Paper: COOP- A cooperative caching service in MANETs

1
Paper COOP- A cooperative caching service in
MANETs

• Author Y. Du and S. K. S. Gupta
• Proceedings ICAS-ICNS 2005. Joint
  International Conference on, Tahiti, French
  Polynesia, pp 58-63, Oct. 23-28, 2005
  
  
  
  Presented By Aarti Munjal
• 
  PhD(CSE)
• 
  Arizona State University
• 
  CSE 535 Mobile Computing
• 
  Paper presentation

2
Contents

• Background
• Motivation
• Related Work Contributions
• COOP - Overview
• Cache Resolution
• Cache Management
• Performance Evaluation
• Related to class
• Use in Project
• Future Work

3
Background

• Mobile ad hoc networks(MANETs) are constrained in
  terms of resources and lack of infrastructure.
• Routing techniques need to take care of these
  facts.
• There are disconnections(among nodes) and
  failures(at node level due to battery power etc.)
  which lead to loss of data.
• Real-time data availability becomes a challenge
  in such a scenario.

4
Motivation

• Performance of routing protocols can be improved
  either by providing MAC layer, network or
  transport layer solutions.
• One more possibility exists that is explored by
  this paper solution at application layer.
• Nodes can cooperate to localize the communication
  which leads to conservation of energy, time as
  well as bandwidth.
• Solution has to be
• efficient - due to constrained resources.
• self-adaptive - due to dynamic nature of the
  network.
• Scalable increase in number of nodes does not
  affect the performance.

5
Related Work Contributions

• Caching has been used as solution to reduce the
  data access time
• Hierarchical caching 1
• Directory based caching 2
• Hash-table based 3
• Cache data, cache path and hybrid cache 4

Node 1 Data i
Node 2 . . . . Data j . . . .



1
Node 1
2
Node 3

6
COOP - Overview

• Aim
• To improve data availability and access
  efficiency.
• Cooperative Caching
• Cache Resolution where to fetch the data
  requested by the user.
• Cache Management due to memory constraints
  involved, which data to purge to make room for
  the other information.

Figure 1 System Model
7
Cache Resolution

• Adaptive Flooding
• Flooding
• to know neighbors and introduce yourself.
• affects the performance of a protocol.
• Calculates proper flooding range.
• Cost of fetching data x
  Ls
• 
  (distance of cache containing data)
  (distance to original data source)
• Average Cost
• Pd probability that each
  node caches data.
• ? average node density.
• ?px2 number of nodes in
  x-hop range.
• 1- (1 - Pd) ?px2 probability to discover
  data in cache within x-hop range
• X Ls(1 - Pd) ?px2 average cost of fetching.

8
Adaptive Flooding contd
Few important points i) To achieve minimum
average data fetching cost, the optimal flooding
range increases very slowly. ii) If ?
increases, x shall be reduced accordingly. iii)
Limited flooding minimizes the average data
fetching cost.
Figure 2 Average data fetching cost vs flooding
range
9
Profile-based Resolution

• Avoids duplicate flooding by storing history of
  previous requests in Recent Requests Table (RRT).
• Each entry of RRT contains
• On a data request, every node checks its local
  cache first.
• Upon cache miss, RRT is searched for the
  corresponding entry.
• Data source selected mindistance
  matching entry , original data source , if a
  match is found in RRT
• Adaptive flooding used and
  corresponding entry removed
  , otherwise

Sender Target Timestamp
10
Cocktail Resolution Scheme

• Roadside Resolution
• Data request needs to be forwarded to original
  data source.
• Request starts from sender to target.
• Any forwarding node in the path checks its own
  cache first, if cache hit then stops forwarding
  the request and sends the data back.
• In case no data is found, if it finds another
  data source nearby, it redirects the data request
  to that node.
• Otherwise, the request is juts forwarded towards
  the target node.

Figure 3 The Cocktail Resolution Scheme
11
Cache Management

• Maximizes distinct data availability by reducing
  duplicated cache in short-distance neighborhood.
• Data categorized
• Primary data - data unavailable in
  neighborhood (neighborhood range is
  customizable).
• Secondary data - data available in
  neighborhood.
• Rules deciding the category of a data item
• Inter-category rule
• Fetched a data item, label of data
  primary, if comes from outside neighborhood range
  or from
• 
  within
  neighborhood but has been labeled secondary
• 
  there
  and the primary copy holder is out of range.
• 
  secondary,
  otherwise.
• Intra-category rule
• For data items within the same category
  (primary/secondary).
• LRU is used for the same purpose.

12
Performance Evaluation

• Access probability of i-th popular data item.
• Request Success ratio shows data availability
• Average Response Delay time efficiency

Figure 4 Average response delay
Figure 5 Request success ratio
13
Relevance to Mobile Computing

• Disconnections and failures in mobile
  environments are quite frequent(due to several
  reasons).
• In order to make sure that data availability is
  not affected due to that, there are several
  solutions that can be looked at.
• Caching the data is one among those solutions.
• Caching can be very extensive depending upon the
  applications.
• For instance, context-awareness can be added
  which leads us to cache the data and change the
  caching strategy depending upon the context.
• These are few topics we have looked at in Mobile
  Computing course CSE 535.

14
Relevance to our project A Context Aware
Caching Scheme for Real-Time Health Monitoring
Systems
Server

• Multi-tier architecture adopted to improve
  scalability.
• To provide real-time data even during
  disconnections and failure, caching is required.
• Cache Resolution
• Cache Management

H8
H7
H3
H2
H1
H4
H5
H6
P5
P6
P7
P9
P10
P11
P12
P8
P4
P2
P1
P3
Figure 6 Multi-tier architecture
15
Conclusions Future Work

• Cooperative Caching higher data availability.
  
• Time cost reduced by using cock-tail approach for
  cache resolution.
• Inter-category and intra-category rules used to
  minimize the data redundancy.
• Caching scheme can be associated with
  context-awareness to make it adaptive so that it
  suits the very nature of MANETs.
• To make it suitable for energy-efficient routing,
  few modifications can be made, for instance
  incorporating remaining energy of a node.
• Different radius values(cooperation zones) can be
  employed to see the effect of cooperation among
  nodes.

16
References

• A. Chankhunthod et al. A hierarchical internet
  object cache. In USENIX Annual Technical
  Conference, 1996.
• L. Fan et al. Summary cache a scalable wide-area
  web cache sharing protocol. In Sigcomm, 1998.
• S. Lyer et al. Squirrel A decentralized
  peer-to-peer web cache. In PODC, 2002.
• G. Cao et al. Cooperative cache based data access
  in ad hoc networks. IEEE Computer, 37(2)3239,
  Feburary 2004.