Design of Equitable Dominating Set Based Semantic Overlay Networks with Optimal Fast Replica Algorithm for Resilient and Balanced Content Distribution

  • 178 views
Uploaded on

Content Distribution Networks (CDNs) are overlay …

Content Distribution Networks (CDNs) are overlay
networks for placing the content near the end clients with the
aim at reducing the delay, network congestion and balancing
the workload, hence improving the service quality perceived
by the end clients. The main objective of this work is to
construct a semantic overlay network of surrogate servers
based on equitable dominating set. This yields any replication
algorithm that can replicate the contents to minimum number
of surrogate servers within the SON. Such servers can be
accessed from anywhere. Then we propose a content
distribution algorithm named Optimal Fast Replica (O-FR)
and apply our proposed algorithm to distribute the content
over the Equitable Dominating set based Semantic Overlay
Networks (EDSON). We analyze the performance of our
proposed Optimal Fast Replica (O-FR) in terms of average
replication time, and maximum replication time and compare
its performance with existing content distribution algorithms
named Fast Replica and Resilient Fast Replica. The result of
such approach improves the service quality perceived by the
end clients. This paper also analyzes the use of equitable
dominating set for the construction of semantic overlay
networks and also investigates how it is useful for maintaining
the uniform utilization of the surrogate servers.

More in: Technology
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Be the first to comment
    Be the first to like this
No Downloads

Views

Total Views
178
On Slideshare
0
From Embeds
0
Number of Embeds
0

Actions

Shares
Downloads
0
Comments
0
Likes
0

Embeds 0

No embeds

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
    No notes for slide

Transcript

  • 1. ACEEE Int. J. on Network Security, Vol. 01, No. 03, Dec 2010 Design of Equitable Dominating Set Based Semantic Overlay Networks with Optimal Fast Replica Algorithm for Resilient and Balanced Content Distribution J.Amutharaj1, M.Gomathy Nayagam2, and S.Radhakrishnan3 Department of Computer Science and Engineering Arulmigu Kalasalingam College of Engineering Anandnagar, Krishnankoil, Srivilliputtur (Via), Tamil Nadu, India Email: amutharajj@yahoo.com1, m_g_nayagam@rediffmail.com2, srk@akce.ac.in3Abstract— Content Distribution Networks (CDNs) are overlay design of equitable dominating set based semantic overlaynetworks for placing the content near the end clients with the network and optimal fast replica content distributionaim at reducing the delay, network congestion and balancing algorithm in Section III and a discussion on the analyticalthe workload, hence improving the service quality perceived study, experimental results and analyze the performance ofby the end clients. The main objective of this work is toconstruct a semantic overlay network of surrogate servers different content distribution algorithms in Section IV.based on equitable dominating set. This yields any replication Finally, the conclusion and future work is presented inalgorithm that can replicate the contents to minimum number Section V.of surrogate servers within the SON. Such servers can beaccessed from anywhere. Then we propose a content II. RELATED WORKdistribution algorithm named Optimal Fast Replica (O-FR)and apply our proposed algorithm to distribute the content Content Delivery Networks provide services thatover the Equitable Dominating set based Semantic Overlay improve network performance by maximizing bandwidth,Networks (EDSON). We analyze the performance of our improving accessibility, and maintaining correctnessproposed Optimal Fast Replica (O-FR) in terms of average through content replication. They offer fast and reliablereplication time, and maximum replication time and compare applications and services by distributing content toits performance with existing content distribution algorithms surrogate servers located close to users.named Fast Replica and Resilient Fast Replica. The result of In order to offload popular servers and improve end-such approach improves the service quality perceived by the user experience, copies of popular content are often storedend clients. This paper also analyzes the use of equitabledominating set for the construction of semantic overlay in different locations. With mirror site replication, filesnetworks and also investigates how it is useful for maintaining from origin server are proactively replicated at surrogatethe uniform utilization of the surrogate servers. servers with the objective to improve the user perceived Quality of Service (QoS). When a copy of the same file isIndex Terms— SON, Dominating set, CDN, DSON, Optimal replicated at multiple surrogate servers, choosing the serverFast Replica. that provides the best response time is not trivial and the resulting performance can dramatically vary depending on I. INTRODUCTION the server selected [1]. Content Delivery Networks (CDNs) have evolved to Laurent Massoulie [2] proposed an algorithm called theovercome the limitations of the Internet in terms of user localizer which reduces network load, which helps toperceived Quality of Service (QoS) when accessing web evenly balance the number of neighbors of each node incontent. A CDN replicates content from the origin server to overlay, sharing the load and improving the resilience tosurrogate servers, scattered over the globe, in order to random node failures or disconnections.deliver content to end users in a reliable and timely manner Rodriguez [3] and Biersack proposed a dynamicfrom a nearby optimal surrogates. parallel-access scheme to access multiple mirror servers. In Apart from the pure networking issues of the CDNs their study, a client downloads files from mirror serversrelevant to the establishment of the infrastructure, some residing in a wide area network. They showed that theirmore issues such as selection of surrogate server for dynamic parallel downloading scheme achieves significantreplication and retrieval, content replication policy, and downloading speedup with respect to a single servercaching.In this paper, we analyze the role of selection of scheme. However, they studied only the scenario wheresurrogate server for optimum replication of content in the one client uses parallel downloading. They failed toCDN, application of different content replication policies address the effect and consequences when all clientsand their working mechanisms. choose to adopt the same schemeColor figures will be L. This paper is organized as follows. The next section Cherksova [4], and J. Kee proposed Fast Replica algorithmdescribes about the related work. Then, we present our to distribute the content, in which a user downloads different parts of the same file from different servers in© 2010 ACEEE 1DOI: 01.IJNS.01.03.24
  • 2. ACEEE Int. J. on Network Security, Vol. 01, No. 03, Dec 2010parallel. Once all the parts of the file are received, the user III. DESIGN OF EQUITABLE DOMINATING SETreconstructs the original file by reassembling the different BASED SON WITH OPTIMAL FAST REPLICA FORparts. CONTENT DISTRIBUTION Al-Mukaddim Khan Pathan and Rajkumar Buyya [5]presented a comprehensive taxonomy with a broad A. EDSON Based Surrogate Server Selectioncoverage of CDNs in terms of organizational structure, Semantic Overlay Network ‘G’ can be defined ascontent distribution mechanisms, request redirection follows.techniques, and performance measurement methodologies. G = {V, E} ------------------ (1)They studied the existing CDNs in terms of their Where V = {V1, V2, V3, .. Vn} be the set of surrogateinfrastructure, request-routing mechanisms, content servers and E is the set of edges between ith surrogatereplication techniques, load balancing, and cache server and jth surrogate server i.e. E= (Vi, Vj) such that Vimanagement.Dominating sets have been used successfully ≠ Vj.in topology control in wireless Ad hoc networks [6, 7] and Let D be the dominating set of G and D ⊂ G, the servervirtual back creation in sensor networks [8]. not in D is adjacent to at least one surrogate server in D. ZhiHui Lu [9] et al proposed a novel content push Hence, all the surrogate servers are either member of D orpolicy, called TRRR i.e. Tree-Round-Robin-Replica which VD.yields an efficient and reliable solution for distributing Equitable Dominating set D is a set of ‘r’ dominatinglarge files in the content delivery networks environment.They carried out some experiments to verify TRRR vertices in V since D = r and VD is the set of all thealgorithm in small scale. They also demonstrated in adjacent vertices of dominating server set D such that theexperiment that TRRR significantly reduces the file difference between the degrees of all the vertices in D candistribution/replication time as compared with traditional differ utmost by 1. Each vertex v in D has more or lesspolicies such as sequential unicast and multiple unicast. same number of neighbor nodes which are members of Amutharaj. J and Radhakrishnan. S [10, 11] constructed VD. So contents are only replicated in the set of surrogatea dominating set based overlay network to optimize the servers D which contains ‘r’ surrogate servers or less thannumber of servers for replication. They investigated the use ‘r’ number of surrogate server’s i.e. D ≤ V .of Fast Replica algorithm to reduce the content transfertime for replicating the content within the semantic overlay B. Algorithm for Formation of Equitable Dominating Setnetwork and compared its performance with sequential based SON (EDSON):unicast, multiple unicast content distribution strategies interms of content replication time and delivery ratio. Step 1:The algorithm begins by marking all the vertices Srinivas Shakkottai and, Ramesh Johari [12] evaluated of the graph white. Step 2:Algorithm selects the vertex with the maximalthe benefits of a hybrid system that combines peer-to-peerand a centralized client–server approach against each number of white neighbors.method acting alone. They investigated the relative Step 3:The selected vertex is marked black and its neighbors are marked gray.performance of peer-to-peer and centralized client–serverschemes, as well as a hybrid of the two—both from the Step 4:The algorithm then iteratively scans the graypoint of view of consumers as well as the content nodes and their white neighbors, and selects the gray nodedistributor. or the pair of nodes (a gray node and one of its white Ye Xia [13] et al considered a two-tier content neighbors), whichever has the maximal number of whitedistribution system for distributing massive content and neighbors.proposed popularity-based file replication techniques Step 5:The selected node or the selected pair of nodes iswithin the CDN using multiple hash functions. They marked black, with their white neighbors marked gray.developed a set of distributed, robust algorithms and Step 6:Once all the vertices are marked gray or black,evaluated the performance of proposed algorithms. the algorithm terminates. All the black nodes form a Oznur Ozkasap [14], Mine Caglar and Ali Alagoz connected dominating set (CDS). Step 7: After forming the CDS, check the degree of eachproposed and designed a peer-to- peer system; SeCond,addressing the distribution of large sized content to a large vertices of the connected dominating set.number of end systems in an efficient manner. It employed Step 8: If the degree of any vertex vary more than one then mark that vertex gray and find the suitable alternatea self-organizing epidemic dissemination scheme for statepropagation of available blocks and initiation of block vertex as the member of the dominating set and mark ittransmissions. They showed that SeCond is a scalable and black. If no alternate node is found then leave as it is.adaptive protocol which took the heterogeneity of the peers C. Working Principles of Optimal Fast Replica in EDSONinto account. In order to offload popular servers and improve end- user experience, copies of popular content are often replicated in multiple surrogate servers which are scattered over geographically different locations based on some content distribution policy. In this paper, content© 2010 ACEEE 2DOI: 01.IJNS.01.03.24
  • 3. ACEEE Int. J. on Network Security, Vol. 01, No. 03, Dec 2010distribution policies such as sequential unicast, multiple Maximum Replication Time: TimeMax reflects the timeunicast, Fast Replica(FR)[5,16,17,18], Resilient Fast when all the surrogate servers in the overlay networkReplica(R-FR)[5,16,17,18], and Optimal Fast Replica(O- receive k-subfiles (1<=k<=m) of the original file.FR), are used to distribute the content from origin server to TimeMax = max {Timei} where i =1...nset of surrogate servers in the EDSON. In idealistic setting all the nodes and links are The objective of Optimal Fast Replica (O-FR) is to homogeneous, and let each node can support ‘n’ networkminimize the maximum replication time. The working connections to other nodes at B bytes/sec. Then,principle of Optimal Fast Replica can be described as Time distribution = Size (F) / (nxB) ………….. (2)follows. Time collection = Size (F) / (nxB) ………….. (3) Step 1: Partition the Original file F into ‘m’ sub files of B. Performance of Content Distribution Algorithms in anequal size. ‘n’ server Semantic Overlay Network Size (Fi) = Size (F)/ m bytes where 1<=i<=m and m=n/2 Step 2: Surrogate server N0 opens ‘m’ concurrent Time taken for distributing the content over theconnections to surrogate servers N1, N2,….Nm. N0 will send Semantic Overlay Network by different content distributioneach node Ni the following file and information. algorithms are presented in Table I. • Surrogate Server list: R = {N1, N2 ... Nm} (In next TABLE I step, sub-file Fi will be forwarded to this server CONTENT DISTRIBUTION TIMES OF DIFFERENT CONTENT DISTRIBUTION ALGORITHMS list. • Sub-file Fi. Algorithm Content Distribution Time(TD) • Replica amount: k (1<= k <= m). Sequential Unicast n * Size (F) / B Step 3: Every surrogate server Ni ∈ {N1,N2, …Nm}opens Multiple Unicast Size (F) / Bk-1 concurrent connections and replicate the sub file Fi to Fast Replica 2 x Size (F) / (nxB) Resilient Fast Replicathe group with k-1 surrogate servers defined in the set { Nj, without Node Failure 2 x Size (F) / (n x B)i<j<i+k, if j<m, then j=(j-1)mod m+1} Resilient Fast Replica with (2+m/n) * Size (F) / (nxB) In this step, every server Ni {N1,N2,…,Nm} has the Failure of ‘m’ serversfollowing output links and input links. Optimal Fast Replica (( k+n ) / n*n*k ) * Size (F) / B • K-1 Output Links : forwarding sub-file Fi to node list { Nj, i<j<i+k, if j<m, then j=(j-1) mod m+1 } Replication Time proportion of different content • K-1 Input Links : receiving sub-file Fj from server distribution algorithms is tabulated in Table II. list { Nj, i-k <j<i, if j<1, then j = j+m } TABLE II Step 4 : At last, every node Ni holds k sub files, { Fj, i-k REPLICATION TIME PROPORTION OF DIFFERENT CONTENT DISTRIBUTION<j <= i, if j<1, then j = j+m } ALGORITHMSIn general case, node list Ni {N1, N2,…, Nm}, as cacheservers and supports concurrent download. Algorithm Replication Time Proportion Sequential Unicast nClient Content Request Processing: Multiple Unicast 1 Fast Replica 2/n When a user client requests file F from origin server Resilient Fast Replica 2/nthat request will be redirected to the surrogate server list without Node Failure{N1, N2… Nm}, and concurrently downloads every sub-file Resilient Fast Replica with (2+m/n)*1/nFi. Then the sub-files will be reassembled in to original file Failure of ‘m’ serversF in the client machine. Optimal Fast Replica (( k+n ) / n*n*k ) In the ideal case, when k=m, every surrogate server Ni C. Performance of Content Distribution Algorithms inholds all of m sub-files of original file F and reorganizes Equitable Dominating Set based Semantic Overlaythem to form the Original file F in the local node. When the Networkuser requests file F from the origin server, the request will Equitable Dominating set D is a set of ‘r’ dominatingbe redirected to one surrogate server in the list {N1, N2... surrogate servers in surrogate server set V and VD is theNm} and download the whole file F. set of all the adjacent vertices of dominating node set D such that the difference between degree of all the vertices IV. RESULTS AND DISCUSSIONS in D can differ utmost by 1.Each vertex ‘v’ in V has more or less same number of neighbor nodes which are membersA. Analytical Study of the adjacent servers set VD. So contents are only Let Time denote the transfer time of file F from the replicated in the equitable dominated set of surrogateorigin server N0 to surrogate server Ni as measured at Ni. servers D instead of V. Suppose Cardinality of D is ‘r’ or aTwo performance metrics: average and maximum value less than ‘r’ then the contents will be replicated inreplication times are considered. utmost ‘r’ number of surrogate servers which is always less Average Replication Time: i = n than ‘n’. i.e. D ≤ V . ∑ Time i Timeavg = 1/n * i = 1© 2010 ACEEE 3DOI: 01.IJNS.01.03.24
  • 4. ACEEE Int. J. on Network Security, Vol. 01, No. 03, Dec 2010 Therefore, Replication Time proportion of different Performance of different content distribution schemes incontent distribution algorithms such as sequential unicast, terms of Maximum Replication Timemultiple unicast, Fast Replica (FR), Resilient Fast We experimented with 12 different size files; 100 KB,Replica(R-FR), and Optimal Fast Replica (O-FR) in 750 KB, 1.5 MB, 3 MB, 4.5 MB, 6 MB, 7.5 MB, 9 MB, 36EDSON can be expressed as follows: MB, 54 MB, 72 MB, 128 MB and 8 surrogate servers. r : 1: 2/r : (2+m/r)*1/r : (( k+r ) / r*r*k ) where r < n. Maximum Replication Time AnalysisD. Simulation Experimental Study and Analysis: 95 90 85 To evaluate the CDN, we used our complete simulation Max. Replication Time ( in ms) 80 75environment, called CDNsim [24], which simulates a main 70 65 60CDN infrastructure. It is based on OMNeT++ library which 55 50 45provides a discrete event simulation environment. 40 35 30 All CDN networking issues, such as surrogate server 25 20selection, SON formation, replicating the content from 15 10 5origin server to surrogate servers, implementing the 0 B B B B B B B B B B B Breplication algorithms, propagation, and queuing are 5M M M M M M 8M 0K 0K 3M 6M 9M 18 36 54 72 5 10 75 12 4. 1.computed dynamically via CDNsim, which provides a Sequential Unicast File Size Multiple Unicastdetailed implementation of the TCP/IP protocol, Fast Replica R-FR w ith m node Failureimplementing packet switching, packet transmission upon Optimal Fast Replicamisses etc. Fig. 2. Maximum Content Replication Times for various schemesPerformance of different content distribution schemes in “Fig. 2,” shows the maximum replication timeterms of Average Replication Time: measured by different, individual recipient nodes for We experimented with 12 different size files; 100 KB, different file sizes of 100 KB, 750 KB, 1.5 MB, 3 MB, 4.5750 KB, 1.5 MB, 3 MB, 4.5 MB, 6 MB, 7.5 MB, 9 MB, 36 MB, 6 MB, 7.5 MB, 9 MB, 36 MB, 54 MB, 72 MB,128MB, 54 MB, 72 MB, 128 MB and 8 surrogate servers. Fig. MB when 8 surrogate servers are in the replication set.1 shows the average replication time measured by different High variability of maximum replication time underindividual surrogate servers for different file sizes of 100 Sequential Unicast and Multiple Unicast is identified.KB, 750 KB, 1.5 MB, 3 MB, 4.5 MB, 6 MB, 7.5 MB, 9 Maximum File replication time under Optimal Fast ReplicaMB, 36 MB, 54 MB, 72 MB, 128 MB when 8 surrogate (O-FR) algorithm across different file sizes in an 8servers are in the replication set. High variability of surrogate servers replication set are much more stable andaverage replication time under Multiple and Sequential predictable. Hence, Optimal Fast Replica (O-FR)Multicast is identified for larger file sizes. algorithm outperforms most of the cases than sequential Average content replication time under Optimal Fast unicast, multiple unicast, Fast replica, and Resilient FastReplica algorithm across different file sizes in an 8 Replica(R-FR) content distribution schemes.surrogate servers replication set is much more stable and Analysis on the impact of Equitable Dominating Set basedpredictable. Hence, Optimal Fast Replica outperforms SONmost of the cases than sequential unicast, multiple unicast,Fast replica, and Resilient Fast Replica(R-FR) content Analyze the impact of E-DS in SON Formationdistribution schemes. 120 Maximum size of SON Average Replication Time Analysis 100 80 Maximum size of E-DS Maximum size of SON 75 80 based SON s) 70 Avg. Replication Tim ( in m 65 60 60 55 e 50 45 40 40 35 30 25 20 20 15 10 0 5 0 40 60 80 100 120 140 160 180 200 B B B B B B B B B B B B No of Surrogate Servers in the CDN M 8M 5M M M M M 0K 0K 3M 6M 9M 18 36 54 72 5 10 75 12 4. 1. File Size Sequential Unicast Multiple Unicast Fig. 3. Impact of Equitable Dominating Set in SON based CDN Fast Replica R-FR w ith m node Failure Formation Optimal Fast Replica By the implementation of equitable dominating set for Fig.1.Average Content Replication Times for various schemes the clustering of surrogate servers in the SON, the average number of surrogate servers for content replication is reduced to 60 percentages or less. Although the number of surrogate servers is reduced, there will not be any change in the redundancy because of the proposed Optimal Fast© 2010 ACEEE 4DOI: 01.IJNS.01.03.24
  • 5. ACEEE Int. J. on Network Security, Vol. 01, No. 03, Dec 2010Replica(O-FR) content distribution algorithm used for Overlay Network of surrogate servers. This is depicted indistributing the content to different surrogate servers and Fig.5.collect the content from the replica servers and reconstructthem locally. Construction of SON Vs Net Utility 1Performance of different content distribution schemes in 0.9Equitable Dominating Set based SON in terms of Average 0.8 Net Utility Ui 0.7Replication Time: 0.6 0.5 Performance of Content Distribution Algorithms in EDSON 0.4 25 0.3 0.2 20 0.1 A e g R p a nT e( inm ) s 0 v ra e e lic tio im 15 S1 S2 S3 S4 S5 S6 S7 S8 Surrogate Server Number 10 Net Utility of ith Server in SON Net Utility of ith Server in DS based SON 5 Net Utility of ith Server in E-DS based SON 0 Fig. 5. Construction of SON Vs Net Utility B B B B B B B B B B B B 8M 0K 0K M 5M M M M M 3M 6M 9M 18 36 54 72 5 10 75 12 4. 1. File Size Fast Replica Optimal Fast Replica R-FR with m node Failure Fast Replica in DSON V. CONCLUSION AND FUTURE WORK R-FR with m node Failure in DSON Optimal Fast Replica in DSON In this work, first we constructed equitable dominating Fast Replica in EDSON R-FR with m node Failure in EDSON Optimal Fast Replica in EDSON set based semantic overlay network (EDSON) of surrogate Fig. 4. Performance of different Content Distribution Algorithms in EDSON servers for replicating the content from the origin server to a set of surrogate servers with the aim at placing the “Fig. 4,” shows the average replication time measured content nearer to the end user.by different, individual recipient nodes for different file We have conducted simulation experiments usingsizes of 100 KB, 750 KB, 1.5 MB, 3 MB, 4.5 MB, 6 MB, CDNsim and analyzed the performance of content7.5 MB, 9 MB, 36 MB, 54 MB, 72 MB,128 MB when the distribution algorithms in terms of average contentfile is replicated in dominated replication set of surrogate replication time and maximum content replication time forservers. We measured the average replication time of large files over SON. It is found that Optimal Fast Replicadifferent content distribution algorithms such as Optimal (O-FR) algorithm outperforms other content distributionFast Replica(O-FR), Resilient Fast Replica(R-FR) and Fast algorithms.Replica(FR) across different file sizes in both traditional We have performed both analytical study and empiricalSON based CDN as well as DSON based CDN of surrogate study for analyzing the performance of the contentservers. distribution algorithms. We observed that average replication time of all the We also investigated the effect of equitable dominatingthree content distribution algorithms such as Optimal Fast set in SON formation and how it was useful in reducing theReplica (O-FR), Resilient Fast Replica(R-FR), and Fast redundancy. It is also observed that equitable dominatingReplica is reduced due to the use of equitable dominating set based SON is useful in keeping the average replicationset for reducing the number of surrogate servers in which time stable and much more predictable even though thereplication of content carry out. content distribution algorithms differs We also investigatedRole of Equitable Dominating Set and surrogate server that how equitable dominating set based semantic overlayutilization: network is useful in maintaining the net utilization of individual surrogate servers much more stable and balance We evaluate the performance of CDN in terms of Net the load of individual surrogate servers.Utility (Ui ) which can be given by the formula. Ui = 2 / ∏ * arctan (α) ------------- (4) ACKNOWLEDGMENT α – ratio between uploaded bytes to downloaded bytes. The authors would like to thank the Project CoordinatorThe resulting utility value ranges to [0..1]. and Project Directors of TIFAC CORE in NetworkThe value Ui can be Engineering, Arulmigu Kalasalingam College of Ui = 1 if the surrogate server uploads only content Engineering for providing the infrastructure facility in Ui = 0 if the surrogate server downloads only content Open Source Technology Laboratory and also thank Ui = 0.5 if upload and downloads are equal Kalasalingam Anandam Ammal Charities for providing We investigated the use of different overlay financial support for this work.construction methodologies such as Semantic OverlayNetwork (SON), Dominating set based SON (DSON), andEquitable Dominating Set based SON (EDSON). It isobserved that Net Utility Ui of individual surrogate serversis uniform in the Equitable Dominating Set based Semantic© 2010 ACEEE 5DOI: 01.IJNS.01.03.24
  • 6. ACEEE Int. J. on Network Security, Vol. 01, No. 03, Dec 2010 REFERENCES with multiple hash functions in massive content distribution”, The International Journal of Computer and[1] Z. Fei, S. Bhattacharjee, E. W. Zegura and M. H. Ammar, “A Telecommunications Networking, Volume 53 , Issue 1, novel server selection technique for improving the response Elsevier, January 2009. time of a replicated service”, proceedings in IEEE [14] Ozkasap O., Caglar M., Alagoz A. "Principles and INFOCOM, vol. 2. San Francisco, CA, March 1998. performance analysis of SeCond: A system for epidemic[2] Laurent Massoulie, Anne-Marie Kermarree, Ayalvadi peer-to-peer content distribution", Journal of Network and J.Ganesh,”Network Awareness and Failure Resilience in Self Computer Applications, Volume 32, Issue 3, Elsevier, May – Organising Overlay Networks”, Proc. of the 2nd, 2009. International Symposium on Reliable Distributed Systems, [15] K. Stamos, G. Pallis, A. Vakali, D. Katsaros, A. 2003. Sidiropoulos, Y. Manolopoulos: "CDNsim: A Simulation[3] P. Rodriguez and E. Biersack, “Dynamic parallel access to Tool for Content Distribution Networks", ACM Transactions replicated content in the Internet”, IEEE/ACM Transactions on Modeling and Computer Simulation, April 2010. on Networking, 10(4), Aug. 2002.[4] L. Cherksova, J. Kee, “Fast Replica: Efficient Large file Amutharaj Joyson received his Bachelor of Engineering Distribution within Content Delivery Networks”, Proc. of the Degree from Manonmaniam Sundarnar University, Tirunelveli in 4th SENIX symposium on Internet Technologies, March 2002 1999 and Master of Engineering from Madurai Kamaraj[5] A. M. K. Pathan and R. Buyya, “A Taxonomy and Survey of University, Madurai in 2002. He is currently doing his doctoral CDNs”, Technical Report, GRIDS-TR-2007-4, The program from Anna University, Chennai. He is a member of CSI, University of Melbourne, Australia, Feb. 2007. IAEng, ISTE and Network Technology Group of TIFAC-CORE[6] Bo Han and Weijia Jia,”Design and Analysis of Connected in Network Engineering. He has published one research paper in Dominating Set Formation for Topology Control in Wireless International Journal of Networks, and presented four research Adhoc Networks”, Proc. of 14th International Conference papers in International Conferences and Twenty research papers on Computer Communication and Networks, Oct 2005. in National Conferences in Network Engineering. His research[7] Lu, K.Bolla,J .A. Huynh, D. T, ”Adapting connected d-hop interest include Content Distribution Networks, Mobile Adhoc Dominating Sets to Topology changes in Wireless Adhoc Networks, Network Secuirty, Distributed Computing, Real time Networks”, Proc. of 25th IEEE International Performance, Systems, and Evolutionary Optimization. Computing and Communication Conference, April 2006. M. Gomathynaygam, received his Bachelor of Engineering[8] Chi Ma, Y. Yang, and Z. Zhang, “Constructing Battery- degree from Manonmaniam Sundaranar University, Tirunelveli Aware Virtual Backbone in Sensor Networks”, in the Proc. and Master of Engineering Degree in Network Engineering from of the International Conference on Parallal Processing Anna University, Chennai in 2006. He is a member of CSI, (IEEE – ICPP ’05), IEEE Computer Society. IAEng, ISTE and and Network Technology Group of TIFAC-[9] ZhiHui Lu, WeiMing Fu, ShiYong Zhang, YiPing Zhong, CORE in Network Engineering. His research interests include “TRRR: A Tree-Round-Robin-Replica Content Replication Network Security, Distributed Computing, Grid Computing, Algorithm for Improving Fast Replica in Content Delivery Cloud Computing, Wireless Networks, and Mobile Adhoc Networks”, in the proceedings of 4th International Networks. Conference on Wireless Communications, Networking and S. Radhakrishnan received his Master of Technology in 1988 Mobile Computing, 2008. and Ph.D degree in 1992 from Institute of Technology, Banaras[10] Amutharaj. J and Radhakrishnan. S, “Dominating Set based Hindu University, Banaras, India. He is the Director and Head of Semantic Overlay Networks for Efficient Content Computer Science and Engineering Department in Arulmigu Distribution”, proceedings of IEEE ICSCN - 2007, vol. 1. Kalasalingam College of Engineering, Srivilliputhur. He is a Madras Institute of Technology, Anna University, Chennai, member of ISTE. He is a Principal Investigator of Research Feb 2007. Promotion Scheme (RPS) project funded by Department of[11] Amutharaj. J, and Radhakrishnan. S, “Dominating Set Science and Technology, Government of India. He is the Project Theory based Semantic Overlay Networks for Efficient and Director of Network Technology Group of TIFAC-CORE in Resilient Content Distribution”, Journal of Networks, Network Engineering. This prestigious project is funded by Academy Publishers, Vol 3, March 2008. TIFAC, Department of Science and Technology, Government of[12] Srinivas Shakkottai, Ramesh Johari, “Demand-Aware India and Cisco Systems. He has produced eight Ph. D’s and Content Distribution on the Internet”, IEEE/ACM currently guiding twelve Research Scholars in the areas of Transaction on Networking, Vol. 18, No.2. April 2010. Network Engineering, Network Processors, Network Security,[13] Ye Xia, Shigang Chen, Chunglae Cho, Vivekanand Sensor Networks, Optical Networks, Wireless Networks, Korgaonkar, “Algorithms and performance of load-balancing Evolutionary Optimization and Bio-medical Instrumentation.© 2010 ACEEE 6DOI: 01.IJNS.01.03.24