Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more
Download
Standard view
Full view
of .
Look up keyword
Like this
2Activity
0 of .
Results for:
No results containing your search query
P. 1
Priority Based Mobile Transaction Scheme Using Mobile Agents

Priority Based Mobile Transaction Scheme Using Mobile Agents

Ratings: (0)|Views: 68 |Likes:
Published by ijcsis
We define a priority based mobile transaction scheme in which mobile users can share data stored in the cache of a mobile agent which is a special mobile node for coordinating the sharing process. This framework allows mobile affiliation work group to be formed dynamically with a mobile agent and mobile hosts. Using short range wireless communication technology, mobile users can simultaneously access the data from the cache of the mobile agent. Data Access Manager module at the mobile agent enforces concurrency control using cache invalidation technique. Four levels of priority are assigned to the requesting mobile nodes based on available energy and connectivity. This model supports disconnected mobile computing by allowing mobile agent to move along with the Mobile Hosts. The proposed Transaction frame work has been simulated in J2ME and NS2 and performance of this scheme is compared with existing frame works.
We define a priority based mobile transaction scheme in which mobile users can share data stored in the cache of a mobile agent which is a special mobile node for coordinating the sharing process. This framework allows mobile affiliation work group to be formed dynamically with a mobile agent and mobile hosts. Using short range wireless communication technology, mobile users can simultaneously access the data from the cache of the mobile agent. Data Access Manager module at the mobile agent enforces concurrency control using cache invalidation technique. Four levels of priority are assigned to the requesting mobile nodes based on available energy and connectivity. This model supports disconnected mobile computing by allowing mobile agent to move along with the Mobile Hosts. The proposed Transaction frame work has been simulated in J2ME and NS2 and performance of this scheme is compared with existing frame works.

More info:

Published by: ijcsis on Feb 15, 2011
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

11/29/2012

pdf

text

original

 
Priority Based Mobile Transaction Scheme UsingMobile Agents
J.L.Walter Jeyakumar
#1
, R.S.Rajesh
#2
 
 Department of Computer Science and Engineering, Manonmaniam Sundaranar University,Tirunelveli, Tamilnadu, INDIA.
1
walterjeya@hotmail.com
2
rs_rajesh@yahoo.co.in
 Abstract
 
We define a priority based mobile transactionscheme in which mobile users can share data stored in the cacheof a mobile agent which is a special mobile node for coordinatingthe sharing process. This framework allows mobile affiliationwork group to be formed dynamically with a mobile agent andmobile hosts. Using short range wireless communicationtechnology, mobile users can simultaneously access the data fromthe cache of the mobile agent. Data Access Manager module atthe mobile agent enforces concurrency control using cacheinvalidation technique. Four levels of priority are assigned to therequesting mobile nodes based on available energy andconnectivity. This model supports disconnected mobilecomputing by allowing mobile agent to move along with theMobile Hosts. The proposed Transaction frame work has beensimulated in J2ME and NS2 and performance of this scheme iscompared with existing frame works.Key words – Transaction, concurrency control, mobile database,cache invalidation, mobility
I.
 
I
NTRODUCTION
 Mobile computing environment consists of Fixed Hosts(FHs), Mobile Hosts (MHs) and Base stations or MobileSupport Stations (MSSs). MH is connected to the Fixednetwork through MSS via wireless channels. TheGeographical area covered by a MSS is called a cell. MobileHosts are portable computers which move around in a cell.When a MH enters into a new cell hand-off or hand-over takesplace. MH communicates only with the MSS responsible forits cell. Transactions and data management functions are doneusing the data base servers installed at MSS. In mobilecomputing, it is necessary that a computation is not disruptedwhile an MH is not connected. The part of the computationexecuting on an MH might continue executing concurrentlywith the rest of the computations while the MH is moving andnot connected to the network [8]. With the evolution of PCS( personal Communication System) and GSM(GlobalSystem for Mobile communication), advanced wirelesscommunication services are being offered to the mobile users.Mobile Database System is a distributed client/server systembased on PCS or GSM in which clients can move aroundfreely while performing their data processing activities inconnected or disconnected mode. Frequent disconnections,mobility, limited battery power and resources pose newchallenges to mobile computing environment. Frequent abortsdue to disconnection should be minimized in mobiletransactions. The low and variable bandwidth of wirelessnetwork together with the expensive transmission cost makesbandwidth consumption an important concern [2]. Correctnessof transactions executed on both fixed and mobile hosts mustbe ensured by the operations on shared data. Blocking of mobile transactions due to long disconnection periods shouldbe minimized to reduce communication cost and to increaseconcurrency [3]. After disconnection, mobile host should beable to process transactions and commit locally. In Mobilecomputing, there is always a competition for shared data sinceit provides users with the ability to access information throughwireless connections that can be retained even while the useris moving. Further, mobile users are required to share theirdata with others. This provides the possibility of concurrentaccess of data by mobile hosts which may result in datainconsistency. Concurrency control methods have been usedto control concurrency. Due to limitations and restrictions of wireless communication channels, it is difficult to ensureconsistency of data while sharing takes place.In this paper, we present a priority based mobiletransaction frame work that allows mobile users to share datacached in the Mobile Agent which is a special node forcoordinating the sharing process. Whenever an MH enters intoa Mobile Agent area it can connect and access the data in thecache. But upon update request by a MH, updation is done atthe local cache and invalidation report is sent to all the mobilehosts which have already accessed the same data. This willforce the mobile hosts to refresh their data values. Thisframework also provides the provision for transaction updateduring disconnection. Data Access Manager (DAM) at theMobile Agent will take care of concurrency control whilesharing takes place. Concurrency control is enforced usingcache invalidation technique. In order to give priority to themobile nodes running on low power and with lowconnectivity, four levels of priority are used.
 
We also take intoaccount the mobility of the Mobile Hosts that has a strongimpact on mobile applications [4].The remaining part of this paper is organized as follows.Section II summarizes the related research. Section III focuseson the Mobile Agent based architecture. Section IV specifiesthe proposed framework for disconnected mobile computing.Section V gives the performance analysis. In Section VI, the
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 9, No. 1, January 201186http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
discussion on the proposed model is presented. Finally,Section VII concludes the paper.II.
 
R
ELATED
W
ORK
 When simultaneous access to data is made at the server,concurrency control techniques are employed to avoid datainconsistency. Conventional locking based
 
concurrencycontrol methods like centralized Two Phase locking anddistributed Two Phase locking are not suitable for mobileenvironment. In centralized two phase locking scheme [17,19], where one node is responsible for managing all lockingactivities, the problem of single point failure cannot beavoided. The distributed two phase locking scheme used in[18], allows all nodes to serve as lock managers. But in theevent of data partition, this algorithm could degenerate into acentralized two phase scheme. In conventional lockingscheme, the communication overhead that arises due tolocking and unlocking requests can create a seriousperformance problem because of low capacity and limitedresources in mobile environment [5]. Moreover, it makesmobile hosts to communicate with the server continuously toobtain and manage locks [6].The timestamp approach for serializing the execution of concurrent transactions was developed for the purpose of more flexibility, to eliminate the cost of locking, and to caterfor distributed database systems [7, 13]. In timestamping, theexecution order of concurrent transactions is defined beforethey begin their execution. The execution order is establishedby associating a unique timestamp to every transaction. Whentwo transactions conflict over a data item, their timestamps areused to enforce serialization by rolling back one of theconflicting transactions. To exploit the dynamic aspects of two phase locking and the static ordering of timestamping, anumber of concurrency control techniques were developedusing a combined approach. In mixed approach techniquescalled Wound-wait and Wait-die [7, 16], the enforcement of mutual exclusion among transactions is carried out usinglocking while conflicts are resolved using timestamps.In [14], optimistic concurrency control scheme is used tominimize locking overhead by delaying lock operation untilconflicting transactions are ready to commit. They rely onefficiency in the hope that conflicts between transactions willnot occur. Without using lock during the execution of thetransactions, this scheme promotes deadlock free execution.In optimistic concurrency control with dynamic time stampadjustment protocol, client side write operations are required.But it may never be executed due to delay in execution of atransaction [1]. In multi version transaction model [9], data ismade available as soon as a transaction commits at a mobilehost and another transaction can share this data. But data maybe locked for a longer time at a mobile host before the lock isreleased at the database server.In [11], AVI (Absolute Validity Interval) was introducedfor enforcing concurrency control without locking. AVI is thevalid life span of a data item. But it calculates AVI onlybased on previous update interval. In [12], a method based onPLP (Predicted Life Period), which takes care of thedynamicity of the life time of data was proposed. Here, lifespan of data is predicted based on the probability of updationof data item. This method makes PLP of data item very closeto the actual valid life span of a data item.In [10], a transaction model for supporting mobilecollaborative works was proposed. This model makes use of Export-Import repository, which is a mobile sharing work space for sharing data states and data status. But in the Export-Import repository based model, locking is the main techniquewhich has the following disadvantages. (i) More bandwidth isneeded for request and reply since the locking and unlockingrequests have to be sent to the server. (ii) Disconnection of mobile host or a transaction failure will result in blocking of other transactions for a long period. Our framework is betterthan the model which uses Export-Import repository forsharing data since it minimizes message communication costsand data update costs to a larger extent. Also disconnectedmobile hosts are treated separately within a mobile affiliationby waiting for their reconnection.
 
Transaction Managementsolution proposed in [15], is for reducing energy consumptionat each MHs by allowing each MH to operate in three modes,Active, Doze, and Sleep thus providing a balance of energyconsumption among MHs.III.
 
M
OBILE
A
GENT
B
ASED
A
RCHITECTURE
 The proposed Mobile Agent based architecture model isillustrated in Fig 1. The model consists of Mobile Hosts,
-
 
Disconnected Mobile Host
MSS
– Mobile SupportStation-
 
Connected Mobile Agent
DBS –
Data BaseServer
-
Mobile Affiliation
BSC –
Base StationController
-
Wireless LAN
 -
Short range wireless
 
Fig.1. Mobile Agent Based Architecture model
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 9, No. 1, January 201187http://sites.google.com/site/ijcsis/ISSN 1947-5500
 
Mobile Support Stations (MSS), and Data Base Servers (DBS).MSS is connected with a Base Station Controller (BSC) [20],which coordinates the operations of MSS using its own storedprogram. Unrestricted mobility is supported by wireless link between MSS and Mobile Hosts. Each MSS serves one cellwhose size depends on the power of its MSS. Data BaseServers are connected to the mobile system through wiredlines as separate nodes. Each DBS can be reached by anyMSS and new DBSs can be connected and old ones can betaken out from the network without affecting mobilecommunication. A DBS can communicate with a MH only viaMSSs.Mobile agent is a special mobile node which connects tothe MSS to cache the frequently accessed data. DisconnectedMobile Hosts can connect to the Mobile Agent using shortrange wireless communication technologies to form mobileaffiliation workgroup.Mobile hosts are allowed to access data from the cache.When data request is made for the first time, data is retrievedfrom the server and stored in the cache. Subsequent requestsare handled by the Data Access Manager module itself. Whena mobile host requests for data update, after local updation of the data item, invalidation report is sent to all the mobile hoststhat have already accessed the same data. This makes all the
 
mobile hosts to refresh their data values. When a mobile hostis disconnected from the Mobile agent after updation request,the updation task is transferred to the Data Access Manager inthe Mobile Agent. Data Access Manager module is used tocoordinate the operations in the cache.After disconnected from the server, Mobile Agent canmove along with the connected MHs and MHs can continuetheir transaction execution. If data update at the server isrequested, mobile agent will wait for reconnection beforeupdation is made.IV.
 
M
OBILE
T
RANSACTION
F
RAMEWORK
 When Mobile Hosts enter into the Mobile Agent area, theyconnect to the Mobile Agent using short range wirelessnetwork technology to form Mobile Affiliation Work Group.Frequently accessed data are cached in the Mobile Agent.Mobile Hosts can access the cached data in the Mobile Agent.The Data Access Manager module at the Mobile Agent isresponsible for enforcing concurrency and cache invalidation.A.
 
 Energy and Connectivity Evaluation
Mobile Hosts all the time maintains its energy availabilityand connectivity. Connectivity is evaluated based on signalstrength. When signal strength goes below one fourth of totalstrength, connectivity is considered as Low. When available
 
energy goes below 25% of total energy level, then
 
energy
 
availability is considered as Low. The status of an MH basedon Energy Availability and Connectivity (A
ij
) can be
 
A
00
 
Low Energy & Low Connectivity, A
01
– Low Energy & HighConnectivity, A
10
–High Energy & Low Connectivity and A
11
 
High Energy & High Connectivity. When Data AccessManager receives a transaction request from a mobile host, itassigns a priority level using A
ij
. A mobile host with lowenergy and low connectivity is assigned the highest priority.Other levels of priority are assigned according to the variouspossibilities as given in Table I.
TABLE IL
EVELS OF
 
P
RIORITY
 
Status of an MH(A
ij
)EnergyAvailability(i)Connectivity(j)PriorityA
00
Low Low 1A
01
Low High 2A
10
High Low 3A
11
High High 4B.
 
Concurrency Control Mechanism
When more number of mobile hosts are accessing datasimultaneously the problem of data inconsistency arises. Thisproblem can be solved if we use an efficient concurrencycontrol mechanism. When data request is made for the firsttime, data is retrieved from the server and stored in the cache.Future requests for data are managed directly by the DataAccess Manager.Data Access Manager uses a suitable data item format tostore data as quintuple [12] in the cache. It has (id, TLU, PLP,dataval, NT) where id denotes unique Id of the data item, TLUindicates time of Last Update, PLP is Predicted Life Period,dataval is current value of the data item and NT denotesnumber of transactions that concurrently access the data item.When Data Access Manager fetches data for the first timefrom the server, it sets TLU to current time, PLP to optimaltime based on the nature of data item and NT to 1. NT isincremented whenever a new data access request is made.Data in the cache becomes invalid, once it is updated in theserver. Life span of a data item is predicted using PLP. Itmakes use of the probability of updation as a basis for settingvalid life span of a data item. In PLP interval, data item isvalid and all the mobile hosts can access same data itemconcurrently.When a MH makes update request or PLP expires, thedata item is invalidated. Now PLP is modified andinvalidation report is sent. The predicted life period of dataitem is computed using the following formula given in [12].PLP=PPLP ± (p*PPLP)Where PPLP is Previous Predicted Life Period and p ispredicted probability of updation of data item. p =Total_updates / NT. It is the ratio of data item update to dataitem access. Since predicted probability of updation is basedon recent past history of updation rate, it is highly probablethat PLP is very close to the actual validity interval of the dataitem.
C. Transaction Execution in the MH 
After connecting to the Mobile Agent ,
 
the MH intimates thestatus of Energy availability and Connectivity (A
ij
) of the MHto DAM at Mobile Agent. Then, the execution of the
(IJCSIS) International Journal of Computer Science and Information Security,Vol. 9, No. 1, January 201188http://sites.google.com/site/ijcsis/ISSN 1947-5500

You're Reading a Free Preview

Download
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->