LSJ1319 - A Scalable Server Architecture for Mobile Presence Services in Social Network Applications
Content
A Scalable Server Architecture for Mobile Presence Services in Social Network
ABSTRACT
Social network applications are becoming increasingly popular on mobile devices. A mobile presence service is an essential component co mponent of a social network application because it maintains each mobile user’s presence information, such as the current status (online/offline), !S location and network address, and also updates the user’s online friends with the information continually. "f presence updates occur fre#uently, the enormous number of messages distributed by presence servers may lead to a scalability problem in a large$scale mobile presence service. %o address the problem, we propose an efficient and scalable server architecture, called !resence &loud, which enables mobile presence services to support large$scale social network applications. 'hen a mobile user oins a network, !resence &loud searches for the presence of his/her friends and notifies them of his/her arrival. !resence &loud organies presence servers into a #uorum$based server$to$server architecture for efficient presence searching. "t also leverages a directed search algorithm algor ithm and a one$hop caching caching strategy to achieve achieve small constant constant search latency. 'e analye analye the performance of !resence &loud in terms of the search cost and search satisfaction level. %he search searc h cost is defined as the total number of messages generated generated by the presence server server when a user arrives* and search satisfaction level is defined as the time it takes to search for the arriving us user er’s ’s frie friend nd list list.. %he %he re resu sult ltss of si simu mula lati tions ons de demo mons nstr trat atee th that at !r !res esenc encee &l &lou oud d ac achi hiev eves es performance gains in the search cost without compromising search satisfaction. satisfaction.
Existing System
"n this section, we describe previous researches on presence services, and survey the presence servic ser vicee of e+isti e+isting ng sy syste stems. ms. 'ell 'ell known known commer commercia ciall " sy syste stems ms levera leverage ge some some form form of centralie centr alied d clusters clusters to provide provide presence presence services. -ennings """ et al. al. presented a ta+onomy of different features and functions supported by the three most popular " systems, A", icrosoft S and ahoo0 essenger. %he authors also provided an overview of the system architectures and observed that the systems use client$server$based architectures. Skype, a popular voice over "! application, utilies the lobal "nde+ (") (") technology to provide a presence service service for users. " is a multi$tiered network architecture where each node maintains full knowledge of all availa ava ilable ble users. users. Since Skype is not an open open protoc protocol, ol, it is difficul difficultt to determ determine ine how "
Contact: 040-40274843, 9703109334
Email id: [email protected], [email protected], www.logicsstems.org.in
A Scalable Server Architecture for Mobile Presence Services in Social Network
technology is used e+actly. oreover, 1iao et al. analyed the traffic of S and A" system. %hey found that the presence information is one of most messaging traffic in instant messaging systems. "n, authors shown that the largest message traffic in e+isting presence services is buddy 2%"3 messages. Proposed System:
4ecently, there is an increase amount of interest in how to design a peer$to$peer S"!. !5!S"! has been proposed to remove the centralied server, reduce maintenance costs, and prevent failures in server$based S"! deployment. %o maintain presence information, !5!S"! clients are organied in a 67% system, rather than in a centralied server. 7owever, the presence service servi ce architectu architectures res of -abber and !5!S"! are distribut distributed, ed, the buddy-l buddy-list ist search search problem we defined later also could affect such distributed systems.
"t is noted that few articles in discuss the scalability issues of the distributed presence server architecture. Saint Andre analyes the traffic generated as a result of presence information between users of inter$domains that support the 1!!. 7ouri et al. Show that the amount of presence traffic in S"!89 can be e+tremely heavy, and they analye the effect of a large presence system on the memory and &!: loading. %hose works in study related problems and developing an initial set of guidelines for optimiing inter$domain presence traffic and present a 67%$based presence server architecture. 4ecently, presence services are also integrated into mobile services. 3or e+ample, ;!! has defined the integration of presence service into its specification in :%S. "t is based on S"! protocol, and uses S"!89 to manage presence information. 4ecently, some mobile devices also support mobile presence services. 3or e+ample, the "nstant essaging and !resence Services ("!S) was developed by the 'ireless <illage consortium and was united into 2pen obile Alliance (2A) "!S in 5==>. "n, &hen et al. proposed al. proposed a weakly consistent scheme to reduce the number of updating messages in mobile presence services of "! ultimedia Subsystem ("S). 7owever, it also suffers scalability problem since it uses a central S"! server to perform presence update of mobile users. "n, authors presented the server scalability and distributed management issues in "S$based presence service.
Contact: 040-40274843, 9703109334 Email id: [email protected], [email protected], www.logicsstems.org.in
A Scalable Server Architecture for Mobile Presence Services in Social Network
MODULES?
@. !resen !resence ce &loud &loud server server overlay overlay.. 5. 2ne$ 2ne$ho hop p cachi caching ng stra strate tegy gy.. ;. 6i 6ire rect cted ed buddy buddy se sear arch. ch.
Modules Description ! Prese Presenc ncee Cloud Cloud ser" ser"er er o"er o"erl# l#y y
%he !resence &loud server overlay construction algorithm organies the !S nodes into a server$to$server overlay, which provides a good low$diameter overlay property. %he low$diameter property ensures that a !S node only needs two hops to reach any other !S nodes. $! One%& One%&op op c#c&i c#c&ing ng str#te str#tegy gy
%o improve the efficiency of the search operation, !resence &loud re#uires a caching cachin g strategy strategy to replicate replicate presence information information of users. users. "n order to adapt to changes in the presence of users, the caching strategy should be asynchronous and not re#uire e+pensi e+pe nsive ve mechani mechanisms sms for distri distribut buted ed agreem agreement ent.. "n !re !resen sence ce &loud, &loud, eac each h !S nod nodee maintains a user list of presence information information of the attached users, and it is responsibl responsiblee for for cachi caching ng the the us user er list list of each node in its !S list, in other words, !S nodes only replic rep licate ate the us user er list list at most one hop away from itself. %he cache is updated when neighb nei ghbors ors establ establish ish connect connection ionss to it, and period periodica ically lly updated updated with with it itss neighbo neighbors. rs. %herefore, when a !S node receives a #uery, it can respond not only with matches from its own user list , but also provide matches from its caches that are the user lists offered by all of its neighbors.
'!
Dir Direc ecte ted d (ud (uddy dy se#rc se#rc& &
Contact: 040-40274843, 9703109334
Email id: [email protected], [email protected], www.logicsstems.org.in
A Scalable Server Architecture for Mobile Presence Services in Social Network
'e contend contend that that minim minimii iing ng search searching ing respon response se time time is im impor portan tantt to mobil mobilee presence services. %hus, the buddy list searching algorithm of !resence &loud coupled with the two$hop overlay and one$hop caching strategy strategy ensures that !resence &loud can typically provide swift responses for a large number of mobile users. 3irst, by organiing !S nodes in a server$to$server overlay network, we can therefore use one$hop search e+actly for #ueries and thus reduce the network traffic without significant impact on the search results. Second, by capitaliing the one$hop caching that maintains the user lists of its neighbors, we improve response time by increasing the chances of finding buddies. &learly, this mechanism both reduces the network traffic and response time. ased on the mechanism, mechan ism, the population population of mobile mobile users users can be retrieved retrieved by a broadcasting broadcasting operation operation in any !S node in the mobile mobile presence service. service. oreover, the broadcasting broadcasting message can be piggybacked in a buddy search message for saving the cost.
System Re)uirement Speci*ic#tion: +#rd,#re Re)uirements:
!rocessor
$
!entium B"""
Speed
$
@.@ 7
4A
$
5>C (min)
7ard 6isk
$
5=
Contact: 040-40274843, 9703109334 Email id: [email protected], [email protected], www.logicsstems.org.in
A Scalable Server Architecture for Mobile Presence Services in Social Network
So*t,#re Re)uirements:
2perating System
?
'indowsD>/DE/5===/1!
Application Se Server
?
%omcat>.=/C.1
3ront 9nd
?
7%8, -ava, -sp
Scripts
?
-avaScript.
Server side Script
?
-ava Server !ages.
6atabase
?
ySF8
6atabase &onnectivity?
-6&.
CO-CLUS.O-
"n this paper, we have presented !resence &loud, a scalable server architecture that supports mobile presence services in large$scale social network services. 'e have shown that !resence &loud achieves low search latency and enhances the performance of mobile presence services. "n addition, we discussed the scalability problem in server architecture designs, and introduced the buddy$list search problem, which is a scalability problem in the distributed server architecture of mobile presence services. %hrough a simple mathematical model, we show that the total number of buddy search messages increases substantially with the user arrival rate and the number of presence servers. %he results of simulations demonstrate that !resence &loud achieves maor performance gains in terms of the search cost and search satisfaction. 2verall, !resence &loud is shown to be a scalable mobile presence service in large$scale social network services.
Contact: 040-40274843, 9703109334 Email id: [email protected], [email protected], www.logicsstems.org.in
