A scalabilty and mobility resilient data search system

33
WELCOME 1

Transcript of A scalabilty and mobility resilient data search system

A Scalabilty And Mobility Resilient Data Search System For Large-Scale Wireless Network

WELCOME1

A Scalabilty And Mobility Resilient Data Search System For Large-Scale Wireless Network

ALEESHA NOUSHADROLL NO : 6 S7 CSE 2

CONTENTIntroductionLORD Data Search SystemFile Querying in LORDArea PartitionRegion Based Data Publishing and QueryingRegion Based Geographic RoutingParallel File Fetching algorithmBack-Tracking AlgorithmColouring Based Partial ReplicationPerformance EvaluationConclusionReferences

3

INTRODUCTIONRecent technical advancements have enabled the development of large scale wireless network.

An important issue in recent wireless networks is data search.

To build a scalable and mobility-resilient distributed data search system for large scale highly mobile and dense wireless networks,we propose

A LOcality based distRibuted Data Search System (LORD).

4

LORD Data Search SystemLORD divides the entire wireless network area into number of geographic regions.

LORD has a Region-based Geographic Routing (RGR) protocol for data publishing and querying.

LORD uses * Parallel File Fetching Algorithm. * A Back-Tracking Algorithm. * Coloring Based Partial Replication Algorithm. 5

Rgr needs only region angle infrmtn.file reqstr use ||,5

File Querying in LORD Steps of File Querying in LORD

Metadata Publishing.Metadata Querying.Metadata Replying.File Querying.File Replying.6

File Querying in LORD

7

Area Partition LORD divides the entire area into a number of regions.

A Region is the neighboring zone in transmission range of a landmark and centered by the landmark.

Each region is identified by an assigned integer number.

Number of landmark is determined by a) Transmission Range. b) Size of entire area. 8

Region Based Data Publishing and Querying

LORD uses Hash function to hash a file to store the metadata.

A files keywords can be its file name.

A File host publishes the metadata to the mapped Region.

To query a file , it calculates the co-ordinate of the files metadata and sent request to destination.

9

Data Mapping Update And Location Update

A Node conducts updates only when it moves from one region to another.

It drops all old regions metadata. Accquire all metadata in the new region from new neighbor.

Conducts location updates by sending message to mapped regions of its files metadata.10

Region Based Geographic Routing LORD propose Region Based Geographic Routing(RGR) to forward messages to their destination.

RGR Routes the message along a shortest path

RGR only needs Angle Of Arrival(AOA),which will be the tight bound of message transmission.

A landmark always located in the center of region.

11

Ri -Source region.Rj -Destination region.

Angle between leftmost vertex of Ri to the leftmost and rightmost vertices of Rj is leftside angle range[,]

Angle between rightmost vertex of Ri to the leftmost and rightmost vertices of Rj is rightside angle range[,]

12

Routing to destination region.

When a node initiates or receives a message ,it calculate its regions LR and RR.

If the node is left of the landmark.It choose the farthest node within the [,] as next hop.

If the node is right of the landmark.It choose the farthest node within the [,] as next hop.

13

Routing in destination region.

If the message is - metadata query :- continously forwarded. -metadata publishing :- broadcast to all nodes. -metadata reply :- target to only file requester. -file query :- target the file host. 14

Parallel File Fetching Algorithm

This algorithm transmits different file segments simultaneously from one file holders to file requester.

Transmitting segment wise reduces the transmission latency.

File requester choose the geographically close file holders.

Ask each file holder to transmit a segment of the file.15

15

Back-Tracking AlgorithmTo keep track of the requesters movement.

If a requester moves out its current region before receiving the response , it sends a back-tracking message to source region.

Message is piggybacked on the hello message between neighbour nodes.

Thus each node in the source region keeps a backtracking message of the requester.16

Pseudo-code for metadata request replying. // Sending a back-tracking message; while have not received the metadata reply do if it moves to a new region then Send a back-tracking message to its old region end ifend while // Receiving a back-tracking message;if receive a back-tracking message then Add the message to its back-tracking message list Broadcast the message to its neighbors in the regionend if17

17

// Receiving a metadata reply;if receive a metadata reply with its region as destination then if its back-tracking message list contains the message from the requester then forward the reply to the requesters current region flood a message in its region to delete the back-tracking message for this reply else if the requester is its neighbor send the metadata to the neighbor else broadcast the metadata to its neighbors end if end ifend if

18

Colouring-based Partial ReplicationColouring based partial replication(CBPR) solves the problem with storage of metadata.

CBPR aims to prohibit two neighbouring nodes holds the metadata for a node.

Because of mobility, neighbouring relationship between nodes change.

To maintain the colouring status,each node in a region needs to check its neighbour.19

Pseudocode for the coloring-based partial replication algorithm for each node// ensuring that at least one neighbor has metafile;if it does not have metafile then for each hello message from its neighbor ni do if ni is a replica node then Record ni in its replica node list end if end for if none of its neighbors is a replica node then Randomly select a neighbor nj send a metadata request to nj with TTL end if end if

20

20

// Handling received metafile request;if received a metafile request from nj then if it is not a replica node then if it has no neighbor which is a replica node then forward the request to a randomly selected neighbor else forward the request to its neighbor that is replica node end if else send a metafile to nj end if end if

21

// Handling received metafileif received a metafile then if it is not the metafile requester then Send metadata to the previous request sender end ifend if

22

Replica Management in node mobility.

Nodes need to transfer replica when they move in or out of a region.

Node join :- when a node moves into a region,it checks whether any of the neighbour has a metafile.If not ,send a metafile request.

Node departure:- when a node moves out of its region,if it is a replica node it transfers its metafile to a node in the region.Otherwise no need to notify any node23

Pseudocode for metafile replication in node mobility // when node ni moves into region Riif ni has neighbors in region Ri then if no neighbor has a metafile then Request a metafile from a randomly selected neighbor end ifelse send a metafile request to region Ri using RGRend if// when node ni moves out of its region;if ni is a replica node then if ni has neighbors in region Ri then

24

send its metafile to a randomly selected neighbor delete its metafile and leave else send a metafile transfer request to region Riend if if ni receives a metafile transfer response from nj then if ni is not a replica node in Rj then send metafile to nj end if ni deletes its metafile and leave end if end if25

Performance EvaluationTo evaluate the performance of LORD,we compare it with Geographic Hash Table(GHT),Global Locating System(GLS),Ad-hoc On Demand Vector(AODV).

We evaluate the * Average success rate :- It is the ratio of number of received files to the number of initiated file queries.

* Average path length :- It is the number of hops for routing the query to the metadata holder.

* Overhead :- It is the total number of all traversed hops in metadata responding,mapping and updating.26

Performance Evaluation

27

Performance EvaluationHigher success rate,shorter path length and lower overhead of LORD indicates its high scalability.

-Due to fewer mapping updates and back tracking of LORD generate higher success rate.

- RGR in the LORD generates the shortest path.

-Colouring based partial replication provide low overhead to LORD.

Backtracking algorithm and redundant replica helps to achive high mobility resilience to LORD.

28

Conclusion

LORD is an efficient data search system for a large scale highly mobile wireless network.

It consists region based data publishing and querying.

It divides the network into number of regions.

It stores the metadata in multiple nodes in region for mobility resilient.29

Conclusion It uses RGR for forwarding data to destination.

It uses parallel file fetching algorithm to minimize the file fetching latency.

It uses coloring based partial replication that reduces replication in a region.

It also works for an unbalanced wireless network with sparse region.

30

References

[1] Haiying Shen, Ze Li and Kang Chen, IEEE transactions on parallel and distributed systems.

[2] S. Guo, Y. Gu, B. Jiang, and T. He, Opportunistic Flooding in Low-Duty-Cycle Wireless Sensor Networks With Unreliable Links,

31

THANK YOU32

ANY QUERIES???33