Distributed Priority Scheduling and Medium Access in Ad Hoc Networks Distributed Priority Scheduling and Medium Access in Ad Hoc Networks Vikram Kanodia E.C.E Rice Univ Houston TX Chengzhi LI C.S Univ of Virginia Asutosh Sabharwal,Bahareh Sadegi,Edward Knighty E.C.E Rice Univ Houston Presented by Abhijit Pandey

Outline Introduction Distributed Priority Scheduling Multi Hop Co-ordination Related Work and Conclusion

Key insight To Utilize the broadcast nature of the medium Store and Forward nature of Multi- hop network Communication and co-ordination of priority information among nodes

Priority Backoff schemes to approximate the idealized schedule. Packet to satisfy end to end quality of service.

Distributed Priority Scheduling A technique that piggybacks the priority tag of a node’s head of Line packet onto handshake and data packets. RTS/Data By monitoring transmitted packets each node maintains a scheduling table into existing Scheduling Table is estimate of its relative priority into medium access control

Methodology Each node issues a Request to Send(RTS), it piggybacks the priority index of its current packet A CTS granted contains priority index of its head of line of the data packet. This is inserted into the table of overhearing nodes. Each node assess the priority index of its own head of line packet, and with prioritized backoff schemes a distributed priority schedule is obtained

Improvement over Distributed Priority Scheduling With probability q=60% of nodes overhearing, the mean delay is reduced from 2.86sec (802.11)to.6 sec Co-ordinated Multi hop scheduling Co-ordination decreases the average delay by 60% as compared to and 25% as compared to distributed priority scheduling without co-ordination.

Scheduling Algorithm In Ad-hoc networks to satisfy packet’s quality of service becomes increasingly difficult Earliest deadline First Packet has a priority index given by arrival time plus its delay bound. This priority can be maintained by base stations.

Distributed Priority Scheduling Packets are serviced in increasing order of priority index. In EDF a packet arriving at time t and having delay bound d has priority index t+d. A packet with size L with service rate r has a priority index of L/r.

Mechanism Due to distributed nature of ad hoc wireless networks Each node is equipped with its own buffer state and partial information about other nodes. The scheduler is distributed with incomplete system information

I.E.E.E distributed coordination function

Distributed Co-ordination function If the channel is sensed idle for a duration of DIFS the node generates a random back off interval before transmitting the packet. The RTS/CTS have information regarding the destination node and the length of the data packet to be transmitted. Any other node which hears either the RTS or CTS can use the data packet length to update its network allocation vector containing the information of the period the network will remain busy

Backoff Timer Contention Window The backoff timer is chosen uniformly from the range[0, w-1] W is the contention window. At the first retransmission attempt w is set to CW min After each unsuccessful transmission the value of w is doubled upto the max value CW max = 2^ m CW min

Piggybacking on IEEE four-way handshake, and the updating of scheduling tables.

Priority Broadcast Hidden nodes which are unable to hear the RTS add an entry in their scheduling table upon hearing the CTS The receiving node appends the priority in the CTS frame. Each node after hearing data packet adds another entry in its scheduling table. Upon successful transmission and Ack, each node removes the current packet from the scheduling table

Simulation Experiments A single broadcast region with link capacity 2Mb/s and data rate of 1.6 Mb/s Each node carries variable rate traffic according to exponential on-off model. Upon receiving a piggybacked RTS, a node enters the priority index into its local scheduling table with probability q.

Delay versus available information

No of collisions versus available information

Probability of correct scheduling vs. number of nodes for different values of q. Increase in probability of correct scheduling as q increases Significant gain even for lower values of q q=0 q=.6 q=.4 q=.2 q=.8 q=1

Multi Hop Co-ordination Downstream node can increase a packet’s relative priority to make up for delays upstream Analytical model to study the probability of overhearing another packets priority index.

Multi Hop Co-ordination All nodes co-operate to provide end to end service. Priority expressed recursively. The index of each packet at its downstream node depends on its priority index at its upstream node. If a packet arrives early downstream node will reduce the priority of the packet and vice versa.

Priority Index assignment schemes Time to Live allocation –Priority of packet increases with time spent in the network –Flows can be differentiated by assigning different TTL’s Fixed Per node allocation –Each node has a certain fixed increment of priority index. Uniform delay budget allocation –The increment of Priority index is D/K Where D =end to end delay target K= no of hops from routing table.

Probability of satisfying end-to-end delay target under different priority schemes Multi-hop coordination IEEE Single hop scheduling

Simulated delay performance of multi-hop coordination.

Conclusion A scheme where priority index of head of line packets is piggybacked onto existing messages. Downstream can make up for latencies upstream by multi hop co-ordination. Co-ordination an important ingredient for targeting end to End QOS. Moderate fraction of piggybacked message overhead

Important Aspects of this paper This paper addresses three fundamental issues of providing Quality of Service in Ad-hoc networks 1Distributed priority scheduling 2Priority based medium access. 3 Multi-hop priority management.