Download Query Directories

Database caching in MANETs
Based on Separation of Queries
and Responses
Author: Hassan Artail, Haidar Safa, and Samuel Pierre
Publisher: Wireless And Mobile Computing, Networking
And Communications, 2005. (WiMob'2005), IEEE
International Conference on
Presenter: Chen-Yu Chung
Date: 2008/09/24
Outline

Introduction

Proposed architecture





Goal
Components of the Architecture
The flow chart of data Discovery
Formation of Query Directories
Simulations and Result
Introduction(1/2)

Most previous researches on MANETs have been
focusing on the development of dynamic routing
protocols that make efficient use of bandwidth and
computational overhead.

Although routing is an important issue, the ultimate
goal of a MANET is to provide mobile nodes with
access to services.
Introduction(2/2)

In MANETs accessing services and data over the
network can be very slow and hence, caching
frequently accessed data is an effective technique
for improving performance.

Therefore, devising an effective caching
strategy for the whole MANET is of special
importance since it allows for improving the
performance of the network as a whole.
Outline

Introduction

Proposed architecture





Goal
Components of the Architecture
The flow chart of data Discovery
Formation of Query Directories
Simulations and Result
Goal

Our aim is to increase the ability of mobile devices
to have access to database data given their highly
dynamic characteristics, such as mobility, power
limitations, and intermittent availability. Our proposal
calls for building a hierarchical architecture for
handling the distribution of data, for providing
services to the requesting nodes, and for managing
the roles of the nodes.
Components of the Architecture

SM (Service Manager)



CNs (Caching Nodes)


To assignment and to Keep track of which nodes have the role of QD and
BQD
To broadcast the available QD information across the mobile network
To store the query response
QDs (Query Directories)


To store the queries without the data
To associate these queries with the CNs that hold the corresponding data

BSM (backup SM)

BQDs (backup QDs)
The flow chart of data Discovery(1/5)

We suppose that there is only one QD in the
mobile network
Broadcasting the
SM
DB
QD
CN
information about QD to all
the mobile node in the
network
MN
The flow chart of data Discovery(2/5)
SM
DB
QD
CN
query
MN
The flow chart of data Discovery(3/5)
SM
QD
DB
Hit
CN
query
MN
The flow chart of data Discovery(4/5)
SM
Miss
DB
QD
CN
query
MN
The flow chart of data Discovery(5/5)

To note that,if ther has several QDs in the
mobile network,the requesting Mobile Node
(MN) will generate a random integer number
between 1 and the number of QDs and then
send its query to the corresponding QD in the
list of QDs.
Formation of Query Directories(1/4)

Initially, the system starts with a single QD that caches
queries.
Caching
new query
Single QD in
the network
Formation of Query Directories(2/4)

When this QD is filled beyond a given threshold, it
informs the SM, which will assign an additional QD.
Caching
new query
Non full QD
Full QD
Formation of Query Directories(3/4)

When this new QD becomes filled, another node will join
the list of QDs that are linked in a circular manner. The
number of QDs will increase until some kind of steady
state level is reached
Caching
new query
Non full QD
Full QD
Full QD
Formation of Query Directories(4/4)

At any point in time, there is one QD that is available to
cache non-cached queries and we refer to it as the nonfull QD (NF-QD). If the submitted query was not found in
the QDs, it is the NF-QD that goes to the database to
fetch the results, adds the query to its local cache,
updates its hash table, and forwards the result to the
requesting node.
Outline

Introduction

Proposed architecture





Goal
Components of the Architecture
The flow chart of data Discovery
Formation of Query Directories
Simulations and Result
Simulations and Result(1/8)
Simulation parameters
800m*800m
MN
Nodes: 25
AODV routing protocol
Random Waypoint mobility model
Query interval: 2 seconds
QD
Number of QD nodes: 3
DB
Implemented as a fixed node
Simulations and Result(2/8)

For the simulations, we made some simplifications
to get preliminary results that confirm the viability of
the approach:
1.
2.
3.
We manually selected three QDs to be near the center of
gravity of the network and made the assumption that all
mobile nodes (MNs) know about it.
The nodes that requested queries were selected randomly
and uniformly. This directly implies that the CNs were
uniformly distributed throughout the network.
No replacement algorithm was implemented to update the
contents of the CNs as their storage capacity was assumed
unlimited.
Simulations and Result(3/8)
4.
The database server was implemented as a fixed node far
from the mobile nodes (node 25 in Fig)
Network Topology using NS-2
Simulations and Result(4/8)
5.
6.
7.
8.
At this stage, the SM and the backup nodes (BSM and
BQDs) were not implemented.
The network comprises equally capable nodes.
It was assumed that the DB server is always accessible
by the network.
We simulated 40 total queries that represent all the
possible queries that the MNs could request
Simulations and Result(5/8)

Time consumed to get query responses
Simulations and Result(6/8)

Hop Count vs. Number of requests
Simulations and Result(7/8)

Hop Count vs. cache size per QD
Simulations and Result(8/8)

Hop Count with Caching/No Caching