Wireless Ad-Hoc Real-Time Multicast Protocol (WARM)

• Streaming of real-time multicast traffic in wireless ad-hoc networks is a challenge

• Simultaneous transmissions by nodes with common multicast member neighbors will result in collisions at those common neighboring nodes.

• An intelligent scheduling mechanism is required. Furthermore, as nodes move, the scheme has to maintain the multicast structure for collision free streaming.

• This is similar to providing QoS – Integrated Services reserved bandwidth along a path from the source to a receiver.

Reference

• G.D. Kondylis et al, “Multicasting Sustained CBR and VBR traffic in Wireless Ad Hoc Networks”, Proceedings of the IEEE International Conference on Communications (ICC), 2000.

• On my website.

• Wireless channel is broadcast in nature– One transmission can reach many

recipients: ideal for multicasting• However, simultaneous transmissions of

adjacent relays can lead to collisions– In a multicast scenario, ACKs cannot be

used: because of the multiple receivers the same problem of collisions exists at the transmitter

• The only way to achieve reliable communications (at the MAC layer) is to reserve bandwidth for the multicast session

• Let us assume that the network size is moderate.

• The problems of synchronizing nodes etc. are taken care of.

• Why is this assumption required ?• Reservations imply we use either a TDMA or

FDMA based method (sometimes CDMA).

Assumptions in WARM

• A single source is transmitting to a multitude of receivers.

• The receivers are told about the session via some advertisement or otherwise.

• The onus is on the receiver to join the multicast session.

• We have the bandwidth divided into channels – either in time (it could be frequency or code but..).•There is bandwidth set aside for the multicast session (which has to be spatially re-used appropriately).•Signaling is done in collision-free manner (a round-robin scheme is assumed)• Half-duplex transceivers – nodes cannot transmit and receive at the same time.

• If nodes 2 and 3 (or 1) transmit simultaneously, node 6 will see a collision

• Therefore, node 2 will have to use different “bandwidth” (TDMA slot / CDMA code / FDMA frequency band)

• That’s a problem of coloring the graph with the minimum possible number of colors (color = TDMA slot, e.g.)

•Joining multicast members should try to attach to already existing relays, before forcing other members to become relays

–they should try to find colors already used by relays, where they can receive reliably

•Relays have a set of “unusable” colors: colors that if used to transmit will cause collisions at some receiver.

• This is with CBR in mind.

• If the traffic is VBR, a set bandwidth is allocated and packets in excess will have to be relayed via random access.

Main Principles of WARM

The TDMA frame format

p1 p1 p0

0 1

Reserved

Random Access

p0 p1

Reserved

Random Access

0 1

Super-frame

Transmit Part Receive Part

• Multiple transmit slots may be reserved. In this packet zero is received in slot 0 and packet 1 is received in slot 2. It transmits packet 1 in slots 0 and 1 (possibly for different children) and packet 0 in slot 6.

• Packets in a frame that are in excess of the reserved number of slots are transmitted/received in the random-access portion of the frame.

• This would happen in the case of the VBR traffic.

• If reservation = mean source rate, then anything in excess of this mean source rate will have to be transmitted using the random access slots.

• Parent node (in the multicast mesh) will inform its children of the specific random-access slots that they will have to listen to by appending the relevant info to packets transmitted during the reserved portion of bandwidth.

• If other nodes can overhear this, (not always possible) they could potentially avoid these slots.

Signaling

• Transmit scheduling for signaling between nodes is done in a round robin fashion on a separate channel.

• This is not a requirement – if routing or other info is to be transmitted they need a control channel MAC-routing interactions.

• Signaling messages usually carry

•The node’s unique ID,

•The hop-count from the source (which helps a node to determine who to connect to)

• A bit to indicate whether it is a transmit frame or receive frame that is to follow.

• List the transmit and receive slots for the node (refreshing the reservations : soft state).

• List the parents of the node. It is required that all parents be at the same level from the source.

• Slots on which the node is precluded from transmitting or receiving to children (to parents).

• This would essentially help when a node is searching for a new parent.

JOINing the Multicast Mesh and Maintenance

• Depending upon what is transmitted in the Unusable Slots which is in fact how collisions (let us qualify what these mean later) are avoided a node can choose its parents (maybe more than one) to receive multicast data from the session.

• If links break due to mobility (slots get jammed) then the receiving node can try and re-establish connectivity by using other usable slots or try and connect to other parents.

• During the time of re-establishment, soft state empty packets may be sent to children so that they do not disconnect from the mesh thereby causing a chain reaction.

• If the links are not re-established within certain time, then the children of the node who loses connectivity might have to attempt re-establishment and so on.

The Metrics for QoS

• To re-iterate we are trying to stream real-time data.

• For each session we set aside certain bandwidth that corresponds to its mean rate or peak rate (certain number of slots).

• Packet losses may occur due to mobility or due to losses for VBR sources in the random access channels.

• The performance is quantified in terms of throughput :percentage of received packets. If the structure is to be re-built often it results in a large packet loss.

• The mesh structure allows nodes to identify new parents quickly when links fail.

What is a collision ?

• A collision is when the interference level is much higher than the required signal level.

• One way of modeling collisions would be that if two nodes transmit such that they are both in the range of a third node, there is a collision at that node.

• However, in reality, because of multiple distant transmissions, the interference level at a node due to the distant transmissions might be significant to cause degradation collision.

• In the following example such effects aren’t considered but are taken into account in the simulation studies reported in the paper. (collision if SIR < ).

Example(one hop interference only)

• One color (slot) is not enough to fully connect the multicast session

• Two slots (colors) can connect the session (but perhaps not every node is at the minimum hop count from the source.

Example of Results(full co-channel interference)

• When source is mobile, packet dropping is more severe and is affected by speed more that when source is static. Can you guess why ?

Power Management• Power Control is popular in cellular networks.

•This would typically require that all nodes transmit such that the received power at the base station is the same.

• It has been shown for both CDMA and TDMA systems that power control helps in reducing interference and improving throughput.

• What is power control or management in an ad hoc network ?

• Difficult to do no centralized control.

• Benefits Improvement in throughput, reduction in power consumption.

• Disadvantages Creating unidirectional links : depends on density.

Reference

• T.ElBatt et al, “Power Management for Throughput Enhancement in Wireless Ad Hoc Networks”, Proceedings of the IEEE International Conference on Communications (ICC), 2000.

• My Web Site.

The Concept of Connectivity

1Node A Node D

Node C

Node B

Node E

Range ofNode A’s Transmission

Range ofNode D’s Transmission

Connectivity of Each Node is 3.

The Key Idea• Each node varies its transmit power so as to define its connectivity.

• What does this mean ?

• Instead of using the constant power level as was done in all our previous discussions (except those on unidirectional links) a node uses the power level such that it reaches only a fixed set of neighbors.

• What is the advantage ?

• The zone of interference is reduced. Thus, frequency re-use is achieved.

• What is the disadvantage ?

• A packet has to potentially travel a large number of hops to reach a destination – does this increase bw consumption ?

The framework

• Time Division Multiplexed Access with round-robin slots for transmitting signaling information.

• Signaling Messages transmitted at maximum power carry routing and connectivity information.

• Actual Data is transmitted by using Slotted Aloha medium access control protocol.

• Power Based Routing is deployed. Two schemes investigated:

• Same power within connectivity domain

• Different powers for different nodes within the domain, I.e., power adaptability at a fine grained level.

How do we form domains ?• Remember, each node has its own slot to transmit signaling information.

• This control info is transmitted at the largest power possible.

• What is the received power ?

• If transmit power is Pt, received power is

Pr = (/d4) Pt.

• Here, is the shadowing co-efficient and the path loss is in accordance to a fourth power law.

• Why is multi-path fading not considered ?

• Each node records the received power from each of its neighbors within its maximum power range.

• It orders these neighbors in accordance to an ascending rule.

• It then identifies the number of neighbors to be included in its connectivity zone (K).

• Thus, the domain is formed.

• We try two types.

• One possible way is to use just the power level required to communicate with each node.

• The second is to use the same power level for all nodes within the domain.

• The first (intuitively) decreases interference further and increases the number of hops (perhaps).

• The second makes links more robust and possibly reduces the number of hops to destination (by a little) but increases interference.

Power Control

Are Unidirectional Links Formed here ?

• The answer is well.. Yes and No.

• What does this mean ?

• Notice that the control info is always carried on bi-directional links.

• This allows the formation of routing tables – all links are bi-directional as far as control signaling goes.

• Data is however carried on unidirectional links.

• We use the aloha protocol at MAC layer – thus it is not very good.

• RTS/CTS might be used but they have same problems that we saw earlier.

Metrics of Performance

• Average Power Consumption at a node.

• Node throughput -- % of packets that were successfully transmitted.

• End to end network throughput -- % of packets that actually reached their intended destinations.

• Network Throughput is maximized at an interim power level. A gain of nearly 8% in throughput.

• When power adjustment is done within a cluster, the maximum is reached at a different point than when no power adjustment is done within the cluster.

• This optimal connectivity range can be dynamically searched for.

Example Results

• Higher power savings (of the order of more than 50 milliwatts on the average) when power management is done.

• Lower connectivity range results in lower power consumption.

• From previous result and this one, we can see that when there we do power adjustments for reaching neighbors within a cluster, we maximize throughput at lower power.

Other forms of Power Management

• Power management to ensure that network does not get partitioned. Use enough power to ensure complete connectivity by Ramanathan and R.Rosales Hein (BBN), Infocom 2000.

• This is another paper on topology control.

• Power Aware Routing allow usage of minimum energy path. Allow nodes to sleep when they are not participating in routing – S.Singh and C.S.Raghavendra, “Power Aware Routing in Mobile Ad hoc Networks”, in Proceedings of ACM Mobicom, August 1998, Seattle .

• SPAN in Mobicom 2001 – allow nodes to go to sleep when they are in dense parts of the network – power savings. by Hari Balakrishnan MIT.