the trends observed in the signal
strengths of packet receptions from neighbors. In [9], the authors propose to split long TCP sessions into
multiple segments. By doing so, even if a link failure occurs in one of these segments, data flow can be
sustained on other segments. In [11], it is shown that the use of multiple paths, concurrently, does not
help in improving TCP performance. The authors of [11] propose using the shortest path as the primary
path and the shortest delay path as a backup path to improve TCP performance
REDUCING LINK FAILURES TO IMPROVE TCP PERFORMANCE
We propose mechanisms that help alleviate packet losses due to mobility. Our mechanisms are based
on measuring the signal strength at the physical layer. As pointed out in Section III, it is important to
first estimate whether a link failure is caused by mobility or by congestion. False link failures, which we
discussed earlier, cannot be overcome by tuning power levels. We propose a simple way to identify and
cope with false link failures. The methods we propose, however, only work when the level of congestion
in the network is not high and will have to be complemented by other techniques that can estimate the
level of congestion in the network. However, we justify the intuitions behind our approach via simulation
experiments (in Section V-C). The design of smart techniques to estimate the level of congestion in the
network is beyond the scope of this paper and is a topic for future research
Effects of Traffic Load
Figure 7 compares the packet losses with the Original scheme and the Combined scheme for one,
three and five TCP connections. With three and five TCP connections, the percentage of dropped packets
is higher than with one TCP connection. The reason for this increase in packet loss is increased link
layer contention, which leads to a higher percentage of false link failures (Figure 8). At higher levels
of congestion, the percentage of dropped packets increases. Thus, the proposed schemes are beneficial
primarily at light loads wherein mobility is predominantly responsible for link failures (see Figure 9).
Figure 9 shows the goodput improvement enjoyed by TCP with the Combined scheme with one, three
and five TCP connections. The total goodput improvement with three and five TCP connections is lower
than that observed with one TCP connection, except when node mobility is low. With increased network
contention, it is more difficult for the Proactive and Reactive LM schemes to salvage packets in transit
Effects of Node Mobility
Since Proactive LM and Reactive LM schemes are used to stem packet losses due to mobility, the
benefits that they provide are significant in scenarios of high mobility. Furthermore, the benefits due to
these two schemes increase with node mobility. In order to demonstrate the benefits of these schemes in
terms of helping TCP cope with mobility induced failures, we perform simulations with highly mobile
CONCLUSIONS AND FUTURE WORK
In this paper our objective is to reduce the packet losses due to mobility in ad hoc networks and
thereby improve the performance of TCP. Towards this, we propose a link management framework that
helps in salvaging TCP packets in transit upon the incidence of link failure. The framework consists of
three individual components. First, we induce a temporary increase in the transmit power level when a
node moves out of range to temporarily re-establish the failed link. This would enable the TCP packets
that are already in flight to traverse the link.
The use of the IEEE 802.11 MAC protocol causes false link failures due to congestion. We propose a
mechanism that allows us to distinguish between true link failures due to mobility and false link failures.This mechanism is based on the measurement of signal strength at the physical layer and is used to
determine if a node is still within range. We then increase power levels to temporarily reestablish a failed
link only if it is determined to be mobility induced. We include a proactive scheme, in which weak
links are identified based on these signal strength measurements and routes are proactively found prior to
failure. This scheme in turn helps in switching to the new route even before the failure occurs and thus
can stem packet losses. The proactive and reactive signal strength based schemes are unified with another
simple MAC layer extension. With our extension, the MAC layer, upon perceiving false link failures,
simply increases the number of RTS attempts in order to salvage transit TCP packets.