Lab Seminar 2009 07 08 Message Drop Reduction
- 1. GITS Seminar Description Based addressing and routing in Cluster-based Ad-Hoc network for Home Environment By Tharinda Nishantha Vidanagama* Supervised by, Prof. Hidenori Nakazato* *Graduate school of Global Information & Telecommunication Studies, Waseda University. 1
- 2. Ad Hoc Networks The Ad Hoc peer network is created spontaneously as two participating nodes come within the reach of one another. As the number of participating nodes and their positions vary the network it self reconfigures allowing all nodes to communicate even if they are not within their transmission range. The main advantage of Ad Hoc networks is that they do not require any infrastructure to operate. 2
- 3. Ad Hoc Networks But the main disadvantages are Highly dynamic topology due to the mobility of the nodes, i.e. the links are not stable and bandwidth is limited. Mobile devices are powered by batteries, putting an energy consumption constraint Security of the network is at risk as anyone can join the network and act as transmission mediators. 3
- 4. Ad Hoc Networks Increasingly wireless technology is build in to many devices. Networking these devices can bring many benefits. Range of application is vast. Military Disaster Service extension. 4
- 5. Description based Ad Hoc Networks Communication failures are considered normal in Ad-hoc networks. Description based addressing instead of host based(IP) addressing is more appropriate because it will not rely on a single path, but uses multicasting . In description-based routing, selective information dissemination can be implemented. Easy to understanding and manage the network for home environment. This will minimize control data among peers and facilitate group communication. 5
- 6. Ad Hoc Network at Home ! Increasingly new functions are built in to home appliances. Home appliances fitted with wireless communication modules would become more common in near future. The system for home networking provides information required for inter- working environment for home appliances. The nodes in this environment could also have the following constraints, Low memory Low transmission distances Some nodes may have limited power supply. 6
- 7. Ad Hoc Network at Home ? What about the Ad Hoc network disadvantages? Household appliances are mostly static i.e. they may not move at all. Therefore the problems arising due to mobility is very limited. Almost all appliances are plugged in to an unlimited power supply, therefore the power limitations doesn’t apply. The whole network is owned by one owner, there is no security risk within the network. 7
- 8. Add Hoc home network - Approach 1. Cluster Based Ad Hoc Network 2. Description Based Routing and Addressing 8
- 9. 1. Cluster Based Add Hoc Network Why? At home most appliances are clustered in places. Kitchen Living room Bed room etc. Clustering also reduces the amount of control data exchanged among the nodes. Only a portion of the nodes are involved in the actual routing process. 9
- 10. Cluster Based Add Hoc Network Organization Everyone knows its one hop neighbors (Neighbor Table). Cluster heads know their neighboring cluster heads (Cluster Adjacency Table). Member nodes send their messages to their respective cluster heads to be forwarded to their destinations. Only cluster heads and gateways forward messages. 10
- 11. Problem of Routing Does a cluster head know about all other cluster heads? This requires high memory capacity ! How to route messages beyond known cluster range? A fixed size memory cache (buffer) is used to keep track of the nodes that are beyond range. Rather than using a table structure this helps reduce the memory usage. There is no need to store the full identities or routs. 11
- 12. 2. Description Based Routing and Addressing We assume that identifiers given to the nodes in the network i.e. home appliances are similar to the following examples. Kitchen oven, Kitchen television, Living room television etc. In this natural language usage, a general location is identified from the beginning words and becomes more specific within that particular area by the next words. Our routing algorithm routes the data messages in the direction of the destination. When the message finally arrives at the particular region, there would be nodes that store the exact identifier in their Neighbor table, thus allowing the message to be delivered to the correct destination node. 12
- 13. Buffer Usage How the buffer is used? If a cluster head or gateway node comes across another cluster head that is beyond its range, it inserts the name in to the buffer. E.g: if the required destination is “kitchen oven” but the buffer entry may be “kitchen”. This partial match is enough to ensure that the message gets routed in the direction of the kitchen. If the entry is in the 1st position of the buffer it would be as follows. 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 k i t c h e n 13
- 14. Buffer storage For the simulation purpose we use a fixed size character buffer. Elements are inserted in a way that some of the information about old entries are overwritten by new entries. If S - Buffer size p - Number of parts the buffer is divided in to P - The next position in the buffer to be occupied B - 1 if data written backwards (If entry number is and odd number) F - 1 if data is written forward (If entry number is an even number) f - Fibonacci series number. When an entry is made in to the buffer the starting position (T) for an entry is given according to the following formula. Starting Position T= ( S / p) *( P − B ) + F 14
- 15. Buffer storage The number of characters allocated for an entry in the buffer is given by the following formula, Number of characters = (s / p) − 1 The actual character positions are given by a modified version of the Fibonacci series (0, 1, 2, 3, 5…), where we have omitted the repetition of number 1. Character position in the buffer is given by the following formula, Character Position = T + f ; F =1. { T − f ; B =1. 15
- 16. Description Based Routing When a node receives a message, If destination is this node Consume data Else if I am Gateway/Cluster head and if this data has not already been forwarded Look for destination in neighbor table If not found Look for destination in Cluster Adjacency table If not found Look for partial match in Cluster Adjacency table If not found Look for partial/complete match in the Buffer. If found Forward message Else Discard message. 16
- 17. Evaluation A simulation program was developed in order to evaluate the above proposal. We have used an environment, which is a flat surface 25*25 square units, where 100 nodes are statically located. The size of the buffer was limited to 128 characters. Each node has various transmission distance (maximum of 3 units) and priorities. Each node will send a number of messages to randomly selected destinations. 17
- 18. Evaluation Average data message drop due to wrong routing Single Buffer Average drop Success 220 200 180 160 187.4 183.6 185.4 193.1 182 181.6 182.9 182 182 179 140 Percentage 120 100 100 100 100 100 100 100 100 100 100 100 80 60 40 20 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 Data messgaes It shows on average a 100% success rate on message delivery with the current environment setting. But… message drop is high !!!. 18
- 19. Evaluation – The drop? The message drop occurred because of the use of single buffer. Buffer entries do not store the direction of that particular entry. Therefore the messages maybe forwarded to the wrong way as well. 128 CH 128 128 CH GW 128 CH Bidirectional link CH Cluster Head 128 Buffer GW Gateway 19
- 20. Solution – Directional buffers Routing nodes are given a buffer for each outward link. The nodes can now compute the next hop of the messages in the right direction. CH 128 12 8 128 128 CH GW 8 12 8 12 CH Bidirectional link CH Cluster Head 128 Buffer GW Gateway 20
- 21. Simulation – Directional buffers Average data message drop due to wrong routing Directional Buffer (Equal-sized) Average drop Success 120 100 100 100 100 100 100 100 100 100 100 100 Percentage 80 60 35.3 34.6 35.8 33.6 35.1 36.2 38 31.8 29.8 32.3 40 20 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 Data messgaes The message drop due to wrong routing is significantly reduced by the directional buffers. But the Memory usage is High !!! 21
- 22. Solution: Proportional Directional buffers The buffer is divided proportionally between each bidirectional link. Initially full size given to first bidirectional link with a buffer entry. If another link has a buffer entry the buffer is proportionally divided among the links according to the number of entries in each direction. 22
- 23. Simulation: Proportional Directional buffers Average data message drop due to wrong routing Proportional Directional Buffer Average drop Success 120 100 100 100 100 100 100 100 100 100 100 100 Percentage 80 60 35.6 35.3 34.8 35.8 36.7 33.9 36.4 34.4 33.5 40 29.9 20 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 Data messgaes 23
- 24. Equal-sized Vs Proportional directional buffer Average Data Message drop due to wrong routing Proportional Directional Buffer Vs Same size buffers Proportional directional buffer Equal-sized directional buffer 36.2 38 40 35.3 34.6 35.8 35.1 33.6 32.3 31.8 29.8 35.6 35.3 35.8 36.7 36.4 30 34.8 33.9 34.4 33.5 Percentage 29.9 20 10 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 Data messages The performance is similar to the equal-sized directional buffers 24
- 25. Directional buffers – remaining message drop When a message is received by a node that has many outward links, it checks each buffer for a match with the destination. If more than one buffer is positively matched the node has to broadcast the message to all outward links. Thereafter only the next hop nodes who know the destination will forward the message, and the others will register a lost message. 25
- 26. Simulation Comparison Data Message loss Comparison % at 100% success rate Broadcasting Single buffer Directional buffer 1300 1157.8 1141.1 1145 1139.2 1151.6 1158.93 1146.37 1142.6 1162.42 1159.44 1200 1100 1000 900 Percentage 800 700 600 500 400 300 187.4 179 183.93 180.9 185.6 190.33 191.23 174.53 191.09 167.88 200 100 0 26.4 34.6 34.8 35.05 36.44 30.47 39.34 33.48 30.44 45.76 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 Data messages The loss due to wrong routing is significantly reduced by the directional buffers. 26
- 27. Conclusion Considering that a home environment does not contain large number of nodes, this algorithm can be effectively implemented. The use of clustering has further supported in reducing the network traffic and routing overhead. Use of descriptions improves the user understandability of the system. Use of directional buffers will significantly reduce the message loss and improve resource usage. Proportional directional buffer is the solution with most preferable settings. 27
- 28. Future extensions We will study the performance and issues when movement is introduced to the network nodes. In the home environment elements such as cell phones and laptops etc, will have movement associated with them. We would also like to find the optimal buffer size for a given number of nodes. Further decrease the amount of control information held and transmitted by nodes. 28
- 29. Questions ? 29
- 30. Thank you ! 30