A Power Saving Mechanism Computer Science Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

In reducing communication energy consumption, various approaches have to be considered and it is reasonable to ask if the network interface contribute significantly to the overall energy consumption of a mobile system. Different operating modes, variety of devices, energy management techniques and usage scenarios makes it impossible to make clear statement about energy consumption in portable devices. Obviously, a measurement of specific systems quickly becomes outdated. Nevertheless, M.Stemm and R.H Katz measurements shows that the network interface represents a significant fraction of the energy consumed by a laptop and is dominant source of energy consumption in some PDA hardware. Recently, Oliver Kasten and Marc Langheinrich implemented a preliminary blue tooth- based sensor devices and the interface accounted for over 40% of the total energy comsumption when the blue tooth is in standby mode. As advances continues to be made to low power hardware and energy efficient operating systems and applications, the relative cost of communication may be expected to increase.

Energy- efficient communication protocol design and evaluation requires practical understanding of the energy consumption behaviour of the underlying network interface. Energy consumed by an interface greatly depends on its operating mode i.e in the sleep mode and idle mode.

IEEE 802.11 base interface have attracted loads of attention because of its wide availability, low cost and relatively stable and open specification [IEEE computer society LAN, MAN standards committee]. The table below shows a summary of some experimental measurements of the power consumption of various networks interface considering the fact that data may vary among various manufacturers, models and measurement methods.

Power consumption measurement







IEEE 802.11 (2.4Ghz)

Airconet PC4800[8]

1.4 - 1.9 W

1.3 -1.4 W

1.34 W

0.075 W


Lucent Bronze [10]


0.97 W

0.84 W

0.066 W


Lucent Silver [10]


0.90 W

0.74 W

0.048 W


Cabletron Roamabout [6}


1.0 W

0.83 W

0.13 W


Lucent wavelan [20]



1.5 W



Diagram 4

Diagram courtesy of . IEEE Computer Society LAN MAN Standards Committee. IEEE 802.11 Standard: Wireless LAN Medium Access Control and Physical Layer Specifications, August 1999.


This chapter looks into energy efficient Mac protocol for wireless networks that tries to reduce the consumption of energy by changing the state of the wireless interfaces in a "doze" state when deemed reasonable. Power saving Mode of IEEE 802.11 is being considered.

In general, the best way to save power for wireless communication devices would be to switch them off. Unfortunately, one can't do this without losing the capability to communicate in both directions that is a station in power saving mode will not be aware of any packet arriving for it at this time. Therefore, 2 problems have to be addressed in the power saving mode;

How does a station in power saving mode receive packet from other station?

How does a station send to another station in power save mode?

The fundamental idea for all station in power saving mode is to synchronize all node to wake up at the same time. When a window wants to start synchronization, the sender announces buffer frames for the receiver. The received station will stay awake until the frame is being delivered. This is done easily in infrastructure network, where there exists a central access point that stored packets for stations in a doze state and to synchronize all mobile stations. This is difficult for Ad-hoc networks, where the packets store, forward and also the timing synchronization has to be done in a distributed manner.

Generally, a wireless network interface can be in 3 states namely awake state, doze state and off state [Eun-sun Jung, Aug 2006]. Consideration will be given to awake and doze state because MAC protocol places wireless interface in either the doze or the awake state at a given time. The transition from the doze state to the awake state requires a small duration of time [A. Kamerman and L. Monteban, 1997]. Likewise the transition from the doze state to the awake state results in additional energy consumption during transition [P.J.M Havinga and G.J.M Smit, Nov 2000].

3.2 Power save Protocol

Power save protocols puts anode's network interface into the sleep mode in order to save energy. Sleep mode cannot forward or receive traffic and its unavailability may interrupt the flow of traffic through a multi hop Ad hoc network. Power save protocols seeks to maximize energy saving while minimizing impact on throughput, latency and route latency. Network layer protocol and MAC layer protocol are the main classes of power save protocol.

Power saving protocols is based on 3 basic strategies

Synchronized power save mechanism: This mechanism helps nodes to periodically wake up to listen to announcements of pending traffic and remain awake to exchange traffic, if necessary. The restricted windows for announcing and forwarding traffic can result in high latencies. Synchronization required can be problematic in a dynamic, multi hop Ad hoc network.

Network Topology: Some covering set of the network is defined such that its topologically representative of the network. The nodes in the covering set provide connectivity equivalent to the of the full network, so that the remaining node can spend most of their time in the sleep state with minimal impact on network performance. This protocol can be synchronous or asynchronous, both in determining the covering set and in traffic forwarding.

The third approach is intended for fully asynchronous operation, which nodes maintain independent and possible even dissimilar sleep-awake schedules. The schedules are designed in such a way that the neighbouring nodes schedules are guaranteed to eventually overlap. Re transmission rules are defined such that a bounded number of attempts are required to permit two nodes to establish connectivity.

In IEEE 802.11 ad hoc mode, stations cooperate to support the power saving mechanism since no infrastructure existed. Active station will buffer packets for those stations in sleep mode and try to notify them for data transmission. Sleeping station will be awakening periodically to listen to possible notification messages. These notification messages are known as ATIM (Ad hoc traffic indication messages). Stations exchange ATIM only during the ATIM window, at times windows with a fixed length. All station will be active to listen to possible ATIMs during ATIM windows. Stations which have buffer packets for other stations compete to send ATIMs to notify the target stations not to enter the sleep mode. Stations that receives the ATIM will reply an acknowledgement (ACK) packets to the sender and keep active in the rest of the beacon interval

Diagram 5

Diagram 5 Shows an example of the first beacon interval, both station A and B can go into the sleep mode since they do not receive any ATIM frame. In the second beacon interval however, station A has to stay active after the ATIM window since it receives an ATIM frame from station B. Then A and B can exchange DATA and ACK frames after ATIM window. Since C doesn't have any packet to send or receive, it dozes off after the ATIM window.

Clock synchronization is of great importance for power management in an IEEE 802.11 ad hoc network because each station determines the beginning and the end of an ATIM window by maintaining a timer. Power-saving mechanism will not function properly if stations cannot synchronize their clocks. Distributed Timing synchronization function (TSF) is being specified by the IEEE 802.11 standard to fulfil clock synchronization in an independent basic service set (IBSS). This IBSS is a fully connected ad hoc network in which all stations are within the transmission range of each other. Each station maintains a TSF timer with modulus 2^64 counting in increments of micro second (µ). TSF timer value is the summation of a variable off set and station's clock. Clock synchronization is achieved by periodically exchanging timing information through beacon frames. Beacon frame contains timestamp declaring when the beacon was sent. After a station receives a beacon and finds that its own TSF timer is slower than the timestamp specified in the beacon. It will add the timing difference to its offset. This implies that clocks only move forward, but never backward [L. Huang and T.H. Lai, 2002].

ATIM frame transmission is performed using the CSMA/CA (collision avoidance) mechanism specified in IEEE 802.11. Power saving mechanism specified in IEEE 802.11 uses the same fixed ATIM window size as well as identical beacon intervals [IEEE Press, Nov 1997]. ATIM window size critically affects energy consumption and throughput, a fixed ATIM window doesn't perform well in all situations as illustrated by [H.Woesner et al]. When ATIM window chooses is also too small, enough time may not be available to announce buffered packets (by transmitting ATIM frames) that degrades throughput. Also when the ATIM window is too large, less time will be incurred for the actual data transmission because data is transmitted at the end of the ATIM window which degrades throughput at high loads. Large ATIM windows can also results in higher energy consumption since all nodes stays awake during ATIM window [Eun- sun, Jung, Aug 2006].

At the beginning of each beacon interval, all station contends with each other to send a beacon frame. Firstly, they wait a random delay, if a beacon arrives before its random delay timer expires, the station will give up its beacon transmission during this beacon interval. Otherwise, when its random delay timer expires, the station transmits a beacon with a time stamp which is equal to the value of the TSF timer at the time the first bit of the time stamp was transmitted to the physical layer plus the transmitting station's delay through its physical layer from the MAC - PHY interface to its interface with the wireless medium. Upon receiving a beacon, the station will adjust the time stamp by adding an amount equal to the delay through the physical layer and then set its TSF timer to the adjusted time stamp if the value of the adjusted time stamp is larger than its TSF timer. Therefore, if faster stations fail in contending to transmit their beacon frames for sometimes, they may go out of synchronization

In the simulation and analysis done by L. Huang and T.H. Lai, it shows that synchronization gets more serious especially when the size of the IBSS becomes larger. It means IEEE 802.11 TSF does not scale very well because stations can only set their timer forward, the station with the fastest clock will suffer from asynchronism with a high probability if it fails to transmit beacon for too many beacon intervals. L. Huang and T.H Lai propose an adaptive timing synchronization procedure (ATSP) to solve these asynchronisms problems. The reason behind the ATSP is to give faster stations a higher priority to send beacons. ATSP helps the fastest station to have a very high probability of successfully sending its beacon, thereby synchronizing all other stations. ATSP helps to provide a simple effective solution to improve the clock synchronization mechanism compared to the IEEE 802.11 TSF.

3.3 Synchronous Power Save Protocol

In synchronous power save protocol, nodes from time to time wake up and exchange traffic announcements and pending traffic. One area that is difficult in this protocol is to determine interval and announcement windows that maximize energy saving while minimizing impact on throughput and latency. This method also requires node to maintain a globally synchronized sleep-wake cycle, meaning that they share arbitrary phase information.

Energy consumption is the reason in evaluating a power-save protocol, latency, throughput and distribution of power consumption must be taken into account as well. The time spent in the sleep state is only an indication of the actual energy savings, which will definitely be reduced by the cost of the state transition, beaconing and ATIM traffic, all of which are sensitive to these configuration parameter.

H. Woesner et al, described a simulation that studied the effectiveness of the IEEE 802.11 power - save protocol for a fully connected eight - node IBSS. This experiment measured throughput and time spent in the sleep state for a variety of beacon intervals, ATIM window lengths and offered loads. The result shows considerable dependence on the beacon interval: short throughput. H. Woesner suggests that "if we were to sacrifice about 10% in throughput, we could save about 30% energy". These savings are only obtained at quite moderate loads; as offered load increases from 15% to 30%, the available savings declines substantially.


L. Huang and T.H Lai proposed ATSP to improve on the problem of clock asynchronization, but it was just designed for single hop ad hoc network. Whereas clock synchronism problem is IEEE 802.11 TSF may worsen in a multi hop mobile ad hoc network because packet delays may vary due to unpredictable mobility and radio interference. Asynchronous powers saving protocol are power management protocol that functions perfectly in a MANET without the support of clock synchronization [shu-min et al]. With asynchronous power saving protocol, stations do not need to synchronize their timer with each other. Station should be able to discover their neighbours without any issue even if they go to sleep sometimes. Precise clock information will make a station be aware of its neighbour while incorrect and incomplete neighbour information may be harmful to most current routing protocols.

There are three (3) types of asynchronous protocol proposed by Y-C. Tseng, et al. They are dominating awake interval, periodically-fully- awake interval and quorum protocols.

Dominating-awake-interval protocol

Asynchronous power saving protocol doesn't count on clock synchronization because the wake up pattern of any two stations must overlap with each other at a given point in time, no matter how their clock drift away. Dominating awake interval protocol is used to order a station to stay awake long enough to ensure that neighbouring station can hear from each other and if desired, deliver buffered packets. Dominating awake means that a station should stay awake for at least half of the length of Beacon Interval (BI). This implies that the station's beacon window will always overlap with any neighbouring station's active windows and vice versa.

Diagram 6

Diagram 7

This diagram shows how beacon window appearing at the beginning of beacon intervals, station B can hear station A's beacon but A always misses B's beacon. In order to solve this issue, beacon interval alternatively labelled as odd and even intervals. Odd can start with an active window; this active window is led by beacon window followed by an MTIM window. Even beacon interval also starts with an active window, but the active window is terminated by an MTIM window followed by a beacon window. Diagram illustrates this and with this design, a station is able to receive its neighbour's beacon frames in every two beacon interval if no collisions occur in receiving the latter's beacon

Periodically-fully- awake- interval protocol

This protocol was designed to reduce on duty time because it was noticed that dominating awake interval protocol doesn't save much energy since the stations have to keep active at least half of the time. In periodically fully awake interval protocol, two (2) types of beacon intervals are designed

Low-power intervals

Fully awake interval

The low power interval starts with an active window that contains a beacon window followed by an MTIM window, such that AW= BW+MW. The station goes to sleep mode in the rest of the time. While fully awake interval also starts with a beacon window which is followed by an MTIM window. The station must remain awake for the rest of the time i.e. AW=BJ. Fully-awake interval protocol arrive periodically every N beacon intervals, and the rest of the beacon intervals are low power intervals. Low power interval is for a station to send out its beacon to infirm others of its presence, while the fully awake intervals are for a station to discover who are in its neighbourhood. Fully awake interval always overlaps with any station's beacon windows, no matter how their clock drifts away. Periodically- fully-awake interval saves more energy as long as N> 2 comparing to the dominating-awake-interval protocol that requires a station to stay awake more than half of the time. However, the response time to get aware of a newly appearing station could be as long as N beacon intervals.

Quorum-based protocol

This protocol was proposed to alleviate the contention of a station having struggled to send a beacon in each beacon interval which the two other protocols have exhibited. Quorum is a set of entities from which one has to obtain permission to perform some critical action (Shu-min chen et al). Basically, two (2) quorum sets always have non empty intersection so as to guarantee the atomicity of a transaction. The concept of quorum is adopted to design stations' wake up patterns so as to guarantee that a station's beacon can always be heard by other station's active windows. The structure of quorum protocol is as follow;

Beacon interval are divided into groups such that each continuous M^2 beacon interval are called a group, where m is a global parameter. Each group, m^2 interval are arranged as a two dimensional m x m array in a row-major manner. On the m x m array, a station can arbitrarily pick one column and one row of entries as its quorum members. 2m - 1 intervals are called quorum intervals, the remaining m^2 - 2m + 1 intervals are called non - quorum intervals. Example of m = 3 is shown below

Diagram 8

An example of a grid-based quorum, representing a 3 x 3 grid and showing where the intersection of Host A and B met each other at interval 2 and 6.

Host A picks the first column and the first row as its quorum intervals, while HOST B picks the Third column and the third row as its quorum intervals. When perfectly synchronizes, interval 2 and 6 are their intersections.

How Quorum and Non Quorum Works;

Each quorum interval starts with a beacon window followed by an MTIM window. The station stays awake for the rest of the interval that is AW= BI.

Non- quorum interval starts with an MTIM window and after that, the station enters the sleep mode that is AW = MW, if there is no expected traffic. The quorum-base protocol designs guarantees that a station always have at least 2 entire beacon windows that are fully covered by any station's active windows in every M^2 beacon interval no matter how much its clocks drift away [Y.C.Tseng et al]. Quorum-base protocol has an edge over the other two (2) protocol in the sense that it only has to contend transmit beacon in 0(1/m) of the beacon intervals. Quorum-base protocol has a problem of some delay in perceiving a newly approaching neighbour.

Asynchronous protocols only guarantee network connectivity in higher dynamic networks since they trade power consumption for network connectivity.


Asynchronous protocol doesn't rely on clock synchronization, but the cost of energy is still relatively high when compared with synchronous protocols because stations need to keep awake for a longer time to discover neighbours. In order to overcome the deficiency of asynchronous power saving protocols, C.S. Hsu and Y.-c. Tseng proposes several cluster- based semi asynchronous power- saving protocol f or multi- hop MANETS. The principle is to cluster neighbouring stations such that synchronous power saving protocols can be adopted within individual clusters and asynchronous powers saving protocols are only used at the boundaries of different clusters.

The semi - asynchronous protocols works with the stations that are clustered first. The station with sufficient energy and the fastest clock among its direct neighbours will be chosen as the cluster head. Since one hop existed from the cluster head to other members, it will be easier for the cluster head to synchronize all member stations in the cluster. When all stations in the same cluster are synchronized, each station can simply adopt a synchronous power saving protocol i.e the 802.11's power saving mechanism as its intracluster protocol. With clustering, the whole network is partitioned into several clusters whose clocks are not necessarily synchronized; so these clusters may have different wake up pattern and this may be out of synchronization. In solving this problem, each cluster should also run an asynchronous power-saving protocol to detect its neighbours. Since each cluster consist of several stations, it is not necessary for every cluster member to run the asynchronous protocol. C- S. Hsu and Y-C. Tseng proposes that within each cluster, some stations should be delegated a s watchers to run an asynchronous protocol for neighbour discovery. The man problem of semi- asynchronous protocol is how to delegate these watchers.

Two schemes are been propose

SNR- Probability based

Location- based schemes

The SNR- Probability-based is designed for loosely coupled clustering and it used to estimate the station's distance to the cluster head. The greater the SNR the shorter the distance becomes. The station with the longer distance to the cluster head will have a higher probability to serve as a watcher because it's closer to the border of the cluster.

Location-based scheme is designed for tightly coupled clustering where the cluster head needs to record the physical locations of all its cluster members, which is obtained via GPS receivers.

Propose IPSM Scheme

IPSM scheme is being proposed by Eun sun Jung to improve on the power saving mechanism. These bought about the name Improve Saving Mechanism (IPSM). Two main differences between IPSM and PSM are being specified in IEEE 802.11.

Dynamic adjustment of ATIM windows: In the propose IPSM scheme, each node independently chooses an ATIM window size based on observed network conditions and might potentially result in each node using a different ATIM window size.

Longer Dozing time: In power saving mechanism, a node transmit or receives an ATIM frame during an Atim window, it stays awake during the entire beacon interval. This method allows a single ATIM frame from station A to station B, to be followed up by multiple data packets during the remaining beacon interval. Which means that a single ATIM frame and ATIM-ACK exchange can be used to deliver several packets in the same beacon interval? This method has it own advantages but at the low loads, it results in much higher energy consumption than needed. While the IPSM scheme, allow a node to enter the doze state after completing any transmissions that are explicitly announced in the ATIM window.

In the Improved Power saving mechanism, a node uses only one ATIM frame to announce pending packets for the same destination during the same beacon interval. When station A transmit and ATIM frame to station B, station A will not transmit another ATIM frame to the same destination in the same beacon interval. Instead, station A include in each packet sent to station B the number of packets still pending for station B. This help station B to know if it has receive will packets that are pending at station A at the time of data transmission. If station A couldn't deliver all pending packets that were announced to station B and the current beacon interval expires, station A and B will both stay up in the next beacon interval while station B anticipating the remaining packets from station A without sending ATIM frame to station B. STtation A delivers the remaining packets to B and moves to the doze state if no packets are pending to transmit or receive are announced. In the same way, station B may enter the doze state after receiving the previously announced packets, if no other pending packets are coming. Like the PSM, if packet arrives after a node has already entered the doze state they will be buffered and announced during a subsequent ATIM window.

3.6 IPSM Protocol

Ipsm operations are presented in details and few terms are defined

ATIMmin: The minimum ATIM window size that a node can have.

ATIMmax: The maximum window size that a node can have.

ATIMinc: The amount of time by which a node can increase the ATIM window at a time.

CIT(Channel idle Time): Current channel idle time measured at the end of the ATIM window. During the ATIM window, each node measure how long the channedl was idle ccontinously. When the channel is busy, the CIT value is set at the end of the ATIM window. It implies that the channed is idle at that time.

CITThreshold (Channel idle time Threshold): At the end of teh ATIM window, if CIT is greater than CITThreshold the channel is idle long enough to assume that no node intend to transmit an ATIM frame.

Diagram 9 illustrate IPSM scheme

At the beginning of each beacon interval, station transmit beacon as described in IEEE 802.11. After, each station begins with ATIM window size which equals to ATIMmin. When the ATIM window finishes, anode increases its ATIM window by ATIMinc and restatrts the ATIM window for ATIMinc. Explaining the diagram above, A and B starts their data transmission after the increase ATIM window finishes. After completing data transmission, they start dozing for the remaining beacon interval. Since C has nothing to send or receive, it dozes off after the ATIM window. All nodes wakes at the beginning of the next beacon interval and the process is being repeated.

It should be noted that there exist a finite delay associated with the doze-to-awake transition as well as high energy consumption. Therefore, with IPSM, a node will not enter the doze state after completing packet transmission if the remaining duration in the current beacon interval is "too small". The delay for both doze to awake and awake to doze transition is assumed to be 800 µs each [W.R. Heinzelman et al and T. Simunic et al].

3.7 IPSM Hidden and Expose Terminal Problem

IPSM protocol may have the exposed terminal problem due to the CSMA mechanism. Diagram 10 (a) and 10 (b) illustrate the hidden and exposed terminal problems.

Diagram 10

Station A wants to send a packet to station B and station C wants to send packet to station D. When station A sends to B, C will not hear, so it assumes the channel is idle. Therefore, when C sends to D, it will cause collision at B. This is known as hidden terminal problem, that is A is hidden from C. This hidden terminal problem causes more waste of channel bandwidth and collisions.

The exposed terminal is when station B send s a packet to A, C has to defer its transmission to avoid collision at B. Therefore when D send a packet to C, C will not respond to D. It means C is exposed to B. While the hidden terminal problem results in data loss, the exposed terminal problem reduces efficiency.

IPSM has the exposed terminal problem during the ATIM window. It can be recalled that ATIM window in IPSM expires when the channel is idle for 128 slot time. Diagram (b) illustrates that A and B communicates ATIM and ATIM-ACK with each other, and C wants to send an ATIM to D. While C is deferring its transmission for B, D will sense the channel to be idle. Thus, D will finish its ATIM window and enter the doze state without receiving packets from C.

Eun-sun Jung, propose a simple solution to overcome this problem. In IPSM, anode can't transmit an ATIM, it will choose a small CW (Contention Window) value in the subsequent beacon interval. If an ATIM-ACK is not received in response to the transmitted ATIM frame, retransmission will occur with the use of CSMA/CA mechanism. Which means the node transmitting the ATIM frame doubles the value of the CW, selects a new back off interval and repeats the procedure to transmit the ATIM frame. Its being stated that this is not a perfect solution to the problem of expose terminal but it works reasonably well. Another solution being proposed by Eun sun Jung is by using a separate busy tone channel. Using the diagram (b) above, station C receives packet from B and defer its transmission to D. By using the busy tone channel, C transmits busy tones to D so that node D can sense it [M.J. Miller and N.H. Vaidya, dec 2004].

3.8 Power Saving Mode With TCP

Transport Control Protocol is known as a connection-orientated, reliable data delivery protocol. TCP make use of a congestion window (CWND) for congestion control at the sender, which indicates the amount of data that can be injected into the network. TCP will assume a network is congested if it fails to receive an acknowledgment for a packet within some retransmission timeout (RTO) interval. Retransmission timeout is determined by estimating the mean and variance of the Round trip time (RTT). Packets are retransmitted when packet loss is detected. The sender will decrease the congestion window to avoid further congestion. Then the data being transmitted will be reduced at the sender. Likewise when the sender receives an acknowledgment, it increase the congestion window to transmit more data.

Power saving mode in PSM or IPSM don't perform well in TCP because in PSM, if the ATIM window is too small, some nodes may not be able to announce their buffered packets. Even when the nodes enter the doze state for a few beacon intervals, TCP will time out and cause packets to be transmitted again. This affects the performance of the TCP. Also if the ATIM window is large, there will be a delay for a sender to receive acknowledgments from its receiver. Thus, the congestion window will increase slowly having effect on the amount of data transmission and the TCP performance. However, Eun sun Jung observed that the ATIM window also delay a sender to receive acknowledgment which makes the congestion window to increase slowly that causes the TCP throughput degradation.

TCP throughput for IPSM may be worse than that of PSM because anode can enter the doze state in the middle of a beacon interval. This gives good energy savings for CBR (Constant Bit Rate) but not for TCP because it reduces time for data transmission. Because TCP through is degraded with any power saving mode, Eun-Sun Jung allow IPSM to give up power saving mode when there is TCP traffic in the network which means nodes will stay awake if there is TCP traffic. This is referred to as IPSM-T (IPSM for TCP). IPSM-T helps nodes maintain a time-to-alive-timer. When a node sends or receivers a TCP packets, it set the timer for a predefined interval. The node will not enter the doze state as long as the timer is not zero. Also, when a sender transmits TCP packets, it indicates TCP traffic using one bit inside the packet headers of ATIM, RTS and DATA. So, when nodes neighbours overhears these packets they look at the packet header and check if the TCP flag is 1. If so, they update their time -to-alive times and will not enter the doze state. These nodes will use power saving mode and enter the doze state after their ATIM window finish. R.Zheng and R. Kravets had propose similar approach that uses a keep - alive time where nodes stay awake for a period of time if they receive routing packets, such as route request, route reply etc.


This simulation was carried out by N. H. Vaidya and E.S. Jung, july 7, 2004.

Evaluation Performance

Proposed IPSM schemes, PSM scheme in IEEE 802.11 and IEEE 802.11 without any nodes in PSM are being simulated. IEEE 802.11 without any node in PSM is referred to as (WOPSM) without Power Saving Mechanisms. Two metrics are used to evaluate the proposed scheme.

Aggregate throughput over all flows in the network

Total data delivered per unit of energy consumption (Kbits delivered per joule). This metric measures the amount of data delivered per joule of energy. Total data delivered is being divided by the total energy consumed by all nodes in the network. The total energy consumption is the sum of each node's energy consumption during simulation time.


Network simulator - 2 (ns-2) with the CMU wireless extension is used in the simulation [The CMU Monarch project. "The CMU Monarch project's wireless and mobility extensions to NS"]. Duration for each simulation is 20 seconds in a wireless LAN. Each flow transmits CBR traffic and this varies for different simulations. The channel bit rate is 2mbps and packet size is 512bytes. Power represented for the energy model used are 1.65W, 1.4W, 1.15W and 0.045W which are consumed by the wireless network interface in the idle, transmit, receive mode and doze state respectively. For the doze to transmit time, 800µs is used [T. Simunic et al] which used less energy better than 250µs that A. Kamerman and L. Montebau proposes. Each node starts with enough energy so it won't run out of energy during the simulations. The results of the simulation are 30 runs on average. 100ms is used for each beacon interval [The editors of IEEE 802.11, wireless LAN (MAC) and (PHY) specification, 1997]. For PSM simulator, ATIM window size varies from 2ms to 40ms; number of nodes is chosen to be 10, 30 or 50. In all cases, half of the nodes transmit packets to the other half. This means node is n, there is a total of n/2 flows in the network.

Simulation Results

Diagram 11(a)

Diagram 11(b)

Diagram 11(c)

Fixed Network Load: The graph above shows the aggregate throughput (Aggregated over all flows) with a fixed network load when using PSM, IPSM, and WOPSM schemes. In the graph, I denote propose IPSM scheme and W WOPSM scheme. The performance of the PSM scheme depends on the given ATIM window size. The diagram plots the throughput of PSM as a function of the ATIM value, where ATIM values are between 2ms and 40 ms.

The graph indicates that the ATIM window size impacts the throughput achieved by PSM quite significantly especially when large number of nodes are used in diagram...(c). IPSM typically yields aggregate throughput comparable to WOPSM and PSM with the optimal ATIM window size.

It can be recalled that in PSM and IPSM, each node only transmits one ATIM frame for many pending packets for the same destination. Thus, a small number of nodes in the network, the ATIM window size for PSM are less sensitive in fig..... (a) as compared to fig....(c). Stating an example from fig...(a), a 4ms ATIM window is enough to achieve a desirable throughput for a network load of 10%. However, with a network load of 40% in (c), an ATIM window of about 15ms gives the best throughput. As the number of nodes increase or the network load gets heavier, the ATIM window size becomes a significant factor for both throughput and energy consumption in PSM. If the ATIM window is too small, there is not enough time to announce all the pending packets which will result in throughput degradation occur which results in throughput degradations well. Aggregate throughput is also degraded in both PSM and IPSM with network needs more time for data transmission, but PSM and IPSM both uses that extra channel capacity for the ATIM window.

Diagram 12(a)

Diagram 12(b)

Diagram 12(c)

The above diagrams illustrate the total data delivered per joule. IPSM performs well than PSM in the sense that PSM allow a node to be in the Power Saving Mode, the node can't enter the doze state if it has at least one packet to transmit or receieve. IPSM overcame this disadvantage by making it possible for node to enter the doze state whenever it finishes the transmission leading to reduced energy consumption as compared to PSM.

When the network load increases, energy saving from the power saving mode becomes smaller. Using the diagram above, the total data delivered per joule in the 30 and 50 nodes. Network (b) and (c) are less than that in the 10 node network (a). It should be noted that the scale for the vertical axis are not the same. When network load is high, less time are incurred for a node to be in the doze state due to data transmission. Which makes a node to take more time to finish data transmission; hence, it is more likely to be awake. The shorter duration in the doze state yields less energy conservation.

Dynamic network load: Using CBR, dynamic network load is simulated in fig..... shows the source nodes start with a network load of 40% and the network load changes from 40% to 10% at 5 seconds. The node then changes from 10% to 20% and from 20% to 30% at 5 second interval.

Diagram 13

Diagram 14 (a)

Diagram 14(b)

Diagram 14(c)

Aggregate throughput: dynamic network load.

Diagram 15(a)

Diagram 15(b)

Diagram 15(c)

Total data delivered per joule: dynamic network load.

Results are now presented for the wireless LAN where the load is time-varying. IPSM scheme is labelled I on horizontal axis, WOPSM scheme labelled W on horizontal axis, and the PSM scheme (with different values of fixed ATIM window size for PSM).

The graph above (a) and (b), ATIM windows of 5ms and 10ms in PSM gives comparable throughput to IPSM and WOPSM respectively. However graph.... (a) And (b) gives a corresponding throughput per joules that is lower than IPSM. And since ATIM window size affects both throughput and energy consumption, PSM doesn't perform well all the time. AYIM window size should be dynamically adjusted according to the number of nodes contending in the network and the network load. This simulation using a dynamic load shows that IPSM always performs well compared to PSM.


Chapter 3 presents energy efficient protocol that improves Power saving mechanism for Distributed control function DCF in an IBSS of IEEE 802.22. ATIM window size is fixed in PSM which makes it a major concern for throughput and energy consumption. When the number of nodes increases, the ATIM window size becomes bigger and won't make the fixed ATIM window perform well at the time. In PSM, when the ATIM window is small, throughput degrades as the network load becomes heavier but if the ATIM window is too large, the energy gain from power saving mode becomes smaller since each node must stay awake during ATIM window.

Several techniques for energy conservation in wireless ad hoc networks such as Basic power management mechanism in IEEE 802.11 standard is being reviewed and since standard power management mechanism doesn't scale well in multi hop ad hoc network due to the difficulty for clock synchronization, several asynchronous and semi -asynchronous power saving protocols are looked at.

Improvement of PSM is being discussed also because a node can also power off its wireless network interface whenever it finishes packet transmission for the announced packets and gives significant energy savings. Simulations carried out by Eun sun Jung shows how IPSM outperforms PSM with respect to energy savings. It's also noted that PSM doesn't work well with TCP traffic, so using the PSM, nodes have to use extra channel bandwidth for the ATIM window which delays a sender to receive acknowledgments. It causes congestion window to slowly increase which leads to TCP throughput degradation. Improving TCP performance, modes should temporarily give up the PSM when there exists TCP traffic in the network.