ENHANCING MULTICAST COMMUNICATIONS WITH DATA-IN-NETWORK SCHEMES GUO HUAQUN NATIONAL UNIVERSITY OF SINGAPORE 2005 ENHANCING MULTICAST COMMUNICATIONS WITH DATA-IN-NETWORK SCHEMES GUO HUAQUN (B. ENG., M. Eng., NUS) A THESIS SUBMITTED FOR THE DEGREE OF PHD OF ENGINEERING DEPARTMENT OF ELECTRICAL & COMPUTER ENGINEERING NATIONAL UNIVERSITY OF SINGAPORE 2005 i To the memory of my mother Guo Lanying ii ACKNOWLEDGEMENTS The author would like to extend her utmost gratitude to the people who, in one way or other have aided and contributed to the successful completion of this Ph.D. thesis. First of all, the author would like to express her sincere gratitude to her supervisors Professor Lawrence Wong Wai Choong and Dr Ngoh Lek Heng for their kind supervision, gracious guidance, a global view of research, strong encouragement and detailed comments throughout the duration of the research. The author would also like to thank the Institute for Infocomm Research (I2R) for offering the research scholarship. Thanks are also given to management of I2R, SingAREN (Singapore Advanced Research & Education Network) and National University of Singapore for providing the opportunity for this research. Special thanks to Ms Yee Poh Cheng, Dr Chai Teck Yoong, Dr Cheng Heng Seng, Ms Tan Joo Geok, Ms Vasudha Ramnath, Dr Dong Ligang in Data-In-Network research group for their help and creating a supportive research environment. Not forgetting Ms Yang Ping and Ms Nazreen Beevi d/o Saifuddin for their help in collecting data. Last but not least, the heartiest gratitude is given to the author’s family for their love and encouragement. iii TABLE OF CONTENTS ACKNOWLEDGEMENTS ii TABLE OF CONTENTS iii SUMMARY ix NOMENCLATURE xi LIST OF FIGURES xviii LIST OF TABLES xxii CHAPTER 1: INTRODUCTION 1.1 Survey of Outstanding Challenges in Multicast Communications 1.1.1 Lack of Multicast across Time 1.1.2 Limited Address Space & Large Delay in Application-level Multicast 1.1.3 Scalability of IP Inter-domain Multicast 1.1.4 Summary of Survey 1.2 Data-In-Network Technology 1.2.1 Background 10 1.2.2 DIN-based Solutions 11 1.2.2.1 VIN: Video-In-Network 14 1.2.2.2 DINCast 14 1.2.2.3 DINPeer 15 1.2.2.4 (G)MPLS-based DINloop_Net 15 1.3 Summary & Research Contributions 15 1.4 Organization of Thesis 17 iv CHAPTER 2:VIN: A SCALABLE VIDEO DISTRIBUTION SERVICE 19 2.1 Introduction 19 2.2 Related Works of Networked Video Services 20 2.2.1 Video-On-Demand 20 2.2.2 Parallel-Server Model 22 2.2.3 Multicast-based Video-On-Demand Services 22 2.2.4 Content Distribution Network 23 2.3 VIN Framework 24 2.4 VIN System Parameters 28 2.5 Analysis of VIN 30 2.5.1 Effect of Core Network Bandwidth 37 2.5.2 Effect of DIN Node Outgoing Link Bandwidth 41 2.5.3 Effect of Cycle Time & Core Network Access Time 42 2.5.4 Effect of Packet Set Size 44 2.6 Performance Study 45 2.7 Cost Analysis 50 2.8 Conclusions 51 CHAPTER 3:COMPARISON OF VIN WITH STAGGERED MULTICAST 53 3.1 Effect of Processor Number on VIN 53 3.1.1 Effect of Core Network Bandwidth on Number of Concurrent Streams 58 3.1.2 Effect of DIN Node Outgoing Link Bandwidth on Number of Concurrent Streams 59 v 3.1.3 Effect of Cycle Time & Core Network Access Time on Number of Concurrent Streams 3.1.4 Effect of Packet Set Size on Number of Concurrent Streams 60 60 3.2 Staggered Multicast 62 3.3 VIN versus Staggered Multicast 65 3.3.1 Resource Requirement 66 3.3.2 Scalability Analysis 67 3.3.3 Startup Latency Analysis 69 3.4 Conclusions 71 CHAPTER 4: DINCAST: OPTIMIZING DELAY IN APPLICATION-LEVEL SHARED-TREE MULTICAST 4.1 Overview of DINCast 73 73 4.1.1 Overview 73 4.1.2 Key Features of DINCast 75 4.2 Details of DINCast 78 4.2.1 Hop Depth Definition 78 4.2.2 Selection of DIN Nodes 79 4.2.3 Automatic Formation of Loop 81 4.2.4 Optimization of Loop 82 4.2.5 DIN Node State 85 4.3 Experimental Results 85 4.3.1 Simulation Model 86 4.3.2 Effect of Various Parameters on Multicast Delay 89 4.3.3 Decision Table for Selecting DINloop_App 95 vi 4.3.4 Performance Improvement of Retrieving Messages directly from DINloop_App 97 4.4 Multicast Delay Analysis 100 4.4.1 Effect of Bandwidth 102 4.4.2 Delay Analysis in Shared-tree Multicast 108 4.4.3 Scheme A 109 4.4.4 Scheme B 124 4.4.5 Scheme C 128 4.4.6 Comparison of Three Schemes 133 4.5 Conclusions 136 CHAPTER 5: DINPEER: OPTIMIZING APPLICATION-LEVEL MULTICAST DELAY IN P2P OVERLAY NETWORK 138 5.1 Solution Overview 138 5.2 Details of DINPeer 140 5.2.1 Generalizing DINloop_App 140 5.2.1.1 A Big Outer Ring 140 5.2.1.2 Spiral-Ring Method 141 5.2.1.3 Bandwidth of DINloop_App 147 5.2.2 Integrating with P2P Overlay Network 149 5.2.3 Optimizing P2P Application-Level Multicast 152 5.3 Evaluation of DINPeer 5.3.1 Multi-point Communications using DINPeer 153 153 5.3.2 Performance of Multi-point Communication with Data Persistency 155 vii 5.3.3 Performance of Divisible Load Computing using DINPeer 157 5.3.4 Effects of DINPeer in Synchronous and Asynchronous Collaboration 5.4 Conclusions 161 166 CHAPTER 6: (G)MPLS-BASED DINLOOP_NET: IMPROVING INTER-DOMAIN MULTICAST SCALABILITY 167 6.1 Introduction 167 6.2 Solution Overview 169 6.3 Implementing DINloop_Net with MPLS 171 6.4 Solution – Combining Control Plane & Data Plane 173 6.4.1 Label Stack 173 6.4.2 Assign Label Stack to Multicast Message 174 6.4.3 Forward Multicast Message in Inter-domain 174 6.4.4 Retrieve Multicast Message 175 6.5 Solution - Separating Control Plane & Data Plane 176 6.5.1 Setting up Steiner Tree for Multicast Traffic 176 6.5.2 Traffic Aggregation 178 6.5.3 Assigning Label Stack to Multicast Packet 179 6.6 Experimental Results 180 6.6.1 Message Load Evaluation 180 6.6.2 Inter-domain Router Forwarding State Size Evaluation 183 6.6.3 Inter-domain Multicast Delay Evaluation 186 6.7 Discussion 187 6.8 Implementing DINloop_Net with GMPLS 188 viii 6.8.1 GMPLS Label 189 6.8.2 Creating DINloop_Net with GMPLS 191 6.9 Conclusions CHAPTER 7: CONCLUSIONS 192 194 7.1 Conclusions 194 7.2 Future Work 196 PUBLICATION LIST 199 REFERENCES 201 APPENDIX A: COST ANALYSIS 216 APPENDIX B: PAMETERS FOR TRANSIT-STUB MODEL 227 APPENDIX C: COMMANDS IN GT-ITM 228 APPENDIX D: READABLE GRAPH FORMAT 229 APPENDIX E: SPIRAL-RING METHOD 230 APPENDIX F: LENGTH OF DINLOOP_NET 237 APPENDIX A: COST ANALYSIS 222 A.4.2 Storage Capacity The storage capacity per fibre is given by: StorageCapacity = Bandwidth * CirculationDelay Length _ of _ Fiber = Bandwidth * Speed _ of _ Light _ in _ Fiber The number of 2-hour Mbps videos which can circulate in VIN is given by: Number _ of _ Videos = = StorageCapacity Length _ of _ Video StorageCapacity * 60 * 60 * * 10 and Number_of_Content_Hours = Number_of_Videos*2 Table A.2 shows the storage capacity for various values of CirculationDelay and its corresponding fibre length when Bandwidth = 81.92 Tbps (refer to Appendix 1, Section 1). Table A.2 Storage Capacity of VIN Delay(ms) Fibre Length(km) Storage Number of 2- Capacity hour Mbps (GBytes) Videos Number of Content Hours 0.25 50 0.711 0.948 1.896 0.5 100 1.422 1.896 3.792 200 10.24 3.793 7.586 400 20.48 7.585 15.170 800 40.96 15.170 30.340 1200 61.44 22.756 25.512 1600 81.92 30.341 60.682 10 2000 102.4 37.926 75.852 20 4000 204.8 75.852 151.704 40 8000 409.6 151.704 303.408 60 12000 614.4 227.556 455.112 80 16000 819.2 303.407 606.814 100 20000 1024 379.259 758.518 APPENDIX A: COST ANALYSIS 223 Note that up to a certain point, it would not be feasible to circulate more videos in VIN as the fiber length required will be astronomical and the circulation delay will be too long for VIN to be effective. It is generally acknowledged that a large number of requests in a VOD system are for a small group of highly popular videos. These will be the videos which are circulating in the network. The other videos are stored in the secondary storage of VIN. A.4.3 Cost Comparisons From the above, we can estimate the cost of the VIN-based VOD system, dimensioned to be comparable to that provided by NCube, as follows: In the following calculation, whenever there is a range of possible values for a parameter, we will use the average value for the parameter. Suppose the amount of fibre in the system is x km and 1000km or less of the fibre is installed in the network while the rest are on a fibre drum. Then, x * 1000 * 0.6875 ⎧ TotalFiber Cost = ⎨ ⎩( x − 1000 ) * 1000 * 0.003125 + 1000 * 1000 * 0.6875 x ≤ 1000 x > 1000 We assume D+Ts = 2tp (The more detail can be referred in Figure 3.10). Then, F= D + Ts 2t p = = ms / kBytes Sp S p 600 Thus, the buffer size of DIN Node is BufferSize ≥ Sp = 2t p * 600 MBytes *1000 Total DIN Node Cost = 100*($21,156 + BufferSize*$0.029) Total Secondary Storage Cost = (200,000 content hours) *60*60*3/8*$0.00004 = $10,800 APPENDIX A: COST ANALYSIS 224 The current cost per stream for a typical client/server based VOD service is in the range of $250 per stream [130]. From [131], the projection disk transfer rate in 2010 is twice that in 2006. For estimation, we assume that the projection disk transfer rate in 2009 is twice that in 2006. Assuming that disk I/O is a dominant cost of the NCube system, the total cost for NCube projected for 2009 can be estimated by: Total NCube Cost = 50,000*$300/2 = $62,500,000 From Section A3.3, the maintenance costs for VIN can be estimated by: Total Maintenance Cost = 0.175*(Total DIN Node Cost) It is not clear if the $300 per VOD stream includes maintenance costs. We therefore compute the total VIN cost with and without maintenance costs as follows: Total VIN Cost = Total Fibre Cost + Total DIN Node Cost +Total Secondary Storage Cost + Amplifier Cost + [Total Maintenance Cost] To facilitate the cost comparison with NCube, we define a cost advantage factor, CAF thus: Cost Advantage Factor, CAF = Total VIN Cost/Total NCube Cost Table A.3 shows the CAF computed for various values of x km of fibre with and without including the maintenance costs. From Table A.3, we observe that when x < 1000 km, as fibre length increases, the installed fibre costs are correspondingly increased, resulting in a more significant increase in CAF for both CAF1 and CAF2. However, beyond 1000 km, the CAF is roughly constant for both CAF1 and CAF2. This is due to the fact that when x >= 1000 km, the costs are mainly dominated by the DIN Node costs, total secondary storage cost, as well as the installed fibre costs, which are fixed regardless of the increased fibre length of the fibre drum. This phenomenon is illustrated graphically in Figure A.3. APPENDIX A: COST ANALYSIS 225 Table A.3 Cost Comparison of VIN-based VOD with NCube VOD using CAF Fibre Length (km) 50 Number of 2hour Mbps Videos 0.948 Total Fibre Cost 34,375.00 Total VIN Cost without Maintenance 2,163,485.44 Total VIN Cost with Maintenance 2,533,715.51 CAF1 CAF2 0.3462 0.4054 0.5 100 1.896 68,750.00 2,197,860.87 2,568,091.02 0.3517 0.4109 200 3.793 137,500.00 2,266,611.74 2,636,842.04 0.3627 0.4219 400 7.585 275,000.00 2,404,113.48 2,774,344.09 0.3847 0.4439 800 15.17 550,000.00 2,679,116.96 3,049,348.18 0.4287 0.4879 1200 22.756 684,531.25 2,813,651.69 3,183,883.52 0.4502 0.5094 1600 30.341 684,687.50 2,813,811.42 3,184,043.86 0.4502 0.5094 10 2000 37.926 685,000.00 2,814,127.40 3,184,360.45 0.4503 0.5095 20 4000 75.852 685,625.00 2,814,769.80 3,185,005.89 0.4504 0.5096 40 8000 151.704 686,875.00 2,816,054.60 3,186,296.78 0.4506 0.5098 60 12000 227.556 688,125.00 2,817,339.40 3,187,587.67 0.4508 0.5100 80 16000 303.407 689,375.00 2,818,624.20 3,188,878.56 0.4510 0.5102 100 20000 379.259 690,625.00 2,819,909.00 3,190,169.45 0.4512 0.5104 Propagation Delay (ms) 0.25 Notes: CAF1 is the Cost Advantage Factor computed without VIN maintenance costs. CAF2 is the Cost Advantage Factor computed with VIN maintenance costs. If the traditional server/client based VOD system adds more servers to a server cluster for more concurrent streams, the VIN system can add more DIN Nodes to increase the concurrent stream number as well. The VIN system can also add more fibre to increase the fibre storage capacity for more popular videos. From the cost analysis of the VIN-based VOD service, we can draw the following conclusions: • The costs of the VIN-based VOD service are dominated by the costs of the optical components in the DIN Node, total secondary storage cost, as well as the installed fibre costs. APPENDIX A: COST ANALYSIS The cost advantage of the VIN-based VOD service over the client/server based NCube service can potentially be by a factor of 0.3462 to 0.5104. Thus, the cost of VIN-based VOD in 2009 is projected to be about half the cost of NCube VOD. 1.0 CAF1 0.9 Cost Advantage Factor • 226 CAF2 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 50 100 150 200 250 300 350 400 Number of Videos Figure A.3 Cost Comparison of VIN-based VOD and NCube VOD APPENDIX B: PARAMETERS FOR TRANSIT-STUB MODEL 227 APPENDIX B: PARAMETERS FOR TRANSIT-STUB MODEL When we create a graph with transit-stub model, we create a specification file with the parameters for the transit-stub model. One example of the specification file ts6000.txt is shown below. # vi ts6000 # [] # # [] [] # number of nodes = 1x25(1+12x20) = 6025 ts 47 12 0 20 1.0 25 20 0.5 20 10 0.5 On running 'itm ts6025', transit-stub graphs with 6025 nodes each will be created starting with initial seed 47. Line 12 0 means each transit node will have 12 stub domains connected to it; there shall be no extra transit-stub edges and no extra stubstub edges. Line 20 1.0 means there will be transit domain in each graph. Line 25 20 0.5 says each transit domain will have, on an average, 25 transit nodes with an edge between each pair of nodes with a probability of 0.5 and Line 20 10 0.5 says every stub domain shall have, on an average, 20 nodes each with an edge between every pair of nodes with a probability of 0.5. The above mentioned values are the example values of one set of parameters. We change the values of each set of parameters for different simulations accordingly. APPENDIX C: COMMANDS IN GT-ITM 228 APPENDIX C: COMMANDS IN GT-ITM vi ts6000 # to edit the specification file for the parameters in transit-stub model itm ts6000 # to generate graphs for one set of parameters ls # list files: ts6000-0.gb, ts6000-1.gb, ts6000-2.gb, ts6000-3.gb, ts60004.gb, ts6000-5.gb. sgb2alt ts6000-0.gb new0 # converts the file’s format sgb2alt ts6000-1.gb new1 # converts the file’s format sgb2alt ts6000-2.gb new2 # converts the file’s format sgb2alt ts6000-3.gb new3 # converts the file’s format sgb2alt ts6000-4.gb new4 # converts the file’s format sgb2alt ts6000-5.gb new4 # converts the file’s format APPENDIX D: READABLE GRAPH FORMAT 229 APPENDIX D: READABLE GRAPH FORMAT GRAPH (#nodes #edges id uu vv ww xx yy zz): 6025 58948 transtub(0,12,0,0,{1,694,3,1.000,0.000,0.000},{25,348,3,0.500,0.000,0 .000},{20,70,3,0.500,0.000,0.000}) 694 VERTICES (index name u v w x y z): T:0.0 235 351 T:0.1 192 350 T:0.2 464 445 T:0.3 205 231 T:0.4 159 427 T:0.5 265 343 . . . 6020 S:0.24/14.16 198 501 6021 S:0.24/14.17 200 450 6022 S:0.24/14.18 183 507 6023 S:0.24/14.19 150 499 6024 S:0.24/14.20 207 499 EDGES (from-node to-node length a b): 321 36 301 54 268 20 264 55 230 25 . . . 6020 6023 48 6020 6024 6021 6024 49 6022 6024 25 6023 6024 57 APPENDIX E: SPIRAL-RING METHOD 230 APPENDIX E: SPIRAL-RING METHOD A system parameter g defines the number of adjacent nodes in the outer ring to be considered for an inner link. The detailed steps of spiral-ring method are described below. In the below algorithm, bandwidth means the end-to-end bandwidth from node1 to node2. The bandwidth measurement is the transmission speed or throughput of the connection between two end nodes. In our method, we assume that the bandwidth is bidirectional and rather static. We assign that DINloop_App in clockwise direction. 1. Form an inner spiral ring. (a) Start from any node (thisnode) in the outer ring and choose anticlockwise or clockwise as one direction for the below distance counting. We choose clockwise direction like DINloop_App direction in Figure 5.3a. (b) thisnode requests the results from near nodes within the g adjacent nodes in the outer ring and obtain maximum bandwidth max_bandwidth from itself to one of g nearby nodes. (c) Pass max_bandwidth with its associated two nodeIds to its neighbour node in the outer ring. (d) Neighbour node requests the results from nearby nodes within the g adjacent nodes in the outer ring and obtain maximum bandwidth bandwidth2 from itself to one of g nearby nodes. (e) max_bandwidth = max (bandwidth2, max_bandwidth). (f) Repeat Steps 1(c), 1(d) and 1(e) g times so as to obtain the first inner link with local maximum bandwidth max_bandwidth and its associated two nodeIds. The first inner link is shown in Figure 5.3a. APPENDIX E: SPIRAL-RING METHOD 231 (g) Assign its associated first nodeId as startnode, and its associated second nodeId as nextnode. bandwidth_this = max_bandwidth. Record bandwidth_this in startnode. startnode changes the entry of Successor-DIN-Node into nodeId of nextnode. nextnode changes the entry of Predecessor-DIN-Node into nodeId of startnode. n = 2. n is the number of nodes in the inner spiral ring. (h) nextnode sends a query message to g nearby endnodes (endnode_1, endnode_2, …, endnode_g) in the outer ring. (i) If startnode ∉ {endnode_1, endnode_2, …, endnode_g}: endnode_1, endnode_2, …, and endnode_g respond to the query message respectively. bandwidth = max (bandwidth, bandwidth, …, bandwidth). Record bandwidth in nextnode. nextnode changes the entry of successor-DIN-Node into nodeId of endnode_i. bandwidth_this = (bandwidth_this, bandwidth). endnode_i changes the entry of Predecessor-DIN-Node into nodeId of nextnode. n = n+1. nextnode = endnode_i. (This process forms the next inner link and is indicated in thick dash-line in Figure 5.3b.) Go to Step 1(h). (j) If startnode ∈ {endnode_1, endnode_2, …, endnode_g}: APPENDIX E: SPIRAL-RING METHOD 232 startnode sends a message to inform nextnode to stop the spiral-ring forming process. startnode sends a query message to g nearby endnodes starting from nextnode in the inner ring at anticlockwise direction. endnode_1, endnode_2, …, and endnode_g respond to the query message respectively. bandwidth = max (bandwidth, bandwidth, …, bandwidth< endnode_g, startnode>). Record bandwidth< endnode_i, startnode> in endnode_i. startnode changes the entry of Predecessor-DIN-Node into nodeId of endnode_i. bandwidth_this = (bandwidth_this, bandwidth). endnode_i informs its successor DIN Nodes to clear the entries of Predecessor-DIN-Node and Successor-DIN-Node. endnode_i changes the entry of Successor-DIN-Node into nodeId of startnode. n = n-i. (This process is indicated in thin dash-line in Figure 5.3d.) Go to Step 2. Using the above steps, we obtain the inner spiral ring (Figure 5.3d) with the ring bandwidth bandwidth_this and the number of nodes n in the inner spiral ring. 2. Drop the inner link with lower bandwidth. (a) Assume N is the desirable number of DIN Nodes and fb is the ring bandwidth increasing rate. APPENDIX E: SPIRAL-RING METHOD 233 (b) If n is smaller than N, go to Step 2(g). (c) n = 1. bandwidth_next = bandwidth record in startnode. dropnode = startnode. (d) If bandwidth recorded in dropnode is less than bandwidth_this*(1+ fb): dropnode sends a query message to the nearby endnodes along the inner ring one by one at clockwise direction. * If the query message is not received by startnode that means endnode exists that bandwidth is not less than bandwidth_this*(1+ fb): dropnode records bandwidth in the bandwidth entry, If n =1, bandwidth_next = bandwidth. If n >1, bandwidth_next = (bandwidth_next, bandwidth). dropnode changes the entry of its Successor-DIN-Node into the nodeId of endnode. endnode informs its Predecessor DIN Nodes along the inner ring before dropnode to clear their entries of Predecessor-DIN-Node and Successor-DIN-Node. endnode changes the entry of its Predecessor-DIN-Node into the nodeId of dropnode. endnode becomes dropnode. n = n+1. Go to Step 2(d). * If the query message is received by startnode: APPENDIX E: SPIRAL-RING METHOD 234 startnode sends a message to inform dropnode to stop the dropping process. Startnode sends a query message to near endnodes starting from dropnode along the inner ring one by one in an anticlockwise direction: endnode_1, endnode_2,…, endnode_i. If endnode_i exists that bandwidth is not less than bandwidth_this*(1+ fb): endnode_i records bandwidth in the bandwidth entry. bandwidth_next = ([...]... overloading and singlepoint of failure Finally, the current inter-domain multicast faces one nagging problem in the area of managing large IP inter-domain multicast routing tables 1.2 Data- In- Network Technology In this thesis, we propose a new Data- In- Network (DIN) technology to overcome the outstanding technical challenges facing multicast The general idea of DIN technology is to configure multiple DIN... Scalability of IP Inter-domain Multicast A nagging problem with IP multicast is large IP multicast routing tables The problem of large IP multicast routing tables also exists in the inter-domain multicast Multicast Source Discovery Protocol (MSDP) and Border Gateway Multicast Protocol (BGMP) were developed for inter-domain multicast However, MSDP requires each multicast router to maintain forwarding state... message in shared-tree multicast Tv Length of a video VBW DIN Node common outgoing link bandwidth xvii VI Core network bandwidth VmA_i Incoming bandwidth/outgoing bandwidth at a node (i.e., Node i) with Function A VmB_x Incoming bandwidth/outgoing bandwidth at a node (i.e., Node x) with Function B VmC_1 Incoming bandwidth/outgoing bandwidth at first DIN Node with Function C VmC_d Incoming bandwidth/outgoing... application-level multicast suffers some other disadvantages in terms of high multicast delay, overloading at Rendezvous Point (RP) and single point of failure Finally, a nagging problem with IP multicast is that IP inter-domain multicast routing tables are large In this thesis, we propose a new Data- In- Network (DIN) technology to overcome the above outstanding technical challenges The general idea of DIN is to... Video -In- Network (VIN), DINCast, DINPeer, and (G)MPLS-based DINloop_Net to address the above problems with multicast communications respectively VIN is a distribution model proposed to enable video distribution across space and time In the VIN system, video data is continuously circulated in a DINloop_Net VIN supports late-joining users with high dynamic multicast membership and still achieves data. .. Comparison of DIN, WDD & OWCache 12 Table 2.1 F values used in analysis 38 Table 2.2 Increasing rate with core network bandwidth 40 Table 2.3 Increasing rate with the outgoing link bandwidth 42 Table 2.4 Decreasing rate with (D+Ts) 43 Table 3.1 VIN versus Staggered Multicast 67 Table 3.2 Increasing rate of VIN versus Staggered Multicast 69 Table 4.1 Link stress comparison 76 Table 4.2 Loop information... switching/tuning to the end of switching/tuning tsl Signalling time from a DIN Node receiving a video request to the DIN Node deciding to switch/tune to a particular wavelength T Total delay time in VIN system TDmd_i Multicast delay for Node i in DINCast, it equals to the delay from the source to Node i via DINloop_App TDmd_i1 Delay from multicast source to a DIN Node in DINCast TDmd_i2 Delay in the DINloop_App... reduce the multicast delay and multiple DIN Nodes also avoid overloading and single-point of failure DINPeer extends DINCast by optimizing application-level multicast based on Peer-to-Peer (P2P) overlay network instead of a multicast tree and exploits a spiralring method to discover an inner ring with the relative largest bandwidth to form a DINloop_App DINPeer integrates the DINloop_App with a P2P... network to construct topologically-aware overlay multicast Finally, (G)MPLS-based DINloop_Net with techniques of MPLS/GMPLS allows implementing DINloop_Net and optimizing inter-domain multicast We implement the DINloop_Net in the core network by establishing Label Switched Paths through the DIN Nodes The DINloop_Net is used to exchange multicast group membership information and facilitate forwarding... (Digital TV) multicast CHAPTER 1 INTRODUCTION 2 services Among the 450 responding stations, 50% of stations are currently multicasting, and 79% among non-multicasting stations are considering multicasting at some point in the future The first was proposed by Steve Deering in 1988 [6] and described as the standard multicast model for IP network, i.e., IP multicast [7] With IP Multicast, a single packet . ENHANCING MULTICAST COMMUNICATIONS WITH DATA-IN-NETWORK SCHEMES GUO HUAQUN NATIONAL UNIVERSITY OF SINGAPORE 2005 ENHANCING MULTICAST COMMUNICATIONS WITH. in Multicast Communications 1 1.1.1 Lack of Multicast across Time 3 1.1.2 Limited Address Space & Large Delay in Application-level Multicast 5 1.1.3 Scalability of IP Inter-domain Multicast. Total multicast delay for all nodes to receive the multicast message in DINCast T Link Link latency T mA Delay time for a multicast message at a node with Function A T mB Delay time for a multicast