Cooperative Caching Techniques for Mobile

Ad hoc Networks
Preetha Theresa Joy, K. Poulose Jacob
Dept. of Computer Science, Cochin University of Science and Technology
Kochi, Kerala, India.

Abstract—Data caching can remarkably improve the important issue in ad hoc networks, other issues such as
efficiency of information access in a wireless ad hoc data access are also very important. High data
network by reducing the access latency and bandwidth availability is of increasing interest in mobile computing
usage. The cache placement problem minimizes total data environments. In ad hoc networks, due to frequent
access cost in ad hoc networks with multiple data items.
The ad hoc networks are multi hop networks without a
network partition, data availability is lower than that in
central base station and are resource constrained in terms traditional wired networks. The two approaches used in
of channel bandwidth and battery power. By data caching distributed computing to enhance data availability are
the communication cost can be reduced in terms of data replication and data caching. Data replication is the
bandwidth as well as battery energy. As the network node maintenance of multiple copies of data at multiple
has limited memory the problem of cache placement is a locations. The copies of data are not necessarily identical
vital issue. This paper attempts to study the existing at any particular point in time. Data caching improves
cooperative caching techniques and their suitability in the data retrieval process in mobile networks. In a cached
mobile ad hoc networks. environment since the requests are served from the local
Keywords — Data Caching, MANET, Cooperative caching,
cache, latency of fetching the data from the server is
cache consistency, cache placement. reduced.

I. INTRODUCTION The main purpose of this work is to explain the basic

concepts concerning cooperative caching in ad hoc
Improvements in wireless networks and small device networks and to expose some of the techniques
processing power have increased mobile computing developed for cooperative caching with its benefits and
usage recently. New wireless network and drawbacks.
telecommunication standards allow devices to
communicate using higher speeds with longer range and The remainder of the paper is structured as follows.
more reliability. The ad hoc networks have self Section II compares data replication and data caching.
organizing capability and can be effectively used where Section III discusses the cooperative caching in MANET
other technologies either fail or cannot be deployed Section IV presents cooperative caching design and
effectively. This situation makes extensive use of mobile architecture. Section V surveys some related work
computing more attractive. It also gives a number of carried out in the field. Section VI discusses about the
problems like message loss, unpredictable open research issues and section VII concludes the paper.
disconnections of mobile devices, and network
II. REPLICATION AND CACHING
partitioning. However, in order to offer high quality and
low cost services, several technical challenges still need The major factors influencing data availability in a
to be addressed in ad hoc networks. mobile computing environment are network partition and
disconnected operation. Mobile users deliberately
Most of the researches on MANETs have focused on the
disconnect their computers or may become
development of Media Access Control protocols and
unintentionally disconnected from a wireless network as
routing protocols to increase connectivity among mobile
they move around. In order to be able to work in these
hosts in a constantly varying topology. Because reliable
circumstances, multiple copies of heavily used data are
and efficient access to information is the ultimate goal of
made. With replication, data generated one time is
networking, the design of higher layer protocols to meet
copied or replicated to multiple nodes. This requires an
user requirements is necessary. Although routing is an

978-1-4673-2149-5/12/$31.00 ©2012 IEEE 175

earlier knowledge of the operational environment and is there is any node in the vicinity that has cached the data
susceptible to node mobility. This makes data replication it requires and where it is, if any. In cooperative caching
unsuitable for highly dynamic network topologies. A in addition to the local cache service, the network’s
common requirement when data is replicated is nodes cooperates with their neighboring nodes for
consistency which can vary in strength between serving data requests.
applications. Increasing the number of replicas will
increase the replication cost and update cost. So the IV. COOPERATIVE C ACHING DESIGN AND
replication protocol should make a tradeoff between data ARCHITECTURE
availability, replication cost and update cost.
In a cooperative cache based data access architecture
In caching-based systems the nodes cache data or the mobile nodes cache the data or the path to the data to
path to a data item, in order to increase data availability reduce query latency and to enhance data availability.
and to reduce query delay. Unlike data replication Here the data is stored in a distributed way and is shared
schemes, caching do not rely on copying data in advance. in the entire network. A good cooperative cache
Client node always caches the received data and management technique should have proper data
subsequent requests are served from the cache. If a cache discovery, cache admission control, cache replacement
miss occurs it has to get the data from the server. Here and cache consistency maintenance schemes. The cache
our focus is on caching protocols proposed for ad hoc admission control module decides whether a received
networks as they are more suitable to improve data data item is cacheable or not, and a cache discovery
availability in a frequently changing network module is to find and deliver requested data items from
environment. the neighboring nodes. Since the network node has
limited memory, it can cache multiple data items subject
III. COOPERATIVE CACHING IN MANET to its memory capacity. The objective of cache admission
control module is to store more distinct data items in the
Cooperative caching is the sharing and coordination of given cache space. The cache consistency module is to
cached data among several nodes to improve the data maintain cache consistency, ie, ensuring that each node
availability in wireless mobile networks. A set of caching the data item is aware of the data update at the
distributed caches can improve their performance by source. The cache replacement policy helps in deciding
cooperating in serving each other’s request and making the items to be removed in the existing cache to make
storage decisions. The criteria for selecting nodes that way for new ones. Replacement algorithms should keep
participate in cooperative caching may vary with track of the usage information to improve the cache hit
different caching techniques. It may be the nodes that lie ratio.
within a geographical area or in a zone or in a cluster.
Cooperative caching takes advantage of local caching to A. Classification of Cooperative caching Techniques
facilitate data access by allowing geographically
neighbouring hosts to share their cached contents. By All cooperative caching techniques implicitly assumes
doing so long-distance data accesses from the data center that a node will cooperate each other in delivering the
can be reduced. Here each node has to know whether

Fig . 1. Taxonomy of different cooperative cache management schemes.

2012 International Conference on Data Science & Engineering (ICDSE) 176

desired data. Based on the system architecture ie, how served from the cache until the cache expires. Here we
the mobile nodes communicate each other for compare some representative cooperative caching
cooperative caching ,the proposed caching techniques strategies for mobile ad hoc networks which combine the
can be divided in to decentralized group based different aspects of the caching policies mentioned
architecture and centralized group based architecture. In above.
decentralized group based architecture, mobile nodes are
arranged in to different groups based on various In [1], three caching schemes, Cache Data, Cache Path
parameters. Each node will belong to only group. There and Hybrid Cache, were proposed to provide efficient
is no control node in this group and the clients can access data access in MANETs. In Cache Data scheme the
data from any other node in the group. In the later case, a intermediate nodes in the routing path between the
mobile node belongs to different groups based on various source and the destination caches the data item when it
criteria. Here each group is associated with group control finds that the data item is popular, i.e., there were many
node to handle query and update operations. requests for the data item, or it has enough free cache
space. There is no cooperative caching among the mobile
Depending on functionality, the different modules in hosts. Each mobile host independently performs the
cooperative cache management can be further divided caching tasks such as placement and replacement. Cache
into different groups. The data discovery process can be Path works in a similar manner to Cache Data, but the
either push based or pull based. In push based approach, forwarding nodes cache the data path information for
the data requesting node can directly find the node which future use, instead of just the data. Hybrid Cache takes
has the cached copy of data. In pull based approach when advantage of the above two schemes for further
a local cache miss occurs, the request is broadcasted to improving the caching performance. Here a mobile node
its neighboring nodes and a node which has the data will caches the data or the path based on some criteria. These
respond. Flooding is the technique used for broadcasting. criteria includes the data item size and the Time to Live
TTL. If data item size is small, Cache Data is used since
In cache admission control, data is admitted to cache in the data item needs only a small portion of the available
two different ways. In the general scheme all the cache; otherwise, Cache Path is used. If TTL is small,
incoming data are cached. In function based cache Cache Data is used because the data item might soon be
admission control, a cost function will decide whether to invalid, using Cache Path can result in changing the path
cache the incoming data. The value function is formed frequently. Here the cache consistency is ensured by a
by combining different parameters like distance and time to live (TTL) mechanism. A routing node considers
access frequency. The cache replacement policies were a cached copy valid if its TTL hasn’t expired. If the TTL
classified in to two groups, local and coordinated. In has expired, the node removes the cached data item.
local replacement the data item to be evicted is
determined independently by each node based on its A caching scheme for Internet based MANETs was
local access information. In coordinated replacement presented in [2]. In this scheme, an Aggregate Caching
policy the mobile nodes which forms cooperative cache mechanism is proposed. Here each mobile node stores
collectively takes the replacement decision. There are data item in its local cache and these local caches are
two kinds of consistencies: strong and weak. A cache aggregated to unified large cache. The data items can be
access mechanism can either use strong or weak received from local caches of the mobile nodes or
consistency depending on applications. Cooperative through an access point or a data center connected to the
caching techniques mainly uses weak consistency model Internet. When a node needs a data item, it broadcasts a
based on time to live parameter. Fig.1 shows the request to all of the adjacent nodes. The node which has
classification of different cooperative caching the data item in its local cache will send a reply to the
techniques. requester; otherwise, it will forward the request to its
neighbors. Thus a request is broadcasted to the other
V. DATA C ACHING APPROACHES FOR MANETS connected nodes and eventually acknowledged by an
access point or some nodes with cached copies of the
To increase the data accessibility in ad hoc networks, requested data item. A broadcast based approach, called
several data caching schemes have been proposed in the Simple Search (SS) algorithm, is implemented on the
literature.These schemes allocate every mobile node a top of existing routing protocols, to locate the requested
certain amount of cache space. In a simple data caching data items.
scheme, a node that requests data always caches the
received data and consecutive requests for the data are

2012 International Conference on Data Science & Engineering (ICDSE) 177

 In [3], COOP addresses two basic problems in co follows. When a node takes data from a remote node, it
operative caching ie, cache resolution and cache puts the data in its local cache for reuse. A least recently
management. In cache resolution COOP tries to discover used replacement algorithm is used to remove the data
a data source which includes less communication cost by from the cache. The data evicted from the cache is stored
utilizing historical profiles and forwarding nodes. For in the idle neighbor node’s storage. If the node needs the
this a mobile node uses three schemes; adaptive flooding, data again, it will retrieve it from the near neighbor that
profile-based resolution and road side resolution. In keeps the copy of the data. This scheme exploits the
adaptive flooding, a node uses constrained flooding to available cache space of neighbor to improve the caching
search for items within the neighborhood. In profile- performance. In order to find a neighbor node with the
based resolution, a node uses the past history of received highest possibility to keep the data for a long time a
requests. In road side resolution, forwarding nodes ranking based prediction algorithm is used.
caching the requested item, reply to the requests instead
of forwarding them to the remote data source. In cache In [6] a scheme called Group Caching (GC) in which
management, in order to minimize the caching each node maintains a localized and group caching status
duplication between neighboring nodes it uses iter- of its 1-hop neighbors is given. Each mobile host and its
category and intra-category rules. Cache consistency is 1-hop neighbors form a group by using the “Hello”
maintained by Time To Live (TTL) scheme. message mechanism. In order to utilize the cache space
of each node in a group, the nodes periodically send their
In Zone Cooperative caching [4] each mobile client caching status in a group. By checking the cache status
belongs to a zone and has a cache to store the frequently of the group, each mobile host knows the remaining
accessed data items. In this scheme each node has a local available cache space of other mobile host in a group, the
cache and these nodes share cache with other nodes that IDs and timestamps of their cached data object. Thus, the
are in a zone. Zone consists of mobile hosts that are mobile host selects the proper group member for cache
accessible in one-hop. So mobile hosts belonging to the placement and replacement. The cache replacement is
zone of a given host forms a cooperative cache system. based on Least Recently Used (LRU) policy and a cache
The data items in the cache satisfy not only the client’s consistency based on Time to Live (TTL) attribute is
own requests but also the data requests passing through it used in this scheme.
from other clients. When there is a miss in the local
cache, the client first searches the data in its zone. If the In [7], a Cluster-based Cooperative Caching (COCA)
data is found it is retrieved otherwise the request is scheme that uses a cross-layer approach to improve
forwarded to the next client that lies on a path towards caching performance is proposed. COCA is a
server. A cache replacement policy based on least utility middleware which implements cross-layer optimization
value (LUV) has been used. For computing the LUV of a for cooperative caching. It consists of a stack profile,
data item the access probability (Ai), size of the data clustering, information search, cache management and
item (Si), coherency which can be known by TTLi field prefetching modules. The stack profile module provides
and distance ( ) between the mobile client and data cross-layer information exchange, the clustering module
source were considered. Eq. for utilityi function for a for cluster formation and maintenance, the information
data item (di) search module for locating and fetching the data item
requested by the client. The basic task of prefetching
utili tyi Ai . i .TT Li module is to anticipate user’s request and to fetch and
cache highly likely data items in advance. The cache
si management module includes a cache admission control,
cache replacement and cache consistency. For cache
A cache admission control based on distance and a cache admission control, a node caches all received data items
consistency based on time to live parameter is also until its cache space is full. When the cache is full the
considered in this scheme. data items are not cached if the cluster has a copy of it.
For cache consistency this scheme uses Time To Live
The concept of Neighbor Caching (NC) [5] is to deploy (TTL) strategy and uses Least Recently Used Minimum
the cache space of inactive neighbors for caching tasks. (LRU-MIN), for cache replacement.
In this method a node uses its neighbor temporarily as

2012 International Conference on Data Science & Engineering (ICDSE) 178

 T ABLE.1. FEATURES OF EXISTING D ATA C ACHING SCHEMES IN M OBILE AD HOC NETWORKS

Cache Cache
Caching Centralized/Dec Cache Performance
admission Data Discovery Consist Advantages Disadvantage
Scheme entralized Replacement Metric Used
Control ency

Cache path
Cache Data Based on
become invalid
/Path Based on no: By message distance and TTL Increased data
Decentralized Average delay when topology
Hybrid of hops exchange access based accessibility
changes, space
cache[1] frequency
needed is more.

Multiple copies of
Aggregate Based on
Based on Broadcast Distinct data is data returned to
Cache [2] distance and Not Throughput, Avg
Decentralized based simple stored increases server extra
access Present no of hops
no :of hops search through put communication
frequency
cost.

COOP [3] Based on Request Success Overhead in

Adaptive TTL Limited
Decentralized LRU ratio, Avg maintaining cache
flooding based flooding
no of hops response delay resolution.

Based on Simple to Latency is more if

Zone Broadcast Avg no of hops,
distance TTL implement, a neighbor does
Cooperative Decentralized request to Value Based Avg query
based Value based not have a cached
[4] neighbors latency
no of hops replacement copy.

Simple and
Data thrown out
Costless,
Neighbor nodes without reusing
Neighbor Not Avg Ending utilizes the
Decentralized Not Present address stored LRU lacks efficient
Caching [5] Present time, traffic cache space of
in local node cooperation
inactive
among nodes.
neighbors

Power efficient
Based on Search in local Extra cost to
Group TTL Cache hit ratio, protocol,
Centralized group member table and group LRU maintain local and
Caching [6] Based Avg latency redundancy is
status table group tables.
reduced

Difficult to
COCA [7] Data accessibility Cross layer
TTL maintain when the
Centralized Cluster based Hierarchical LRU Min ratio, Avg query design with pre
Based number of nodes
delay fetching
increases.

Mean Query Reactive

Duplication of
Adaptive Data is Shared TTL response time, caching,
Decentralized Based on TTL data is high if the
Cache [8] replicated Common index Based Hop Count, resource
replicas are closer.
Energy saving efficient

In [8], a Resource Efficient Adaptive Caching which uses towards a Dominant Request Path ie, the network path
an adaptive cache distribution and replication is responsible for majority of request for a particular data
implemented. The adaptive cache replication scheme item at the given node. When the demand wanes, the
known as Tidal Replication, considers a global demand replicas are de-allocated and the burden for the data item
for data for making replication decision. Here replication returns to the control node. When replication request is
is done by caching the data items that consume received the available storage space for the node (RSA)
significant portion of bandwidth for a control node is given by eqn,

2012 International Conference on Data Science & Engineering (ICDSE) 179

 RSA = s m – c it’s time to- live has not expired. But for some
s–m applications like application that depend upon location
Where s represents the total storage capacity of the node, dependent data emphasis should be given to ensure
m represents the total storage utilization of its non replica strong consistency.
data items and c represents the total storage utilization of
any cached replica it is holding. For resource efficiency VII. CONCLUSION
an adaptive de allocation methodology based current
utilization and age has been included. The current In this paper we reviewed the challenges and basic
utilization is given by concepts behind cooperative caching in MANETs and
U = xi provided a summary of the work done in the area of data
zi caching. Existing techniques were compared based on
Where xi is the windowed cumulative number of how they addressed the identiÞed issues. It is observed
requests, and zi is the age of the replica, for the given that the different schemes referred above produce
data item i. The age based de- allocation threshold Tag e varying results with respect to the parameters like
is represented in Eq. where yi is equal to the age latency, hop count and cache hit ratio depending on the
threshold time in minutes, and zi is the age of the cache size, node density etc. None of the existing works
replication. give a combined and complete solution for efficient
Tag e = yi cache placement, discovery, replacement and
zi consistency.
REFERENCES
The cache distribution scheme, Magnetic Distribution
redistributes data items according to their demand, [1]. L. Yin, G. Cao: Supporting cooperative caching in ad hoc
networks, IEEE Transactions on Mobile Computing, 5(1):77-89( 2006).
balancing the demand in a place of equilibrium, in order
to reduce hop count and response time. The different [2]. S. Lim, W. C. Lee, G. Cao, C. R. Das:A novel caching scheme for
cooperative caching schemes are summarized in Table 1. internet based mobile ad hoc networks. Proc .12th Int. Conf. Computer
Comm. Networks (ICCCN 2003), 38-43 ( Oct. 2003).

[3]. Y. Du,Gupta, S. K. S.:COOP-A cooperative caching service in

VI. OPEN RESEARCH ISSUES MANETs," in Proc. Joint Int. Conf Autonomic and Autonomous
Systems and Int. Conf. Networking and Services,. 58-63( 2005).
Though different techniques for cooperative caching in
MANET were proposed the area is still largely open to [4]. Chand, N. Joshi R.C., and Misra, M., Efficient Cooperative
research. In order to find the location of a cached copy of Caching in Ad Hoc Networks Communication System Software and
Middleware.(2006).
a data item in cooperative caching mechanism the
protocols relies on flooding. But flooding can increase [5]. Joonho Cho, Seungtaek Oh, Jaemyoung Kim, Hyeong Ho Lee,
network contention and collision and this causes many Joonwon Lee, Neighbor caching in multi-hop wireless ad hoc
networks, IEEE Communications Letters, Volume 7, Issue 11, 525 –
problems in network communication. In group based
527(2003).
approach used in cooperative caching, the group
maintenance is hard and costly. When the number of [6]. Yi-Wei Ting ,Yeim-Kuan Chang. A novel cooperative caching
nodes increases, it is difficult to handle the local and scheme for wireless ad hoc networks: Groupcaching. In NAS ’07: Pro-
ceedings of the International Conference on Networking,
group tables. So the algorithms must be improved to deal Architecture,and Storage,( 2007).
with these issues. Data dissemination in a mobile
computing environment is characterized by three [7]. Denko, M.K., Tian, J.,Cross-Layer Design for Cooperative
resource constraints including energy, communication Caching in Mobile Ad Hoc Networks, Proc .of IEEE Consumer
Communications and Networking Conf( 2008).
bandwidth and storage. The data caching techniques
proposed deals with one or two aspects of these [8]. Dan Hirsch, Sanjay Madria A Resource- EfÞcient Adaptive
constraints but not all the three. Although data caching in Caching Scheme for Mobile Ad-Hoc Networks.29th IEEE International
Symposium on Reliable Distributed Systems.( 2010).
mobile ad hoc networks has been well studied, the cache
management issues like consistency and cache [9].V. Srivastava and M. Motani, "Cross-layer design: a survey and the
replacement are not adequately addressed. In the data road ahead," IEEE Communications Magazine, 43(12): 112-119, 2005.
caching techniques presented in this paper, to maintain
cache consistency all of them uses a weak consistency [10].T. Hara. Replica allocation in ad hoc networks with periodic data
update. Intl. Conf. on Mobile Data Management (MDM), 2002.
model which uses TTL (Time-To-Live) based approach,
in which a client considers a cached copy up-to-date if

2012 International Conference on Data Science & Engineering (ICDSE) 180