(Slides) A demand-oriented information retrieval method on MANET

2,823 views
2,914 views

Published on

Enomoto, M., Shibata, N., Yasumoto, K., Ito, M. and Higashino, T.: A demand-oriented information retrieval method on MANET, International Workshop on Future Mobile and Ubiquitous Information Technologies (FMUIT'06).

http://ito-lab.naist.jp/themes/pdffiles/060510.makoto-e.fmuit06.pdf

In urban areas including shopping malls and stations
with many people, it is important to utilize various information
which those people have obtained. In this paper, we
propose a method for information registration and retrieval
in MANET which achieves small communication cost and
short response time. In our method, we divide the whole application
field into multiple sub-areas and classify records
into several categories so that mobile terminals in an area
holds records with a category. Each area is associated with
a category so that the number of queries for the category
becomes the largest in the area. Thus, mobile users search
records with a certain category by sending a query to nodes
in the particular area using existing protocol such as LBM
(Location-Based Multicast). Through simulations supposing
actual urban area near Osaka station, we have confirmed
that our method achieves practical communication
cost and performance for information retrieval in MANET.

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
2,823
On SlideShare
0
From Embeds
0
Number of Embeds
1,843
Actions
Shares
0
Downloads
8
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide
  • “ A Demand-Oriented Information Retrieval Method on MANET” presenter is Makoto ENOMOTO from Nara Institute of Science and Technology
  • Background of the Study is, “ Mobile terminals’ processing power is rapidly increasing.” “ These devices now have wireless communication capabilities such as IEEE 802.11 and Bluetooth.” We came to think “ It is desired to utilize MANET to allow people to exchange various information which those people have obtained as if MANET is a database system.”
  • Purpose of the Study. Evaluated Criteria are -short average response time(from sending a query to receiving a first reply) -small traffic amount Policy is -Maintenance free, stateless routing -no concentration of load on particular terminals This leads us to aim to realize an efficient method which satisfies these criteria.
  • Contents of Study are as follows …
  • Information retrieval system from user’s perspective when user register data item. We call a data item a record, hereafter. A record is combination of ID,timestamp,location ,main data , main data is like text,images and so on,keyword, [prices,… ] ,prices and other auxiliary fields can be omitted.Records sample are as follows. User register a record from his or her own terminal. The record is transferred by some other nodes and reaches the designated area that preserves the record.
  • When user search a record, searching example is like this, he or she sends a query from his or her terminal, the query is transferred through some other nodes. The query finally reaches the nodes that have requested records. Reply messages are sent back.
  • Here is the assumption we made …
  • Problem definition is as follows, Assume that query frequency varies widely according to location and category. Problem is to find the distribution of records that minimizes the average response time This system is assumed to use dynamically changing information that static web sites don’t have, such as happy hour of the shopping store, train delay, available restaurant, congestion at event site
  • Basic ideas, the approach is that “ If all of registered records are replicated in all nodes, the average response time becomes minimum”, but in this case “replication cost also becomes high.” In the proposed method, we replicate records only in the nodes in the area with the largest demand. In this figure, target region is divided into sub-areas, if this area is with the largest demand, we replicate records in this area.
  • Proposed method as STEP1 is classifying records into classes. All records are divided into multiple sets by keywords that are included in records, such as On Sale, Event, Train Delay
  • Proposed method as STEP2 is that whole application field is divided into sub-regions, hereafter area s This is a map around Osaka station. We can divide the region in an arbitrary way. But for simplicity, we divide the region into squares. We name each sub-areas such as ‘A1’,’B1’, and so on.
  • Proposed method as STEP3 is that All nodes count the number of matched queries for each class. Demand is measured by the count of query sent to the area Like these, each area counts queries by class.
  • Proposed Method as STEP4 is to find the most demanding area. This step is the key idea of the proposed method. This is the example. First, Area ‘A1’ flood the its query count table. Second, Area ‘A2’receive the table. Third, Area ‘A2’ compares entries of the table and its own table, and updates the table so that each table entry has higher value between two tables and flood again. In this way, after all areas update the table, all nodes know which areas are the most demanding areas.
  • Proposed Method as STEP5 is to replicate records in the most demanding area . This is the Example. User registers a record which ClassID = 3 from this node If most demanding area of the Class is ‘C1’, This node sends a registering message to ‘C1.’ When sending the message, using the communication protocol that can send a message to designated area like LBM, nodes inside these curves receive the message, and then transfer to the next node. Replications can be made in the second most demanding area to ease sudden change of the demand.
  • Proposed Method as STEP6 is to send query to the most demanding area The method of sending a query and receiving a reply is the same as that of sending a data. The example is that A node in ‘A2’ sends a query to ‘C1’ like this. Records that matched the query are sent back, and nodes in the area in which the query-sending node is receive the result. In case a node in C3 that is the most demanding area send a query, No message is made, because replications are made in the area.
  • Communication protocols used in our implementation are the following:
  • Experimental Result. This is the Average Response Time as the number of node varies. Our proposed method takes most demanding. For comparison, least demanding and flooding. Least demanding means records are replicated in the least demanding area, and Flooding means the communication is performed using flooding protocol. Most demanding show the shortest, as compared to the other two.
  • This is the number of messages. Proposed method needs the least messages, compared to the other two .
  • This is the Reply Loss. Proposed method shows the least reply loss as the number of nodes increases.
  • (Slides) A demand-oriented information retrieval method on MANET

    1. 1. A Demand-Oriented Information Retrieval Method on MANET Makoto ENOMOTO , Naoki SHIBATA † , Keiichi YASUMOTO , Minoru ITO , Teruo HIGASHINO † † Graduate School of Information Science, Nara Institute of Science and Technology † Department of Information Processing and Management, Shiga University † † Graduate School of Information Science and Technology, Osaka University Presenter: Makoto ENOMOTO
    2. 2. Background <ul><li>Mobile terminals’ processing power is rapidly increasing. </li></ul><ul><li>These devices now have wireless communication capabilities such as IEEE 802.11 and Bluetooth. </li></ul>It is desired to utilize MANET to allow people to exchange various information which those people have obtained as if MANET is a database system.
    3. 3. Purpose of the Study <ul><li>Evaluated Criteria </li></ul><ul><ul><li>short average response time(from sending a query to receiving a first reply) </li></ul></ul><ul><ul><li>small traffic amount </li></ul></ul><ul><li>Policy </li></ul><ul><ul><li>Maintenance free, stateless routing </li></ul></ul><ul><ul><li>no concentration of load on particular terminals </li></ul></ul><ul><ul><li>This study aims to realize an efficient method which satisfies these criteria. </li></ul></ul>
    4. 4. Contents of Study <ul><li>User’s Perspective </li></ul><ul><li>Assumption and Problem Definition </li></ul><ul><li>Basic Ideas </li></ul><ul><li>Detailed Proposed Method </li></ul><ul><li>Experiment and Evaluation </li></ul>
    5. 5. User’s perspective ( when register ) <ul><li>Data item( hereafter record ): </li></ul><ul><li>combination of </li></ul><ul><li>    ID , timestamp , location , main data(text , images),keyword, [prices ,… ] </li></ul><ul><li>Record example:   (0,19:00,135.29-34.41,“Yamato”,“Japanese Style Restaurant, Happy Hour”,-) </li></ul>Registering record Preserving record
    6. 6. User’s perspective ( when search ) Searching Example :   Time = 19:00-20:00,   Location = less than 500m from Osaka Station,   Category = Japanese Style Restaurant Searching record Preserved record
    7. 7. <ul><li>Mobile Terminal ( laptop PC,PDA,cell phone ) </li></ul><ul><ul><li>has wireless communication capabilities. </li></ul></ul><ul><ul><li>has some memories. </li></ul></ul><ul><ul><li>is carried by a person and therefore each node moves. </li></ul></ul><ul><ul><li>knows its geographical position by GPS. </li></ul></ul><ul><ul><li>has a unique ID number. </li></ul></ul><ul><li>The geographical region where the proposed method is operated is predefined . </li></ul>Assumption
    8. 8. Problem Definition <ul><li>Assume that query frequency varies widely according to </li></ul><ul><ul><li>location  ( station, event site, … ) </li></ul></ul><ul><ul><li>category  ( traffic , shopping, … ) </li></ul></ul><ul><li>Find the distribution of records that minimizes the average response time </li></ul>available restaurant Train delay info congestion at event site Shopping info, happy hour
    9. 9. Basic Ideas <ul><ul><li>If all of registered records are replicated in all nodes, the average response time becomes minimum. </li></ul></ul><ul><ul><li>  replication cost also becomes high </li></ul></ul>In the proposed method, we replicate records only in the nodes in the area with the largest demand Approach :
    10. 10. Proposed Method STEP1: Classify records into classes Record: ID , timestamp , location , main data(text , images),keyword, [prices ,… ] All records are divided into multiple sets Class1: On Sale Class2: Event Class3: Train Delay
    11. 11. Proposed Method STEP2: whole application field is divided into sub-regions (hereafter area s) A B C 1 2 3 A1 B1 A2 A3 B2 B3 C1 C2 C3
    12. 12. Proposed Method STEP3:All nodes count the number of matched queries for each class Demand ; measured by the count of query sent to the area Count Class 10 Train 11 Event 23 On Sale Count Class 17 Train 7 Event 4 On Sale Count Class 21 Train 86 Event 70 On Sale Count Class 1 Train 11 Event 5 On Sale Count Class 3 Train 10 Event 8 On Sale Count Class 32 Train 119 Event 54 On Sale Count Class 16 Train 11 Event 32 On Sale Count Class 9 Train 11 Event 66 On Sale Count Class 23 Train 150 Event 68 On Sale
    13. 13. Proposed Method STEP4:Find the most demanding area A1’s query count table (i) Area ‘A1’ flood the its query count table (ii)   Area ‘A2’ receive the table (iii) Area ‘A2’ compares entries of the table and its own table, and updates the table so that each table entry has higher value between two tables and flood again A2’s query count table 3 10 8 Count Train Event On Sale Class A1 A1 A1 Area Area Count Class 3 11 8 Train Event On Sale A1 A2 A1 Area Count Class 1 11 5 Train Event On Sale A2 A2 A2 C3 B3 A3 C2 B2 A2 C1 B1 A1
    14. 14. Proposed Method STEP5:Replicate records in the most demanding area A B C 1 2 3 Data sending node Example 1.registering a record which ClassID = 3 2.most demanding area of the Class is ‘C1’ 3.Sending registering message to ‘C1’
    15. 15. Proposed Method STEP6:Send query to the most demanding area A B C 1 2 3 Query sending node Example A node in ‘A2’ sends a query to ‘C1’
    16. 16. Experiment and Evaluation <ul><li>The target geographical region </li></ul><ul><ul><li>500m × 500m region around Osaka Station </li></ul></ul><ul><li>Distribution of Nodes </li></ul><ul><li>Nodes are placed on the roads according to the actual observed density. </li></ul><ul><li>For simplicity, nodes are assumed not to move. </li></ul>
    17. 17. Experiment - implemention Protocols Flooding used when sending a query or a record to nodes in a particular area used when a message is sent to the entire ad-hoc network Location-Based Multicast (LBM)
    18. 18. Experimental Result – Average Response Time
    19. 19. Experimental Result – Number of Messages
    20. 20. Experimental Result – Reply Loss
    21. 21. Conclusions <ul><li>We have proposed an information retrieval method on MANET that achieves low communication traffic and short response time . </li></ul><ul><li>Simulation using actual geographical information with measured pedestrian density in Osaka city. </li></ul>Future Works <ul><li>Estimating the overhead of control messages for finding the most demanding areas and record replication cost </li></ul><ul><li>Simulations under the conditions where pedestrians move along the roads with realistic pedestrian density on each road </li></ul>

    ×