AN EFFICIENT COOPERATIVE CACHE APPROACH

IN MOBILE INFORMATION SYSTEM

Thu Tran Minh Nguyen and Thuy Thi Bich Dong

IS Department, Faculty of IT, University of Science, VNU, HCMC, Vietnam

Keywords: Mobile information systems, Cooperative computing, Cooperative caching, Cache discovery, Cache

replacement, Mobile P2P network.

Abstract: Data availability in mobile information systems is lower than in traditional information systems due to its

limitations of wireless communication such as low bandwidth, instability, and easy disconnection.

Cooperative data caching is an attractive approach related to this problem because it allows sharing and

coordinating the cached data among multiple mobile users in the network. So it can improve the

performance of the system and the accessibility of data items. However, if we do not have a good

cooperative caching approach then the cache hit ratio and response time may become ineffective. In this

paper, we propose an efficient approach for cooperative caching in mobile information systems namely

MIX-GROUP. We evaluate and demonstrate the experimental results on datasets by using NS2 simulator.

MIX-GROUP is also proved the effectiveness by comparing it with the existing other approaches such as

COCA and GROUPCACHING. The experimental results present that MIX-GROUP is more effective than

those other approaches.

1 INTRODUCTION

Wireless communications and mobile devices have

become more and more popular in our life. People

use mobile devices such as mobile phones, pocket

PCs, laptops … to retrieve information from other

applications anywhere and at anytime they need.

This type of applications is based on mobile

information systems or mobile databases. There are

many mobile clients connected to a server through

wireless links in a mobile information system.

However, mobile information systems raise new

problems that do not exist in traditional information

systems. Wireless links are unstable, which means

the bandwidth of the uplink channel (mobile client

to base station) is usually narrower than that of the

downlink (base station to mobile clients). Mobile

devices often have small storage capacity and weak

power. Furthermore, a mobile user (client) is usually

disconnected from the base station (server) because

of unstable wireless links or out-of-battery itself.

Therefore, researchers have explored several

solutions for effective data management in mobile

platform. Cooperative caching in mobile peer-to-

peer systems is more interesting recently. In mobile

peer-to-peer system, the clients will look up the

required data at their own local cache or other peer’s

local cache before sending the data requests to the

server. This means that the cooperative caching

explores how to collaborate multiple caches to

achieve a better performance. By cooperatively

caching the frequently accessed information, mobile

clients do not always have to send requests to the

base station. This technique can get lower mobile

host communication overhead and energy

consumption as well as reduce query latency.

However, mobile peer- to-peer systems have brought

up new challenges for researching on routing,

resource discovery, data retrieval, and data

consistency control, security and privacy

management.

In this paper, we will propose a cooperative

cache approach in mobile information system. Our

proposed approach addresses three problems

including cache discovery, cache admission and

cache replacement. We also evaluate the

effectiveness of our approach by some simulation

experiments. Especially, experiment results are

compared with COCA (Chow, 2004) and

GroupCaching (Ting, 2007) to prove the

effectiveness of proposed approach.

153

Tran Minh Nguyen T. and Thi Bich Dong T..

AN EFFICIENT COOPERATIVE CACHE APPROACH IN MOBILE INFORMATION SYSTEM.

DOI: 10.5220/0003504901530159

In Proceedings of the 13th International Conference on Enterprise Information Systems (ICEIS-2011), pages 153-159

ISBN: 978-989-8425-56-0

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

2 RELATED WORKS

Cooperative caching improves the system

performance because it allows sharing and

coordinating the cached data among multiple mobile

users in the network. A variety of popular

cooperative caching strategies have already been

studied such as: Zone Cooperative scheme (Chand,

2007); proactive approach for cooperative Caching

(Kumar, 2010); Cache Data, Cache Path, Hybrid

Cache (Yin, 2004); COCA (Chow, 2004);

GroCOCA (Chow, 2004); Cluster Cooperative

Caching (Chand, 2006); COOP (Yu Du, 2005 &

2009). However, none of all these researches

provides a complete solution for cooperative

caching. Each of them has both advantages and

disadvantages. In this section, we only analyse

approaches which are used and compared with our

approach in more detail.

COCA (Chow, 2004) is proposed by Chow et al.

In this cooperative caching protocol, the mobile

node shares its cache contents with each other to

reduce the number of server requests and access

miss ratio. However, access latency is quite high

when the mobile hosts encounter a global cache hit.

Yu Du & et al proposed a cooperative caching

scheme called COOP for MANETs (Yu Du, 2005 &

2009). To improve data availability and access

performance, COOP addresses two basic problems

of cooperative caching. For cache resolution, COOP

uses the cocktail approach that consists of two basic

schemes: hop-by-hop resolution and zone-based

resolution. By using this approach, COOP discovers

data sources, which have less communication cost.

For cache management, COOP uses the inter- and

intra-category rules to minimize caching

duplications between the nodes within a same

cooperation zone. This improves the overall capacity

of cooperated caches. Disadvantage of this scheme

is the flooding, which introduces extra discovery

overhead.

Group Caching Scheme (Ting, 2007) maintains

localized caching status of one-hop neighbors for

performing the tasks of data discovery, cache

placement, and cache replacement when a data

request is received in a mobile host (MH). Each MH

and its one-hop neighbors from a group by using

“Hello” message mechanism. In order to utilize the

cache space of each MH in a group, the MHs

periodically send their caching status to its group.

Thus, when caching placement and replacement

need to be performed, the MH selects appropriate

group member to execute the caching task. In this

scheme, the biggest concern is the energy

consumption in MHs and constrain of wireless

bandwidth. Therefore, in this GroupCaching

scheme, each MH only maintains one – hop

neighbors in a group (Radhamani, 2010).

The cooperative caching scheme in our study

named MIX-GROUP. MIX-GROUP scheme is also

built basing on the favourable characteristics of the

other ones, such as it inherited the idea of marking

data item’s label when data items are cached into

MU’s local cache. This idea is proposed in COOP

scheme (Yu Du, 2005 & 2009). However, MIX-

GROUP resolved the problem of flooding that

COOP scheme as well as other ones has met. We

will present the proposition of MIX-GROUP scheme

in detail in Section 3.

3 COOPERATIVE CACHING

MODEL

3.1 Description of Model

We assume that the cooperative caching architecture

has only one server and many mobile users (MU) as

shown in Figure 1. The server is also called the base

station (BS). Each BS controls MUs in its service

zone. The communication between the BS and MUs

is wireless link. MU is the mobile devices used to

send requests to server and can move freely in one

cell or from one cell to another cell. We are

especially interested in cooperating data among

MUs in network in this model. Each MU has a

service zone to communicate with other one. The

connection among MUs is P2P wireless network.

Figure 1: Proposed Cooperative Caching Architecture-

MIXGROUP.

In this architecture, the BS acts as the server

provide data items for MUs and it can be updated

data items. MU plays the role of both the client and

the server. MU acts as the client when it sends the

requests to retrieve data items from the BS or the

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

154

other MU and as the server when it provides data

items for another MU in network. However, MU

does not allow updating data items in system.

3.2 Principle of Data Search Process

When the MU resource (example as MU 2 in Figure

2) emits the data request, MU resource will look for

requested data items in its local cache. Local cache

is the place in which caches data items requested by

previous queries. If the required data items are not

found in the local cache (or cache miss), MU

resource will send the data request to the MUs

neighbour to seek data items. MUs neighbour are the

MUs that belong to MU resource’s region. MU

resource links to MUs neighbour by one hop

(example as MU 8 in Figure 2). Then if they can’t be

found out in the MU resource‘s region, the data

request will be sent to the MUs network to find

requested data items. MUs network are the MUs that

do not belong to MU resource’s region. MU

resource links to MUs network by at least two hops

(example as MU 5 in Figure 2). Lastly, if requested

data items are not found at MUs neighbour or MUs

network, the data request will be sent to the BS for

requesting data.

Figure 2: Illustration for cooperative caching among MUs.

3.3 Processes at each MU Peer

In this section, we will present the architecture of

each MU in more detail. Each MU has three main

layers: User Application, Middleware and Network

Protocol as in Figure 3.

• User Application layer is responsible for

receiving the data requests emitted by users.

Then it transmits these requests to

Middleware layer in order to be processed.

User Application layer also receives the

requested data responded from Middleware

layer and then displays them to users.

• Network Protocol layer is used to

communicate and transmit messages among

the system’s elements.

• Middleware layer is the main layer that

serves for researching requested data items.

Middleware layer has three modules: Local

Query Process, In-zone Query Process and

Out-zone Query Process. Local Query

Process module is responsible for seeking

data items from local cache. It also resolves

for data admission and data replacement. In-

zone Query Process module takes the part of

searching data in the region that includes the

neighbours of the MU resource. Out-zone

Query Process module assigns a task to find

requested data at the MUs network of the MU

resource. The detail of processes and

metadata will be presented in the next

section.

Figure 3: Detail of the middleware layer.

3.4 Detail of Processes and Metadata

3.4.1 Local Query Process

a) Local Cache Discovery

Each MU has a local cache (LC). LC stores the

frequently accessed data items to serve local data

search process. Caching data items in the LC of each

MU helps reduce latency and increases accessibility.

The LC structure includes {id, f, t, L, TTL, D, S}

where id is the identification of the cached data item,

f is the frequency of cached data item accessed by

other MUs; L is the label of the cached data item.

Basing on L, we will recognize a data item as the

primary data or the secondary data; t is the

timestamp of cached data item. TTL value is the

living time of a data item, D is data value of data

item d, S is a size of data item d. When a MU

requests for data item d, it first will look for d from

its local cache. A local cache hit occurs if the request

is satisfied in local cache.

b) Cache Admission

When a MU receives the response for the

requested data item, a cache admission control is

triggered at MU to decide whether the data item

should be cached. The idea of cache admission

control is known as each cached data item will be

AN EFFICIENT COOPERATIVE CACHE APPROACH IN MOBILE INFORMATION SYSTEM

155

marked as the primary data (PD) or the secondary

data (SD). This process must always ensure that

there is only one primary data copy in cooperative

zone. That means once a MU fetches a data item, it

labels the data item as the primary copy if the data

item comes from a MU beyond the zone radius.

Otherwise, if a data items comes from within the

zone radius, we need to consider whether the data

provider labels the data item as primary or

secondary. If one of the providers already labels its

copy as primary, the new copy would be secondary

since we do not intend to have duplicated primary

copies in cooperative zone. On the other hand, if all

providers tag its own copy as secondary, the new

copy is primary one. The idea of classifying cached

data items is the same way that Y. Du et al used in

COOP scheme (Yu Du, 2005 & 2007). However, the

difference between COOP and MIX-GROUP

scheme is MIX-GROUP’s zone has one hop whereas

COOP’s zone has multi hops.

Figure 4: The Illustration for classifying cached data.

An example is illustrated in Figure 4. When MU 3

requests data item A, MU 4 replies and marks A’s

label as primary. Because MU 4 is in MU 3’s zone,

so data item A is copied at MU 3 as the secondary

data. Then MU 1 requests data item A, MU 3

responds A to MU 1 as the secondary data. MU 1

will caches A as the primary data because MU 1 has

the same zone with MU 3, but it has other zone with

MU 4.

c) Cache Replacement

Cache replacement means that MU decides

which data item should be removed from its local

cache when the caching space is not enough for

caching the new data items. The policy of MIX-

GROUP’s cache replacement is based on factors: L,

f, TTL, t and S. These factors, as described above f is

the access count of the data items that reflects a data

item’s popularity in the region; L is used to classify

the data items as primary or secondary (primary data

item is more priority than secondary one); TTL is the

living of the data item, t is timestamp of data item, S

is size of data item.

When the local cache has not enough space for

storing new data item, cache replacement policy will

choose the data item that has invalid TTL value. In

the case, there is not invalid data item; we choose

the secondary data item whose β value is smallest.

This decision is selected because cooperating

caching is not only on the behalf of the caching MU

but also basing on the other MUs need. β value is

evaluated as β = f/ (t*S).

3.4.2 In-zone Query Process

When the data request is missed and is not

responded from LC, the request will be sent to the

neighbors of the MU request. In this process, each

MU has an information profile named ZoneData.

ZoneData stores the information of the neighbors

and data items that they sent to the MU request

before. The ZoneData structure includes {id, L,

sender

, t } where id is the identification of requested

data item; L is the label of data item cached at MU

sender; id

sender

is the identification of MU sender; t is

the time that MU resource gives this data item.

Based on the ZoneData, MU resource can find out

the MU destination that is caching the request data

in its zone. ZoneData is as a way to avoid blind

broadcasting to all MUs in the region that the other

cooperative caching schemes met. By using this

way, our proposed scheme will reduce the

communication cost and increase the system

performance.

3.4.3 Out-zone Query Process

When the data request is not responded from the

neighbours of the MU resource, it will be sent to

MUs network. In this process, we used the

information of the OutZoneData table to discover

the MU destination. The OutZoneData structure

includes {id, ID

sender

, m, t}, where id is the

identification of the data item d; ID

sender

is the

identification of MU network which is storing the

data item d; m

is the number of hops from MU

requester to that MU network; t is the time that MU

requester gets the data item d. The data discovery

path selects MUs network that m value is minimum.

This approach will increase data response with the

least hops.

4 SIMULATION RESULTS

4.1 Simulation Environment

and Experiments

We built a simulation scenario and implemented by

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

156

NS2 simulation (Sinh, 2011). The simulation

parameters are configured as in Table 1. The goal of

the simulation is the quality test of the MIX-GROUP

scheme. The performance metrics used for quality

testing of MIX-GROUP scheme include: average

response time for a data request, average number of

messages, and cache hit ratio. These metrics are

evaluated basing on the MU’s cache size and the

number of MUs varying in system. The

experimental results are compared with those of

other schemes such as GROUPCACHING (Ting,

2007) and COCA (Chow, 2004). The results showed

in the below plots prove the effectiveness of our

proposed approach.

Table 1: Simulation parameters.

Parameters Settings

Simulation area 950m x 510m

Total #MU 10 – 100

Bandwidth 2 Mb/s

Radio range 250 m

Cooperation zone radius 2-3 hops

Total # data items 1000

Client cache size 2000 – 20000 KB

Data item size 1 – 1000 KB

4.2 Simulation Experiment Results

4.2.1 Effect based on Cache Size

a) Average number of messages

The experiment result in Figure 5 showed that the

average number of messages of three approaches

reduces when MU’s cache size increases. However,

MIX-GROUP has the average number of messages

is lowest. This result is archived because MIX-

GROUP has used the information of ZoneData and

OutZoneData to avoid flooding messages while

searching data in system.

Figure 5: Effect of average number of messages based on

MU’s cache size.

b) Average response time

The experimental result of average response time

is shown in Figure 6. Average response time is

steadily decreased in each scheme when MU’s cache

size increases. The plot showed that

GROUPCACHING’s average response time rapidly

decreases but still higher than MIX-GROUP.

Figure 6: Effect of average response time based on MU’s

cache size.

c) Ratio of cache hit

Cache hit ratio result is shown in Figure 7.

Cache hit means the data request is responded from

cooperative zone, while cache miss means the data

request is responded from BS. The plot of

experiment results presented GROUPCACHING’s

cache hit ratio is quite effective. However, cache hit

ratio of MIXGROUP is more effective when MU’s

cache size is large. Achieving this result based on

effective cache admission and replacement strategy.

Figure 7: Effect of cache hit ratio based on MU’s cache

size.

4.2.2 Performance Results based on Number

of MUs

a) Average number of messages

The average number of messages of three

schemes that evaluated based on the different MU

number shown in Figure 8. The result plot presented

the average number of messages of COCA and

GROUPCACHING steadily increases when the

number of MU increases while MIX-GROUP’s

average number of message reduces. This result

Averageofnumbermessage

Averagenumber ofmessagesbased on

MU'sCache Size

MIXGROUP

GROUPCA CHING

COCA

0.5

1.5

2.5

3.5

4.5

Averageresponse time(second)

Averageresponsetimebasedon

MU'sCacheSize

MIXGROUP

GROUPC A CHING

COCA

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Cachehitratio(%)

Cachehitratiobased onMU's CacheSize

MIXGROUP

GROUPCA CHING

COCA

AN EFFICIENT COOPERATIVE CACHE APPROACH IN MOBILE INFORMATION SYSTEM

157

proves MIX-GROUP is very efficient energy

consumption when the MU number increases in

system.

Figure 8: Effect of average number of messages based on

MU number.

b) Average response time

The larger the MU number of system is, the

more cooperative cache the MUs get. The average

response time results are shown in Figure 9. The plot

is shown that MIX-GROUP has average response

time is lowest. This result has got because MIX-

GROUP used the OutZoneDate talbe for searching

data at the MUs network.

Figure 9: Effect of average response time based on MU

number.

c) Ratio of cache hit

The experimental result is shown in Figure 10.

Similar to cache hit ratio based on cache size, plot of

cache hit ratio based on the different MU number in

three schemes increases when the MU number

increases. GROUPCACHING is quite effectiveness

but when the MU number highly increases MIX-

GROUP is more effective than GROUPCACHING.

This result shows that our proposed cooperative

caching scheme is very efficient.

Figure 10: Effect of cache hit ratio based on MU number.

5 CONCLUSIONS

Cooperating caching is more interesting problem in

mobile environment. In this paper, we proposed a

cooperative caching model named MIX-GROUP,

which addresses three basic problems of cooperative

caching: cache discovery, cache admission and

cache replacement. Experimental results show that

MIX-GROUP improved data availability and access

performance. A supplemental evidence for our

success is the average number of messages, the

average response time and cache hit ratio that MIX-

GROUP has archive is more effective than COCA

and GROUPCACHING. However, we still have not

taken care about cache consistency in MIX-GROUP

model yet. This will be eventually considered in our

future work.

REFERENCES

N. Chand, R. C Joshi, M. Misra, 2007.“Cooperative

Caching in Mobile Ad hoc Networks based on Data

Utility”. International Journal of Mobile Information

System 3(1), 19-37.

Prashant Kumar, Naveen Chauhan, L. K. Awasthi,

Narottam Chand, 2010. “Proactive Approach for

Cooperative Caching in Mobile Adhoc Networks”. In

IJCSI 2010, IJCSI International Journal of Computer

Science Issues, Vol.7, Issuse 3, No 8.

Yin, L. Cao, G., 2004. ”Supporting Cooperative Caching

in Ad hoc Networks”. In Infocomm 2004, The 23

Conference on the IEEE Communications Society.

C. Y. Chow, H. V. Leong and A. Chan, 2004. ”Peer to

Peer Cooperative Caching in Mobile Environments”.

In ICDCSW 2004, The 24

International Conference

on Distributed Computing Systems, Workshops,

pp.528-533.

C. Y. Chow, H. V. Leong, and A. Chan, 2004. “Cache

Signatures for Peer to Peer Cooperative Caching in

Mobile Evironments”. In AINA 2004, The 18

Conference on Advanced Information Networking and

Applications, pp.96-101.

C. Y. Chow, H. V. Leong, and A. Chan, 2004. “Group

based Cooperative Cache Management for Mobile

Clients in Mobile Environments”. In ICPP 2004, The

Conference on Parallel Processing, pp.83-90.

N. Chand, R. C Joshi, M. Misra, 2006. ”An efficient

Caching Strategy in Mobile Ad hoc networks Based on

Clusters”. In WCON 2006, The 3

International

Conference on Wireless and Optical Communications

Networks IEEE, Los Alamitos.

Yu Du, Sandeep K. S. Gupta, 2005. “COOP – A

cooperative caching service in MANETs”, In

ICAS/ICNS 2005, Autonomic and Automous System

and International Conference on Networking and

Service.

100

120

10 20 30 40 50 60 70 80 90 100

Averagenumberof message

Averagenumberofmessagebased on

numberofMU

COCA

GROUPCACHING

MIXGROUP

0.5

1.5

2.5

3.5

4.5

10 20 30 40 50 60 70 80 90 100

Averageresponse time(second)

AverageResponse timebased onnumber

ofMUs

MIXGRO UP

GROUPCA CHING

COCA

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

10 20 30 40 50 60 70 80 90 100

Cachehitratio (%)

Cachehitratiobased onnumberofMU

MIXGROUP

GROUPCACHING

COCA

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

158

Yu Du, Sandeep K. S. Gupta, 2009. “Improving on-

demand data access effeciency in MANETs with

cooperative caching”. In Journal Ad hoc Network, vol

7, issue 3, pp.579-598.

G. Radhamani and S. Umamaheswari, 2010. “Comparison

of Cooperative Caching Strategies in Ad-Hoc Network

(MANET)”. In Communications in Computer and

Information Science, Vol 90, part 2, Springer.

Y. W. Ting, Y. K. Chang, 2007. “A Novel Cooperative

Caching Scheme for Wireless Ad hoc Networks:

Group Caching”, In NAS 2007, The International

Conference on Networking, Architecture, an Storage),

IEEE, Los Alamitos.

Sinh V. Vo, Quang X. Pham, 2011. “A Group

Cooperative Caching Approach for Mobile

Information Systems and Building some Experiments”,

Student Final Project Report, Unversity of Science,

Vietnam National University - Ho Chi Minh (report in

vietnamese).

AN EFFICIENT COOPERATIVE CACHE APPROACH IN MOBILE INFORMATION SYSTEM

159