energy efficient coordinated cooperative cache replacement algorithms for social wireless networks

8/12/2019 Energy Efficient Coordinated Cooperative Cache Replacement Algorithms for Social Wireless Networks

1/8

Energy efficient coordinated cooperative

cache replacement Algorithms for Social

wireless networks

surabattina sunanda1, abdul rahaman shaik21Computer science and engineering.2Computer science and engineering

1surabattina.sunandagmail.com, 2rahaman.!"!1gmail.com

Cooperative caching is a techni#ue

used in wireless networks to improve theefficiency of information access by reducing

the access latency and bandwidth usage.inthis paper,we discuss about cooperative

caching policies for minimi$ing electronic

content provisioning cost in Social %ireless&etworks 'S%&E(). S%&E(s are formed

by mobile devices, such as data enabled

phones, electronic book readers etc., sharingcommon interests in electronic content, and

physically gathering together in public

places. Electronic ob*ect caching in suchS%&E(s are shown to be able to reduce the

content provisioning cost which depends

heavily on the service and pricing

dependences among various stakeholdersincluding content providers 'C+), network

service providers, and End Consumers

'EC).Cache replacement policy plays asignificant role in response time reduction

by selecting suitable subset of items for

eviction from the cache. n addition, this

paper suggests some alternative techni#uesfor cache replacement. -inally, the paper

concludes with a discussion on future

research directions.

Keywords:Data Caching, CacheReplacement,SWNETs,Cooperative caching,

content provisioning, ad hoc networks

1.INTRODUCTION:ECE&( emergence of data enabled

mobile devices and wireless/enabled dataapplications have fostered new content

dissemination models in today0s mobile

ecosystem. A list of such devices includes

Apple0s i+hone, oogle0s Android,Ama$on0s indle, and electronic book

readers from other vendors. (he array of

data applications includes electronic bookand maga$ine readers and mobile phone

Apps. (he level of proliferation of mobile

applications is indicated by the e3ample factthat as of 4ctober 2!1!,Apple0s App Store

offered over 1!!,!!! apps that are

downloadable by the smart phone users.

%ireless mobile communication is a fastest

growing segment in communication

industry516. t has currently supplemented orreplaced the e3isting wired networks in

many places. (he wide range of applications

and new technologies5"6 simulated this

enormous growth. (he new wireless trafficwill support heterogeneous traffic,

consisting of voice, video and data. %irelessnetworking environments can be classified

in to two different types of architectures,

infrastructure based and ad hoc based. (he

former type is most commonly deployedone, as it is used in wireless 7A&S and


2/8

global wireless networks. An infrastructure

based wireless network uses fi3ed network

access points with which mobile terminalsinteract for communication and this re#uires

the mobile terminal to be in the

communication range of a base station. (head hoc based network structure alleviates

this problem by enabling mobile terminals to

cooperatively form a dynamic networkwithout any pre e3isting infrastructure. t is

much convenient for accessing information

available in local area and possibly reaching

a %7A& base station,which comes at nocost for users.

8obile terminals available today have

powerful hard ware, but the capacity of thebatteries goes up slowly and all these

powerful components reduce battery life.(herefore ade#uate measures should be

taken to save energy. Communication is one

of the ma*or sources of energy onsumption.9y reducing the data traffic energy can be

conserved for longer time. :ata caching has

been introduced51!6 as a techni#ues to

reduce the data traffic and access latency. 9ycaching data the data re#uest can be served

from the mobile clients without sending it to

the data source each time. t is a ma*ortechni#ue used in the web to reduce the

access latency. n web, caching is

implemented at various points in thenetwork. At the top level web server uses

caching, and then comes the pro3y server

cache and finally client uses a cache in the

browser.

(his paper provides algorithms,for Energy

efficient cooperative cache replacement inwireless networks. (he topic of caching in

ad hoc networks is rather new, and not much

work has been done in this area. %eclassified the replacement policies for

8A&E(s in to two groups uncoordinated

and coordinated. n uncoordinated

replacement, the data item to be evicted is

determined independently by each node

based on its local access information. n

coordinated replacement policy,the mobilenodes which forms cooperative cache

collectively takes the replacement

decision.n this paper we discuss aboutcoordinated replacement policy.

2.RELATED WORK:(here is a rich body of the e3isting

literature on several aspects of cooperative

caching including ob*ect replacements,reducing cooperation overhead, cooperation

performance in traditional wired networks.

(he Social %ireless &etworks e3plored in

this paper, which are often formed using

mobile ad hoc network protocols, aredifferent in the caching conte3t due to their

additional constraints such as topologicalinsatiability and limited resources. As a

result, most of the available cooperative

caching solutions for traditional staticnetworks are not directly applicable for the

S%&E(s. (hree caching schemes for

8A&E( have been presented5;65116.

in. n the first scheme, Cache:ata, aforwarding node checks the passing/by

ob*ects and caches the ones deemed usefulaccording to some predefined criteria. (hisway, the subse#uent re#uests for the cached

ob*ects can be satisfied by an intermediate

node. A problem with this approach is thatstoring large number of popular ob*ects in

large number of intermediate nodes does not

scale well.(he second approach, Cache+ath,is different in that the intermediate nodes do

not save the ob*ects< instead they only

record paths to the closest node where the

ob*ects can be found. (he idea in Cache+athis to reduce latency and overhead of cache

resolution by finding the location of ob*ects.

(his strategy works poorly in a highlymobile environment since most of the

recorded paths become obsolete very soon.

(he last approach in is the =ybridCache inwhich either Cache:ata or Cache+ath is


3/8

used based on the properties of the passing/

by ob*ects through an intermediate node.

%hile all three mechanisms offer areasonable solution, and that relying only on

the nodes in an ob*ect0s path is not most

efficient. >sing a limited broadcast/basedcache resolution can significantly improve

the overall hit rate and the effective capacity

overhead of cooperative caching. Accordingto the protocols the mobile hosts share their

cache contents in order to reduce both the

number of server re#uests and the number of

access misses. (he concept is e3tended infor tightly coupled groups with similar

mobility and data access patterns. (his

e3tended version adopts an intelligent bloom

filter/based peer cache signature to minimi$ethe number of flooded message during cache

resolution. A notable limitation of thisapproach is that it relies on a centrali$ed

mobile support center to discover nodes with

common mobility pattern and similar dataaccess patterns. 4ur work, on the contrary, is

fully distributed in which the mobile devices

cooperate in a peer/to/peer fashion for

minimi$ing the ob*ect access cost. nsummary, in most of the e3isting work on

collaborative caching, there is a focus on

ma3imi$ing the cache hit rate of ob*ects,without considering its effects on the overall

cost which depends heavily on the content

service and pricing models. (his paperformulated two ob*ect replacement

mechanisms to minimi$e the provisioning

cost, instead of *ust ma3imi$ing the hit rate.

Also, the validation of our protocol on a realS%&E( interaction trace with dynamic

partitions, and on a multiphone Android

prototype is uni#ue compared to the e3istingliterature. -rom a user selfishness

standpoint, 7aoutaris et al.investigate its

impacts and mistreatment on caching. Amistreated node is a cooperative node that

e3periences an increase in its access cost

due to the selfish behavior by other nodes in

the network. n, Chun et al study selfishness

in a distributed content replication strategy

in which each user tries to minimi$e its

individual access costby replicating a subset of ob*ects locally 'up

to the storage capacity), and accessing the

rest from the nearest possible location.>sing a game theoretic formulation, the

authors prove the e3istence of a pure &ash

e#uilibrium under which network reaches astable situation. Similar approach

has been used in which the authors model a

distributed caching as a market sharing

game.4ur work in this paper has certain similarity

with the above works as we also use a

monetary cost and rebate for content

dissemination in the network. =owever, asopposed to using game theoretic approaches,

we propose and prove an optimal cachingpolicy. Analysis of selfishness in our work is

done in a steady state over all ob*ects

whereas the previous works mainly analy$ethe impact of selfishness only for a single

data item. Additionally, the pricing model of

our work which is based on the practical

Ama$on indle business model issubstantially different and practical

compared to those used earlier.

3.Network Model,Ca!e Re"lae#e$t

%ol&&e':

3.1 NETORK MODEL

-ig. 1 illustrates an e3ample S%&E( withina >niversity campus. End Consumers

carrying mobile devices form S%&E(

partitions, which can be either multi/hop

'i.e.,8A&E() as shown for partitions 1, ?,and @, or single hop access point based as

shown for partition 2. A mobile device can

download an ob*ect 'i.e., content) from theC+0s server using the CS+0s cellular

network, or from its local S%&E( partition.

%e consider two types of S%&E(s. (hefirst one involves stationary S%&E(


4/8

partitions. 8eaning, after a partition is

formed, it is maintained for sufficiently long

so that the cooperative ob*ect caches can beformed and reach steady states. %e also

investigate a second type to e3plore as to

what happens when the stationaryassumption is rela3ed.(o investigate this

effect, caching is applied to S%&E(s

formed using human interaction tracesobtained from a set of real S%&E( nodes.

-ig. 1. Content access from an S%&E( in a

>niversity Campus.

3.2. Ca!e Re"lae#e$t

Caching in wireless environment has uni#ue

constraints like scarce bandwidth, limited

power supply, high mobility and limitedcache space. :ue to the space limitation,

the mobile nodes can store only a subset of

the fre#uently accessed data. (heavailability of the data in local cache can

significantly improve the performance

since it overcomes the constraints inwireless environment. A good replacement

mechanism is needed to distinguish the

items to be kept in cache and that is to

be removed when the cache is full. %hile it

would be possible to pick a random ob*ectto replace when cache is full, system

performance will be better if we choose anob*ect that is not heavily used. f a heavily

used data item is removed it will probably

have to be brought back #uickly, resultingin e3tra overhead. So a good

replacement policy is essential to achieve

high hit rates. (he e3tensive research on

caching for wired networks can be adapted

for the wireless environment withmodifications to account for mobile terminal

limitations and the dynamics of the wireless

channel.

3.3. Ca!e Re"lae#e$t %ol&&e' &$ Ad

!o $etwork'

:ata caching in 8A&E( is proposed as

cooperative caching. n cooperative caching the

local cache in each node is shared among the

ad*acent nodes and they form a large unified

cache. So in a cooperative caching

environment, the mobile hosts can obtain data

items not only from local cache but also from

the cache of their neighboring nodes. (his aimsat ma3imi$ing the amount of data that can be

served from the cache so that the server delays

can reduced which in turn decreases the

response time for the client. n manyapplications of 8A&E( like automated

highways and factories, smart homes and

appliances, smart class rooms, mobile nodes

share common interest. So sharing cache

contents between mobile nodes offers significant

benefits. Cache replacement algorithm greatly

improves the effectiveness of the cache by

selecting suitable subset of data for caching.(he available cache replacement mechanisms

for ad hoc network can be categori$ed in to

coordinated and uncoordinated depending on

how replacement decision is made. n

uncoordinated scheme the replacement decision

is made by individual nodes. n order to cachethe incoming data when the cache is full,

replacement algorithm chooses the data items

to be removed by making use of the local

parameters in each node.


5/8

3.(.)"l&t Ca!e Re"lae#e$t

*&+.2. Ca!e "art&t&o$&$+ &$ '"l&t a!e "ol&.

(o reali$e the optimal ob*ect placementunder homogeneousob*ect re#uest model we

propose the following Split Cache policy in

which the available cache space in eachdevice is divided into a duplicate segment -fraction) and a uni#ue segment 'see -ig. 2).

n the first segment, nodes can store themost popular ob*ects without worrying

about the ob*ect duplication and in the

second segment only uni#ue ob*ects are

allowed to be stored. (he parameterin -ig.2 '! B 1) indicates the fraction of cache

that is used for storing duplicated ob*ects.

%ith the Split Cache replacement policy,soon after an ob*ect is downloaded from the

C+0s server, it is categori$ed as a uni#ue

ob*ect as there is only one copy of this

ob*ect in the network. Also, when a nodedownloads an ob*ect from another S%&E(

node, that ob*ect is categori$ed as a

duplicated ob*ect as there are now at leasttwo copies of that ob*ect in the network.

-or storing a new uni#ue ob*ect, the leastpopular ob*ect in the whole cache is selected

as a candidate and it is replaced with the

new ob*ect if it is less popular than the newincoming ob*ect. -or a duplicated ob*ect,

however, the evictee candidate is selectedonly from the first duplicate segmentof the

cache. n other words, a uni#ue ob*ect isnever evicted in order to accommodate a

duplicated ob*ect.

(.E$er+ eff&&e$t oord&$ated

oo"erat&/e a!e re"lae#e$t

Al+or&t!#' for )o&al w&rele''

$etwork'

(.1 Coord&$ated Ca!e re"lae#e$t%ol&&e'

Coordinated cache replacement strategy for

cooperative caching schemes in mobile

environments should ideally consider cache

admission control policy. Cache admissioncontrol decides whether the incoming data is

cacheable or not. Substantial amount of

cache space can be saved by properadmission control, which can be utili$ed to

store more appropriate data, therebyreducing the number of evictions. f a nodedoesn0t cache the data that ad*acent

nodes have it can cache more distinct data

items which increase the data availability.

(here is coordination between theneighboring nodes for the proper placing of

data. Another feature of coordinated

replacement is that the evicted data may bestored in neighboring nodes which have free

space. Some of the replacement policies

which make use of coordinated cachereplacement are given below.

TD)

(he cache replacement 5"6 is based on twoparameters distance ':) which is measured

as the number of hops and access fre#uency.

As the network is mobile the value ofdistance ':) may become obsolete. So the

value is chosen based on the time at which it

is last updated. (he ( value is obtained by

the formula 1tcur/ tupdate. :istanceis updated by looking at the value of (.

9ased on how the distance and time isselected three different schemes are

proposed (:SD:,(:SD( and

(:SD&.(:SD: considers distance as the

replacement criteria. f two data items havethe same distance least value of ':() is


6/8

replaced. n (:SD( the replacement

decision is made by selecting the data with

lowest ( value. n the third scheme productof distance and access fre#uency is

considered. n these algorithms (:SD: has

the lower success rate and (:SD( has thehigher hit ratio.

LU0 M&

(his replacement scheme 5F6, has two parts

replacement and migration. (he replacement

decision is based on a utility value formed

by combining the parameters accessprobability, distance, si$e and coherency. n

the migration part the replaced data is stored

in the neighboring nodes which have

sufficient space. -or migration the data withhighest utility value is given preference.

=ere even though the replacement decisionis made locally migration is a coordinated

operation. n order to save the cache space

the data item is cached based on the locationof the data source. f it is from the same

cluster the data is not cached. (he limitation

of this scheme is that no checking is done

whether the data is already present in themigrating node.

ECOR%

Energy efficient cooperative cache

replacement problem 'EC4+) 5G6 is an

energy efficient cache replacementpolicy used in ad hoc networks. =ere the

replacement decision is done based on the

energy utili$ation for each data access. -or

this, they considered the energy for in$one communication, energy for sending the

ob*ect,energy for receiving and energy cost

for forwarding the ob*ect. 9ased on this theyproposed a dynamic EC4+ :+

and EC4+ Dgreedy algorithms to replace

data. (he neighboring nodes will not cachethe same data item in its local cache which

reduces the redundancy and increases

hit ratio.

Co$t 0etor

n this scheme 5H6, each data item maintains

a count which gives the number of nodeshaving the same data. %henever the cache is

full, data item with ma3imum count is

removed first as this will be available in theneighboring nodes. %henever a data item is

removed from the cache the access count

will be decremented by one. nitially whenthe data is brought in to cache the

count is set to $ero. (able1 shows a

comparison of different cache replacement

policies.

. D&'''&o$ a$d *tre work

8ost of the replacement algorithms used in

ad hoc networks is 7> based which uses

the property of temporal locality. (his isfavorable for 8A&E( which is formed for a

short period of time with small memorycapacity. -re#uency based algorithms will

be beneficial for long term accesses. t is

better if the function based policies canadapt to different workload condition. n

these schemes if we are using too many

parameters for finding the value function,

which are not easily available theperformance can be degraded. 8ost of the

replacement algorithms mentioned above

uses cache hit ratio as the performancemetric. n wireless network the cost to

download data item from the server may

vary. So in some cases this may not be thebest performance metric. Schemes which

improve cache hit ratio and reduce access

latency should be devised. n cooperative

caching coordinated cache replacement ismore effective than local replacement since

the replacement decision is made by

considering the information available in theneighboring nodes. (he area of cache

replacement in cooperative caching has not

received much attention. 7ot of work needsto be done in this area to find better

replacement policies.

Al+or&t!# %ara#eter' E/&t&o$ %erfor#a$e Ad/a$ta+e D&'ad/a$ta+e


7/8

Co$'&dered #ea're

TD) :istance and

access

fre#uency

7ow access

rate and

lowest distance

Success rate,

cache hit ratio.

Ialue of

distance is

updated.

:oesn0t consider

recency of data

item.

LU0 M& Access

probability,si$e, coherence

and

distance.

7ow access

probability,bigger si$e,

low

consistency,lowest

distance

9yte hit ratio,

average #uerylatency,

message

overhead

Evicted data

isstored in

ad*acent

nodes

&o checking is

done beforestoring data.

ECOR% Energy for in$one

communication,

sending ob*ect,receiving ob*ect

7owest energyvalue

Cache hit ratio,average access

delay

Energy istaken

as the

importantparameter

Computing energyfor each task is

not easy

Co$t 0etor Access count 8a3imumaccess count

Average accesstime

Coordinatedsimple to

implemen

:ata redundancyis high.

Tale.1 Co#"ar&'o$ of a!e re"lae#e$t "ol&&e' for oo"erat&/e a!&$+

4. Co$l'&o$

n this paper we made a general comparisonof the ma*orreplacement policies in wireless

networks and summari$ed the main points.

&umerous replacementpolicies are proposedfor wireless networks, but a few for

cooperative caching in ad hoc networks. %e

alsosummari$ed the operation, strengths anddrawbacks of these algorithms. -inally we

provided some alternatives for cache

replacement and identified topics for futureresearch.

5.Refere$e'516 C. Aggarwal, J.7. %olf, and +.S. Ku,

LCaching on the %orld %ide %eb,M EEE(rans. nowledge and :ata Eng.,

vol. 11, no. 1, pp. ;@/1!G, Jan.-eb. 1;;;526. :enko, 8.., (ian, J.,Cross/7ayer

:esign for Cooperative Caching in 8obile

Ad =oc &etworks, +roc .of EEEConsumer Communications and &etworking

Conf' 2!!H).

5?6. 7. Kin, . CaoN Supporting cooperative

caching in ad hoc networks, EEE

(ransactions on 8obile Computing, "'1)NGG/H;' 2!!F).

5@6. Chand, &. Joshi .C., and 8isra, 8.,

Efficient Cooperative Caching in Ad =oc&etworks Communication System

Software and 8iddleware.'2!!F).

5"6 S. 7im, %. C. 7ee, . Cao, C. . :asN Anovel caching scheme for internet based

mobile ad hoc networks. +roc .12th

nt. Conf. Computer Comm. &etworks

'CCC& 2!!?), ?H/@? ' 4ct. 2!!?).5F6 &arottam Chand, .C. Joshi and 8ano*

8isra, OCooperative Caching Strategy in

8obile Ad =oc &etworks 9ased on

Clusters,O nternational Journal of %ireless+ersonal Communications special issue on

Cooperation in %ireless &etworks, Iol. @?,ssue 1, pp. @1/F?, 4ct 2!!G

5G6 7i, %., Chan, E., P Chen, :. '2!!G).

Energy/ efficient cache replacement policiesfor cooperative caching in mobile


8/8

ad hoc network. n +roceedings of the EEE

%C&C 'pp??@;Q??"@).

5H6. 9. R heng, J. u, and :. 7 ee. Cacheinvalidation and replacement strategies for

location dependent data in mobile

environments,. EEE (ransactions onComputers, "1'1!) N 11@1Q11"?, 4ctober

2!!2.

5;6. 8ary 8agdalene Jane.-, Kaser &ouhand . &adara*an,L&etwork :istance 9ased

Cache eplacement +olicy for 7ocation/

:ependent :ata in 8obile EnvironmentM,

+roceedings of the 2!!H &inth nternationalConference on 8obile :ata 8anagement

%orkshops ,EEE Computer Society

%ashington,:C,>SA,2!!H.

51!6.umar, A., Sar*e, A.. and 8isra, 8.T+rioritised +redicted egion based Cache

eplacement +olicy for location dependentdata in mobile environment0, nt. J. Ad =oc

and >bi#uitous Computing , Iol. ", &o. 1,

'2!1!) pp."FQFG.5116. 9. Rheng, J. u, and:. 7. 7ee. cache invalidation and

replacement strategies for location/

dependent data in mobile environments.

EEE (rans. on Comp, "1'1!)N1@Q21, 2!!2.5126 U. en and 8. :hunham. >sing

semantic caching to manage location

dependent data in mobile computing. +roc.4f AC8EEE 8obiCom, ;;N21!Q221,

2!!!.

51?6 . 7ai, R .(ari, and +. 9ertok .8obility aware cache replacement for

location dependent information services. n

(echnical eport ( / !@/!@ ' 8 (

School of C S P ( ),2!!@.

surabattina sunandapursuing her 8.(ech degr

computer engineering from Agudur. She completed her 9

from narayana

college,gudur.=er resinterests focus on networking

Abdul rahaman shaik receiv8.(ech degree in com

engineering from avs cet .

currently working as Asprofessor in ASCE( gud

research interests focus

networking

energy efficient coordinated cooperative cache replacement algorithms for social wireless networks

Documents