Locality-Aware Clustering Application Level Multicast for Live Streaming Services on the Internet

Size: px
Start display at page:

Download "Locality-Aware Clustering Application Level Multicast for Live Streaming Services on the Internet"

Transcription

1 JOURNAL OF INFORMATION SCIENCE AND ENGINEERING 27, (2011) Locality-Aware Clustering Application Level Multicast for Live Streaming Services on the Internet Department of Computer Engineering Chulalongkorn University Bangkok, Thailand With the rapid growth of Internet, media streaming plays an important role for the growing demand of media service. Application Level Multicast (ALM) has emerged as a key alternative to enable broadcasting the streaming media over the Internet. Several studies have proposed solutions to construct ALM multicast trees with various aspects such as end-to-end delay optimization, quick joining delay, small maintenance join/leave overhead, and balance multicast tree. However, these approaches focus on the optimization of a simple measurement, end-to-end delay, while ignoring other factors such as bandwidth consumptions, which can be very critical for large-scale ALM. In addition, the tier-based nature of the Internet further increases the difficulties of constructing scalable ALM multicast tree. This paper proposes a hierarchical structure approach called Locality-Aware Clustering (LAC), which utilizes the knowledge of network topology to construct an efficient ALM multicast tree. Based on simulation studies, LAC algorithm can reduce the overlay stress by more than 80% comparing to traditional algorithms such as ZIGZAG and MBMT/MSMT while maintaining low overlay delay even when there are large number of nodes. This simulation results show that LAC algorithm can construct ALM multicast trees that are more localized, efficient and scalable than the traditional approaches. Keywords: locality-aware clustering, application level multicast, landmark, bandwidth localization, bandwidth efficiency 1. INTRODUCTION The success of live multimedia services on the Internet, such as event broadcasting, Internet TV, and Internet radio, has increased the amount of multimedia traffics drastically. Many content providers adopt media streaming technology such that an end user can start watching the media as soon as the download starts. Nevertheless, these services rely on the network capability to satisfy traffic requirements such as bandwidth and jitter delay. This usually requires support from underlying network equipment. Application Level Multicast (ALM) has emerged as a key alternative to enable broadcasting the live streaming media on the Internet without any support from network equipment. ALM typically requires an efficient and scalable structure of unicast communication to support a large number of users. To be a good ALM structure for live streaming services, the structure has to meet two goals; (1) bandwidth efficiency, which deliver the streaming to the end-host with sufficient bandwidth that matches the requirements of the streaming application, (2) scalability, which minimize system resource usage to sustain the large number of users. To satisfy these goals, many ALM overlay structures, which are usually tree-based so called multicast tree, are proposed. Many works have Received December 22, 2008; revised September 9, 2009; accepted November 18, Communicated by Rong-Hong Jan. 319

2 320 been focusing on optimizing the overlay construction problem in many aspects, such as the delay [1, 2], total number of hops to forward the packet to all destinations [1], join/ leave overhead [1, 3], link stress [1, 4, 5], and tree bandwidth maximization [5]. Some algorithms [1, 6] consider group of node or cluster level, instead of the node level. Using the cluster level, the structure can reduce the join/leave overhead and consume less network resource in transferring data. Examples of cluster based approaches are NICE [6] and ZIGZAG [1]. ZIGZAG is very famous as it aims to construct overlay network based on latency optimization while limits the out-degree of each node to a constant bound. Thus, ZIGZAG can construct multicast trees that can be deployed on the Internet as the bandwidth availability is limited. Nevertheless, it cannot fully utilize the network resources as all links are not the same. Limiting out-degree of all nodes to a certain constant number is too pessimistic. To utilize the underlying network resources effectively, several topology-aware approaches have been proposed. In general, topology-aware approaches, such as MSMT/ MBMT [5], have to measure all physical links in the network. This can be very impractical for large-scale network such as the Internet as it is almost impossible to ping all nodes in the network and massive amount of information must be exchanged. In addition, the network environment is very dynamic. Thus, the measured information can become obsolete very quickly and re-measurements are needed. This paper proposes a Locality-Aware Clustering ALM (LAC) that can construct a multicast tree with good scalability, bandwidth efficiency, and practicality for live streaming services. To utilize the fact that the Internet spans several countries, links of the Internet are tier-based by nature as they can be classified into transit links and stub links. Hence, a good ALM approach focuses on the bandwidth localization, which is to minimize the bandwidth consumption on the transit links and utilize more bandwidths on the stub links. Moreover, this paper also proposes a simple geographical information called landmark to classify the links. By using landmarks, LAC can construct a multicast tree with maximum bandwidth localization and minimal latency measurements. The performance of the approach is evaluated using simulation. The simulation results show that LAC algorithm demonstrates better overlay utilization and offers better scalability than other traditional ALM approaches. The organization of this paper is as followed. Section 2 defines all necessary models, including physical model, logical model, ALM model and ALM performance metrics. Section 3 proposes Locality-Aware Clustering algorithm. This study evaluates and analyzes LAC algorithm by comparing with ZIGZAG, MSMT/MBMT through simulation in section 4. Section 5 discusses the related work and finally section 6 concludes the paper. 2.1 Physical Model 2. MODELS AND ASSUMPTIONS In general, a physical network is represented as a general graph G = (V, E), where V denotes a set of nodes and E denotes a set of physical links. A node in physical network, v i, can be either a host or a router. For a physical link e(v i, v j ) E, which is a link between v i and v j, there are two associated costs, link delay and link bandwidth represented

3 LOCALITY-AWARE ALM FOR INTERNET LIVE STREAMING 321 by d(e(v i, v j )) and bw(e(v i, v j )), respectively. Let the application bandwidth, denoted by abw(e(v i, v j )), to be the bandwidth that an application uses for transmitting data over a physical link e(v i, v j ), including all duplicated packets on that particular link. The utilization of a link, denoted by u(e(v i, v j )), is ratio of application bandwidth and the link bandwidth. Thus, abw(( e vi, v j)) uev (( i, vj)) =. bw(( e v, v )) (1) i j Typically, the effectiveness of a streaming application is usually limited by the bottleneck link, which is the link being utilized the most. The bottleneck link, denoted by BL, is the link with maximum utilization. In this case, ubl ( ) = max { uev ( (, v))}. (2) ev (, v) E i j i j n i In reality, Internet is the networks of networks. Thus, it exhibits the characteristics of the tier-based environment [7], such as domestic tier and international tier. Therefore, to represent the real network such as Internet, this paper will extend the physical network model with transit-stub architecture [8]. This architecture consists of two domain categories, Transit-Domain (TD) and Stub-Domain (SD). Transit-domain is consisting of international-level Transit-Nodes (TN) acting as connection points between transit-nodes and stub-domains. Thus, the transit-domain provides Internet backbone to connect between the two transit-nodes, as well as, between a transit-node to stub-domains via transit link or inter-area link. For stub-domain, all nodes in the same domain are assumed to be in nearby physical location or area. Thus, these nodes communicate other nodes in the same stub-domain via stub links. In this study, the data traffics within a stub-domain are called stub traffic and the traffics within the transit-domain or between a transit-node and stubdomain are called transit traffic. The extended model for a physical network G = (V, E) is formally defined as followed. Suppose the network has n stub-domains. Let a i be a stub-domain and V i be a set of nodes in a i. V 0 is defined as the set of transit nodes. The set of nodes V can be partitioned into n subsets, V = =1 Vi. In addition, a set of all links E composes of a set of transit links E t, which is the collection of links connecting transit-nodes with other transit nodes or stub-domains, and a set of stub links E s, which is the collection of links connecting nodes in the same stub-domains. Then, the set of link E is E t E s. Suppose e t is a transit link and e s is a stub link. This works make general assumptions that bw(e t ) < bw(e s ). Fig. 1 shows an example of Internet transit-stub architecture with 4 stub-domains and one transit-domain. Based on this model, nodes and links in physical topology will be classified into transit-domain and stub-domain. This process requires the measuring of node information such as delay and bandwidth using many popular tools such as traceroute, pathchar [9], Max-Delta [10]. However, these approaches usually take long time to measure the information, require frequent measurement, and consume a lot of resources. To simplify the identification process, this study utilizes the fact that nodes in the same domain should be in the nearby geographical area. Thus, area identification can be done easily such as asking user to provide the area information or inquiring from some standard Internet services

4 322 Fig. 1. Physical network. Fig. 2. Logical network. such as IPwhois [11]. These location identification approaches may not be 100% accurate. However, large-scale location information such as country code should have quite minimal error rates. In this study, the country code from IPwhois service will be used to identify the area of node, such as th, jp, and tw that represents Thailand, Japan, and Taiwan, respectively. The obtained country code will be converted to the code number, called landmark, that is the running number from 1 n country code. The landmark is used for 2 purposes. First, it is used for stub-domain identification. Let l(v i ) be a landmark of the node v i. For any node v i and v j, if l(v i ) is equal to l(v j ), both nodes are considered to be in the same stub-domain. Second, the landmark is used to identify the type of network link. For a link e(v i, v j ), if v i and v j have the same landmark and both nodes are not belong to the transit-domain, the link e(v i, v j ) is considered to be a stub link. Otherwise, it is a transit link. Fig. 1 presents a sample network consisting of 17 end-hosts (v s, v 1 v 16 ) and 8 routers (R 1 R 8 ). Let a 1 a 4 are the country codes of hosts obtained by using IPwhois service. The set of links E composes of stub links, e.g., e(v s, R 5 ), and transit links, e.g., e(r 1, R 2 ). 2.2 Logical Model In general, most applications rely on logical network model as they utilize point-topoint communication or end-to-end paths between nodes, while ignoring the physical components such as routers. Based on a physical model G = (V, E), a logical network model is a complete graph, G o = (V o, E o ) where nodes in the logical network (V o ) are all hosts (excluding routers) from G. For v i and v j V o, each logical link or end-to-end path p(v i, v j ) E o, E o is the set of physical links that constitutes the shortest path between v i and v j. Fig. 2 shows a logical network derived from the physical network in Fig. 1. The costs in logical network model include path delay and path bandwidth, which are defined as follows, pd( p( v, v )) = { d( e( v, v ))}, (3) i j i j ev ( i, vj) pv ( i, vj) pbw( p( v, v )) = min { bw( e( v, v ))}. i j ev (, v ) pv (, v ) i j i j i j (4)

5 LOCALITY-AWARE ALM FOR INTERNET LIVE STREAMING 323 Fig. 3. Packet flow of IP multicast. Fig. 4. Packet flow of ALM. 2.3 Application Level Multicast Model Streaming applications often require multicast services for delivering their streaming feeds. Typically, multicast services for streaming can be based on either IP multicasting or Application Level Multicasting (ALM). The main concept of ALM is to implement the multicast service at the application level. At this level, ALM utilizes pointto-point communication to create overlay network atop of logical network model. This enables ALM to become independent from physical network topology. Note that this study focuses on the ALM with single source. Thus, the ALM structure in this study is assumed to be based on tree. Figs. 3 and 4 show the differences between IP multicast and ALM. The arrow dash lines represent the packet flows on the transit link, while the arrow dot lines represent the packet flows on the stub link. Obviously, IP multicast mechanism relies on the routers to forward data packets from the sender to receivers without duplication. On the contrary, ALM utilizes unicasts to deliver packets. Thus, some unicasts may share the same physical link. This can lead to multiple sending of the same packet over this link, called link stress. The link stress of a physical link e(v i, v j ), denoted by st(e(v i, v j )), refers to the number of duplicated packets on that link. This paper defines the stress on transit link to be the maximum stress of all transit links on the multicast tree. The stress on stub stress is the maximum stress of all stub links, and the overlay stress is the maximum stress of all links. Given an ALM tree, T = (G o, v s, D, S). Let G o be a logical network model. Let v s be the source which is the root of the tree and let D be the set of nodes in G o which are the destinations of ALM traffic. Streaming data traffic from the source to all destinations may require multiple unicasts. For a destination v i D, a streaming path s(v s, v i ) constitutes the path between v s and v i. SP is the set of all streaming paths in the tree. Streaming path delay and streaming path bandwidth are defined as followed. The streaming path delay, denoted by sd(s(v s, v i )), is the sum of end-to-end delay of the logical link along the streaming path. The streaming path bandwidth, denoted by sbw(s(v s, v i )), is the minimum path bandwidth on the streaming path. Thus: sd(( s v, v )) = { pd(( p v, v ))}, (5) s i s i pv ( s, vi) sv ( s, vi) s i pv ( s, vi) sv ( s, vi) s i sbw( s( v, v )) = min { pbw( p( v, v ))}. (6)

6 324 Fig. 5. Overlay network. Fig. 6. The relationship of the three models. Fig. 5 presents the overlay network of the ALM tree in Fig. 4. To perform multicasting, v s establishes unicast connections to v 1, v 3, v 6, v 7, and v 16. After the nodes receive the data, they forward the data to other members, such as v 1 to v 2 and v 3 to v 4. The stress on transit link and stub link are 4 and 2, respectively. Finally, the overlay stress is 4. In the overlay network, the overlay delay, denoted by od(t), is the maximum streaming path delay comparing to all of them on the overlay network. The overlay bandwidth, denoted by obw(t), is the minimum streaming path bandwidth of the overlay network. Thus, od( T) = max { sd( s( v, v ))}, (7) sv ( s, vi) SP sv ( s, vi) SP s i obw( T) = max { sbw( s( v, v ))}. (8) s i Fig. 6 summarizes the relationships between the Physical model, the Logical model, and the ALM model. Some ALM algorithms base entirely on physical model, while some others focus only on logical model. With the introduction of landmark in the extended physical model, LAC algorithm can utilize both physical model and logical model to construct ALM trees with good performance. ALM performance matrices are discussed in the next section. 2.4 ALM Performance Metrics To evaluate the effectiveness of ALM on the Internet, this study uses the following metrics: link stress, overlay utilization, transit byte ratio, and overlay delay. The definitions of these metrics are as follows, 1. Link stress: This metric measures the quantity of the flooding packets. To utilize network bandwidth effectively, the stress must be minimized, especially at the link with small bandwidth like a transit link and highly-congestion link like a bottleneck link. The transit stress, denoted by st t, and the bottleneck link stress, denoted by st b, are defined as follows,

7 LOCALITY-AWARE ALM FOR INTERNET LIVE STREAMING 325 st = max { st( e( v, v ))}, (9) st t ev (, v) E i j b i j t = max { st( e( BL))}. (10) ebl ( ) E 2. Overlay utilization: As mentioned earlier, the bottleneck link is very critical to the streaming application. Thus, the utilization of the bottleneck link is used to represent the overlay utilization. Therefore, u(t) = u(bl). (11) 3. Transit byte ratio: This metric is a metric that shows the quality of localization-aware. It is used to quantify the quality of traffic localization by considering traffic in the stub links. The transit byte ratio, tbr, is the ratio of total bytes, tb, on the transit links by total bytes of all links. tbr( T ) = ev (, v) E i j t ev (, v) E i j { tb( e( v, v ))} { tb( e( v, v ))} i i j j (12) For good localization, the ALM should avoid using the transit links. Thus the transit byte ratio should be closer to 0. In addition, the stub byte ratio, denoted sbr(t), is defined as follows, sbr(t) = 1 tbr(t). (13) 4. Overlay delay: This metric is the maximum streaming path delay in the ALM tree. In this assumption, the overlay delay is not quite critical for the live streaming services. The delay just has to be within the pre-defined bound. Hence, the overlay delay is not the main metric in this evaluation. 3. LOCALITY-AWARE CLUSTERING (LAC) In this section, a new algorithm called Locality-Aware Clustering (LAC) is proposed to construct an efficient overlay tree for live streaming over the Internet. LAC algorithm uses landmark information to partition nodes and links into groups and constructs the overlay tree accordingly. 3.1 Locality-Aware Clustering Hierarchical Structure Traditionally, ALM overlay trees are constructed based on the optimization of the overlay delay. This may be significant in some ALM applications such as network game and video conference [12]. In some other applications such as live streaming, some delays within bound are usually acceptable and scalability becomes an important issue as it needs to support large number of audiences. To solve this problem, a new algorithm called

8 326 Locality-Aware Clustering (LAC) is proposed to construct an efficient overlay tree for live streaming. LAC structure is a tree of clusters, similar to the structure constructed by ZIGZAG and NICE, where the source is the root node. The size of the cluster is limited to be 3k, where k is a constant. This is to avoid the overhead of streaming data and protect the lost of members in merging cluster [6]. In each cluster, there are one Header Node (HN) and one Distributor Node (DN). Other nodes in the cluster are considered member nodes or Other Nodes (ON). For a cluster C i, the definitions of these nodes are as follows, 1. Header Node (HN i ): HN i is responsible for receiving streaming feed from the parent cluster and forwards the streaming feed to its child, which is the distributor node. This node is also a member of the parent cluster. For the top cluster, the source or the root of the tree acts as both header node and distributor node. 2. Distributor Node (DN i ): DN i is responsible for receiving streaming feed from the HN of this cluster and distributing data to other members of the cluster. DN also maintains the membership of the cluster. In addition, DN is the backup of HN. Thus, when HN fails, DN will become HN and a new DN is elected from one of the members. 3. Other node (ON): The rests of members are called the Other Nodes (ONs), which are the children of DN. ONs receive data from DN. If the cluster has child clusters, HNs of the child clusters are also ONs of this cluster. In addition, we define {ON} i to be a set of all ONs in C i. Fig. 7 shows an example of an LAC tree that generate from the physical network in Fig. 1. The tree has three overlapped clusters, C 0, C 1, and C 2. C 0 is the root cluster, which also includes the source and the root node (v s ). v s acts as both HN 0 and DN 0 of C 0 and v 8 is the only ON in this cluster. In C 1, there are three members (v s, v 1, and v 2 ). v s is an HN 1, v 1 is the DN 1, and v 2 is the only ON. In C 2, the HN 2 is v 8, the DN 2 is v 9, and v 10, v 12, v 16, v 3, v 5 are the members of {ON} 2. Based on the structure, the streaming feed comes from the source node v s to its only child in C 0, v 8. Upon receiving data, v 8 as HN 2 forwards the streaming feed to DN 2 (v 9 ), which forwards the feed to all {ON} 2. In addition, acting as HN 1, v s also forwards its data to v 1, which is DN 1. Although, LAC structure is based on ZIGZAG and NICE, LAC structure is more efficient because it combines the information from physical model and logical model to implement the traffic localization concept called Locality-Aware. This concept bases on the fact that the communication between two nodes within the same stub domain should be more efficient than those across two different domains. Thus, the concept focuses on constructing a cluster such that stub links (or local links) are preferred over tran- C0 HN0, DN0 Vs a1 HN1 DN1 V1 a1 ON0 Vs: Sender V8 a4 HN: Header Node HN2 DN: Distributor Node ON: Other Node DN2 V9 a4 a1- a4: Landmark ON1 ON2 V2 V10 V12 V16 V3 V5 a1 a4 a4 a4 a2 a2 C1 C2 Fig. 7. An example of LAC tree.

9 LOCALITY-AWARE ALM FOR INTERNET LIVE STREAMING 327 sit links as intra-cluster links. This minimizes the stresses in the transit links, which lead to better performance and greater scalability. Considering the LAC structure, DN has to communicate with many ONs. If DN and ON are on the same physical area, they can communicate directly via stub links. With this concept, LAC algorithm tries to group nodes with the same landmark into the same cluster. This will be done when a new node joins the tree. 3.2 Joining a New Node To join a new node, LAC joining algorithm, presented in Fig. 8, is used to find the target cluster. Based on the localization, LAC joining algorithm selects the target cluster (C x ) by keeping nodes in the same area within the same cluster. Suppose a new node v i want to join the structure. If the v i is the source node, it becomes the root of LAC multicast tree. Otherwise, v i will requests all DN list and their landmarks from the DN of the root cluster (DN 0 ) to construct a shortlist of the near-by clusters. A near-by cluster is a cluster that shares the same landmark with v i. If there are more than one cluster in the near-by cluster shortlist, v i will select the C x which is not full and has the smallest latency. Note that if all clusters in the shortlist are full, v i will select the C x based on the latency only. If the near-by cluster shortlist is empty, v i measures the latency from v i to all DNs in the multicast tree and joins the cluster with the smallest latency. The algorithm of LAC: Joining new node 1. If (source(v i )) 2. root = v i 3. else 4. if (#C == 0) 5. HN 0 = DN 0 = root 6. else DN x = FindC x (v i, l(v i )) 7. end if 8. JoinNode(DN x, v i, pd(dn x, v i )) 9. if (#Member(C x ) > 9) 10. SplittingAndMerging(C x ) 11. end if 12. end if Fig. 8. The algorithm of joining newcomer. When the target cluster is selected, v i joins as a child of the DN of the target cluster. LAC joining algorithm then evaluates the size of the cluster. If the cluster is too large (more than 3k), the cluster is partitioned by splitting and merging algorithm. Fig. 9 shows an example of joining nodes to the multicast tree. Suppose the multicast tree already has 3 nodes, as demonstrated in Fig. 9 (a), and there are 7 new nodes, v 9, v 10, v 8, v 1, v 2, v 5, and v 12, want to join the multicast tree. When the first 6 nodes join the multicast tree, all nodes will join the root cluster, C 0, as presented in Fig. 9 (b). When v 12 joins the multicast tree in Fig. 9 (c), the root cluster becomes overflow as its size is 10 (including the root node), which is more than 3k (k = 3). In this case, the cluster will be partitioned into 2 clusters by calling SplittingAndMerging in line 10.

10 328 (a) v s, v 16, v 3 joining. (b) Cluster with maximum members. (c) Cluster overflows. Fig. 9. An example of joining nodes in LAC tree construction. The algorithm: SplittingAndMerging 1. sortlandmark(l({on} x )) 2. sortnoofnodeinlandmark(#(l{on} x )) 3. sortlatency(pd({on} x ) ) 4. sp = position(finddifflandmark({on} x )) 5. In the right of sp 6. HN y = child(dn x, sp) 7. DN y = child(dn x, sp + 1) 8. For child (i = sp + 2 to n) 9: addchild(dny) = child(dnx, v i ) 10. end for 11. If (l(hny) == l(dnx)) 12. addchild(dnx) = HNy 13. else 14. addchild(parentofcx) = HNy 15. end if 16. end of right of sp 17. In the left of sp 18. If (#C > 1) 19. for child (i = 0 to sp 1) 20. clearchild(dnx, v i ) 21. end for 22. else 23. DNx = child(hnx, 0) 24. for child (i = 1 to sp 1) 25. addchild(dnx) = child(hnx, v i ) 26. end for 27. addchild(hnx) = DNx 28. end if 29. end of left of sp Fig. 10. The algorithm of SplittingAndMerging. When the cluster is too large, there are two stages of the SplittingAndMerging procedure. (1) Splitting the overflow cluster into two clusters, and (2) merging the new cluster to the multicast tree. Fig. 10 shows the detail of SplittingAndMerging algorithm. Suppose C x is too large. Splitting C x will perform in 3 steps, sorting the {ON} x, which is the set of all ONs of C x (lines 1-3), finding the splitting position to form a new cluster (called sp, line 4), and selecting the HN and DN for the new cluster (lines 6-7, 23). In the sorting stage, LAC first group {ON} x by their landmarks, then sort the group of landmark by the number of nodes in decreasing order, and finally re-sort these nodes in any landmark by latency in increasing order. Note that the group of ONs with the same landmark of DN x will stay on the left side of the list. Then the splitting position is assigned by choosing the first child, whose landmark is different from the others. If all children have the same landmark, the cluster is split in halves. Using the splitting position, HN x, DN x, and all ONs on the left side of the splitting position remain in C x. In this case, HN x and DN x remain unchanged. The ONs on the right side of the list will form a new cluster, C y. The first two nodes in the right side of the list will be chosen as new HN y and the DN y. To prevent the new cluster from being too small, the splitting position must be in between the third and the sixth child of the list. To complete the process, C y will be merged to the multicast tree using the merging function in lines In merging process, the landmark of HN y is compared to the

11 LOCALITY-AWARE ALM FOR INTERNET LIVE STREAMING 329 landmark of the DN x. If the both landmark are the same, the HN y will join C x and become the child of the DN x. Otherwise, the HN y will become a child of the parent cluster of C x. In this case, C x and C y become siblings. Once merged, if the parent cluster becomes overflow, the parent cluster will be split and merged in similar fashion. There is a special case when splitting the root cluster, C 0. Since the C 0 is the root cluster, it has no parent cluster. In this case, all clusters in the multicast tree will be rearranged by renaming all clusters and creating a new C 0. Suppose the C 0 is full. It will be split C 0 and form C 1 similar to normal splitting process. Note that the root node still remains as HN 0 and DN 0. After splitting, the tree must be re-arranged. To re-arrange the tree, the names of all clusters in the multicast tree will be increased by one. Thus, C 0 becomes C 1, etc. A new C 0 is created from the HN 1 (after being renamed), or the root node, and the HN 2 (after being renamed). For the new C 0, the root node will become the HN 0 and DN 0 and the HN 2 will become the only ON of C 0. Fig. 11 shows the example of splitting and merging cluster. Fig. 11 (a) shows C 0 after the {ON} 0 are sorted by their landmarks, number of nodes in each landmark, and latencies. Then the ONs will form a list in order to choose the splitting position. In C 0, the first position that the landmarks are different is the third position, so the splitting position of this cluster is the node v 8. Thus, the C 0 is split. HN 0 and DN 0 (v s ), as well as, all ONs from the left side of the splitting position {v 1, v 2 } remain in C 0. HN 0 and DN 0 remain unchanged. The remaining ONs ({v 8, v 9, v 10, v 12, v 16, v 3, v 5 }) form a new cluster C 1. The first (a) C 0 in splitting process. (b) LAC tree after merging C 2. (c) C 2 overflows. (d) C 2 in splitting process. (e) LAC tree after merging C 3. Fig. 11. An example of splitting and merging cluster.

12 330 Fig. 12. ALM with locality-aware. Fig. 13. Overlay network. two nodes in the list, v 8 and v 9, are chosen as the HN 1 and the DN 1, respectively. Since the C 0 is split, all clusters in the multicast tree must be re-arranged. C 0 is renamed to be C 1. C 1 is renamed to be C 2. A new C 0 is created from HN 1 (v s ) and HN 2 (v 8 ). The root node becomes HN of the new C 0. Similarly, Fig. 11 (c) shows LAC tree after v 7, v 4, and v 6 join the C 2 and then the cluster is overflow. C 2 is partitioned into C 2 ({v 8, v 9, v 10, v 12, v 16 }), C 3 ({v 3, v 5, v 4, v 7, v 6 }) as illustrated in Fig. 11 (d). The new cluster C 3 is merged to the multicast tree and HN 3 becomes a child of C 0, which is also the parent cluster of C 2 in Fig. 11 (e). Finally, Figs. 12 and 13 show an example multicast tree generated by LAC algorithm. Using localization strategy, LAC tree is constructed such that most streaming traffics have been transmitted over the stub links leading to fewer stresses in the transit links. In this example, the transit stress of this tree is 2, which is lower than the transit stresses of the tree generated by ZIGZAG in Fig Leaving Node There are two possible cases of node leaving the multicast tree, voluntary and involuntary (e.g. node crashes). When a node leaves voluntarily from a cluster in LAC tree, the leaving process is invoked. By considering the type of the leaving node, there are three cases: 1. If the leaving node is the HN of the cluster, the DN of that cluster is promoted to be a new HN. The ON node with the shortest latency from the new HN will become the new DN. 2. If the leaving node is the DN of the cluster, the ON node with the shortest latency from the existing HN will become the new DN. 3. If the leaving node is the ON, no action is necessary. After node leaves from the cluster, if the cluster size is not between k and 3k, all nodes in the cluster must join a cluster whose HN has the same landmark as the HN of the undersize cluster. After joining, the joined cluster may be split again if the cluster size is overflow. In the case that a node crashes, if the crashed node is an ON, this will not affect the leaving process. If the crashed node is either HN or DN, however, part of the multicast

13 LOCALITY-AWARE ALM FOR INTERNET LIVE STREAMING 331 tree is disconnected. To prevent this problem, a heart-beat mechanism must be implemented between HN and DN. Once a crash is detected, the leaving process is invoked, similarly to a node leaves voluntarily. 3.5 Practical Issues To implement LAC algorithm, node s memory requirement, clustering time, and packet loss must be considered. In LAC algorithm, cluster information during normal operation is distributed among clusters. In this case, HN and DN must maintain the information of their own cluster. Since cluster size is limited to 3k, the memory requirements of maintaining cluster information during normal operation is quite minimal. Memory requirements during clustering time, however, require special consideration. In general, node s memory requirement and clustering time are quite related, especially during the joining period. In LAC algorithm, the time consuming step of the joining process is to construct a shortlist of near-by clusters. This requires examining the landmark information of DN of each cluster starting from the root cluster and goes further down in the multicast tree. To shorten this construction process, high-level nodes such as the root node and HN of the high-level clusters may cache cluster information of their child clusters. Since only landmark information of each DN is needed to represent each cluster, caching child clusters information for shortlist construction is not memory-demanding. Packet loss is possible during the leaving process. For voluntarily leaving, the leaving node will not leave right away. It will wait until the leaving process is complete and then actually leave. For involuntarily leaving, the leaving node is crashed. Thus, it cannot wait until the leaving process is complete. In this case, other nodes in the cluster will be disconnected from the multicast tree and the streaming service is interrupted. To minimize the length of interruption period, the heart-beat mechanism between HN and DN must be frequent enough in order to detect the crash as soon as possible. The other practical issue is the accuracy of location identification. Using country code as a landmark may not always be accurate as the IPWhois service may return a country code that is not corresponding to the actual location where the node is located. The inaccurate landmark information can lead LAC to construct an inefficient multicast tree. However, since the landmark is based on a country code, which is very large-scale, it is very unlikely that this information is incorrect. In addition, LAC algorithm has been designed such that it can tolerate this problem because latency is also considered during the joining process. If the landmark is incorrect, the DN that is claimed to be in the same area of the joining node will have longer latency than other DNs in the shortlist whose landmark are correct. Thus, the joining node will choose to join the DN with shorter latency, which should have a correct landmark. At the worst, LAC will perform as good as other cluster-based algorithms such as NICE and ZIGZAG if the inaccuracy rate is too high. 4.1 Simulation Environment 4. PERFORMANCE EVALUATION To evaluate LAC algorithm, simulations using NS2 simulator are conducted. The physical network topologies in all simulations are randomly generated based on the tran-

14 332 sit-stub architecture GT-ITM [8]. Each network composes of a transit-domain with five transit-nodes. Each transit node connects to three stub-domains and each stub-domain has 199 nodes. The size of physical network is 1000 nodes. Furthermore, nodes are randomly chosen to form multicast trees. Table 1. Testing condition. Parameter Stub Tier Transit Tier Propagation Latency (msec) Link Bandwidth (Mbps) Message size (bytes) Streaming rate (Kbps) Table 1 lists the parameters and their values being used in this simulation. The bandwidth distribution and end-to-end latency are the measurement results of [13]. The Bandwidth distribution is randomly assigned between 80 to 100 Mbps for stub link and 8 to 10 Mbps for transit link. All cluster size is limited at 3k where k = 3. Each streaming is 5 sec. long, and the streaming rate is 250 Kbps, which is quite typical for both audio and video streaming applications. Average values of simulating ALM trees on three different physic-cal topologies are being used in all performance comparison. 4.2 Simulation Results As discussed in section 2.4, stress is considered the most important metric for evaluating ALM. To achieve good scalability on the Internet, stress in the transit links must be minimized. Fig. 14 shows the overlay stresses of all algorithms with different number of nodes. For ALM trees with 500 nodes, the overlay stress of ZIGZAG, MSMT/MBMT, and LAC, are 182, 52, and 9, respectively. The result shows that LAC performs quite well, especially when the number of nodes is large. LAC trees can out-perform ZIGZAG and MBMT/ MSMT by 95% and 82%, respectively. Fig. 15 compares the stresses in the transit links and in the stub links of ALM trees generated by LAC and MBMT algorithms. With localization strategy, LAC tries to utilize the stub links as much as possible. Thus, the stresses in transit links of LAC tree remain quite low. On the contrary, the stresses in the transit links of MBMT increase as ALM tree becomes larger. This is because MBMT first uses a latency-oriented spanning tree as a basis to construct an ALM tree. Then, MBMT uses link utilization as criteria to optimize the tree. With this approach, the ALM trees constructed by MBMT are basically latency-optimized, which may not be stress-optimized. To provide a clear explanation, let s consider a simple example. In an MBMT tree, suppose a node v i is a parent of 3 other nodes. And suppose these 3 nodes are in the same area and v i belongs to the different area. Since all 3 nodes are in the same area, the latencies between these nodes and v i should be almost the same. Thus, for a spanning tree, these 3 nodes will link to the same parent, v i. In this case, the connections between parent and all 3 children will be over the same transit link, which lead to more stresses in the transit link. On the contrary, LAC will group these 3 nodes into the same cluster and have only one node connecting to v i. In this case, the transit stress is minimized.

15 LOCALITY-AWARE ALM FOR INTERNET LIVE STREAMING 333 Fig. 14. Overlay stress. Fig. 15. Transit stress and stub stress. Fig. 16. Overlay delay. Fig. 17. Transit bytes ratio. Fig. 18. Overlay utilization on bottleneck link. Fig. 19. Stress on bottleneck link. Fig. 16 shows the overlay delay. The results indicate that the overlay delay of LAC tree is less than the overlay delay of other trees in almost all of network sizes. Since LAC relies on the stub links, which have much lower latency than transit links, the overlay delay of LAC scheme is minimized. Fig. 17 shows the transit byte ratio as number of nodes change. The results confirm the characteristics of LAC algorithm that it utilizes the stub links more than the transit links. Comparing to other algorithms, LAC has the lowest transit bytes ratio, 0.043, while ZIGZAG has the largest transit byte ratio, This means less data is transferred on the transit links of LAC tree. Thus, with the same number of nodes in the multicast tree, LAC is the most bandwidth efficient. Fig. 18 investigates the scalability of the ALM tree on the Internet by evaluating the overlay utilization on the bottleneck link. This results show the performance of ALM tree

16 334 when taking the bottleneck link of each tree into consideration. In this case, the lower is better. As number of nodes increases, the overlay utilization on the bottleneck link of all algorithms increases. It also quite clear that LAC can out-perform all algorithms as its bottleneck links is quite minimal. This is because LAC scheme tries to place a new node in the cluster that matches its landmark first, while other algorithms ignore this fact and utilize either the minimum spanning tree or the shortest path algorithms. Thus, for these algorithms, the new node may ends up on different physical area of its peers. Fig. 19 confirms the previous results. LAC focuses on minimizing stresses on the bottleneck links and thus it has less stresses and utilizes bandwidth very effectively. 5. RELATED WORK Various ALM overlay construction have been proposed for streaming services [1-6]. They can be classified into two categories, tree-based (ZIGZAG [1], STAG [3], MSMT/ MBMT [5], and NICE [6]), and mesh based (IOO [2] and SHM [4]). IOO [2] proposed an inter-overlay optimization scheme using mesh-based structure aiming for constructing efficient paths for different overlay, guaranteeing streaming service quality, and improving resource utilization. SHM [4] constructs the hierarchical structure that composes of Rendezvous Point (RP) and domain. The structure is a global share tree based on IP Multicast Island that does not require extra detection mechanism between end-host during the member join. Some structures [1, 3, 5, 6] construct multicast trees using the underlying network information such as delay, bandwidth, and location information. These algorithms are considered topology-aware. STAG [3] proposed a tree construction that uses transit-stub network topology on top of physical network. This mechanism minimizes the overhead of the joining process, has low relative delay penalty, and reduce link stress. As this algorithm focuses mainly on latency, it is not bandwidth efficient. NICE [6] and ZIGZAG [1] construct a hierarchical cluster of nodes with each cluster having a head represent in the higher layer. In these algorithms, nodes are being grouped into a cluster to reduce the control overhead and allow fast joining process. NICE was designed to support network architecture with limited bandwidth. Its cluster size is bounded by a constant to avoid the bottleneck at the distributor. ZIGZAG is an extension of NICE with two additional techniques, cluster size balancing and capacity-based switching. It aims to minimize the overheads during joining and leaving processes, as well as, reducing the number of network hops. However, ZIGZAG assumes all nodes are similar in term of latency and bandwidth availability. Unfortunately, this is not applicable for the Internet. MSMT/MBMT [5] optimizes ALM tree based on the underlying network topology. Using complete information of network topology, the algorithm focuses on minimizing stress in any link and generates an ALM tree with high tree bandwidth and low link stress with low penalty in end-to-end delay. However, obtaining information required by MSMT/ MBMT, complete network topology, is not quite practical. 6. CONCLUSION This paper proposes an overlay construction for live streaming application on the

17 LOCALITY-AWARE ALM FOR INTERNET LIVE STREAMING 335 Internet, called Locality-Aware Clustering or LAC. Based on the assumption of tier-based networking structure, LAC combines the physical model and logical model to construct an efficient ALM structure for the Internet using landmark and end-to-end delay. This paper evaluates the performance of LAC scheme by comparing with ZIGZAG, MSMT/ MBMT using simulation. The experimental results clearly show that LAC tree can outperform all other algorithms by more than 80% when comparing overlay stress and by more than 30% when comparing the utilization on bottleneck link. In addition, the overlay delay of LAC tree is minimal. The simulation results also indicate that the effectiveness of LAC mechanism is due to its localization strategy as LAC relies on the stub tier, which can greatly reduce the bandwidth requirements on the transit tier, whose bandwidth is usually quite limited. Therefore, LAC algorithm can generate a bandwidth-efficient ALM structure with low latency and good scalability for live streaming. The future works will extend LAC algorithm to support multiple source services. REFERENCES 1. D. A. Tran, K. A. Hua, and T. T. Do, ZIGZAG: An efficient peer-to-peer scheme for media streaming, in Proceedings of International Conference on Computer Communications, 2003, pp X. Liao, H. Jin, Y. Liu, and L. M. Ni, Scalable live streaming service based on inter-overlay optimization, IEEE Transactions on Parallel and Distributed Systems, 2007, pp J. Cui, Y. He, and L. Wu, More efficient mechanism of topology-aware overlay construction in application-layer multicast, in Proceedings of International Conference on Networking, Architecture, and Storage, 2007, pp S. Lu, J. Wang, G. Yang, and C. Guo, SHM: Scalable and backbone topology-aware hybrid multicast, in Proceedings of IEEE Conference on Computer Communications and Networks, 2007, pp X. Jin, W. P. K. Yiu, S. H. G. Chan, and Y. Wang, On maximizing tree bandwidth for topology-aware peer-to-peer streaming, IEEE Transactions on Multimedia, Vol. 9, 2007, pp S. Banerjee, B. Bhattacharjee, and C. Kommareddy, Scalable application layer multicast, in Proceedings of ACM SIGCOMM, 2002, pp H. Haddadi, M. Rio, G. Iannaccone, A. Moore, and R. Mortier, Network topologies: inference, modeling, and generation, IEEE Communications Surveys and Tutorials, Vol. 10, 2008, pp E. W. Zegura, K. Calvert, and S. Bhattacharjee, How to model an Internetwork, in Proceedings of IEEE Conference on Computer Communication, 1996, pp V. Jacobson, Pathchar, X. Jin, W. P. K. Yiu, S. H. G. Chan, and Y. Wang, Network topology inference based on end-to-end measurements, IEEE Journal on Selected Areas in Communications, Vol. 24, 2006, pp The APNIC Whois Database, S. Hajime, S. Hiroki, and T. Yoshito, P2P grouping for the delay sensitive applications, IEIC Technical Report, Vol. 106, 2006, pp

18 K. Silawarawet and N. Nupairoj, Minimizing multicast communication latency in WAN-based environment, in Proceedings of International Conference on Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed, 2006, pp Kanchana Silawarawet received the B.E. in Computer Engineering from Siam University in 1996, and the M.S. degree in Computer Science from Chulalongkorn University in She has been working as a Lecturer at Department of Computer Engineering, Siam University for 5 years and currently pursuing the Ph.D. degree in Computer Engineering at Chulalongkorn University. Her research interests include multicast, parallel and distributed computing, and grid computing. Natawut Nupairoj received the B.E. in Computer Engineering from Chulalongkorn University in 1990, M.S. and Ph.D. in Computer Science from Michigan State University in 1993 and 1998, respectively. Since 1998, he has been a Lecturer at Department of Computer Engineering, Chulalongkorn University. His research interests include grid computing, distributed system, and service oriented architecture.

Traceroute-Based Topology Inference without Network Coordinate Estimation

Traceroute-Based Topology Inference without Network Coordinate Estimation Traceroute-Based Topology Inference without Network Coordinate Estimation Xing Jin, Wanqing Tu Department of Computer Science and Engineering The Hong Kong University of Science and Technology Clear Water

More information

International Journal of Advanced Research in Computer Science and Software Engineering

International Journal of Advanced Research in Computer Science and Software Engineering Volume 2, Issue 9, September 2012 ISSN: 2277 128X International Journal of Advanced Research in Computer Science and Software Engineering Research Paper Available online at: www.ijarcsse.com An Experimental

More information

Peer-to-Peer Networks. Chapter 6: P2P Content Distribution

Peer-to-Peer Networks. Chapter 6: P2P Content Distribution Peer-to-Peer Networks Chapter 6: P2P Content Distribution Chapter Outline Content distribution overview Why P2P content distribution? Network coding Peer-to-peer multicast Kangasharju: Peer-to-Peer Networks

More information

A Topology-Aware Relay Lookup Scheme for P2P VoIP System

A Topology-Aware Relay Lookup Scheme for P2P VoIP System Int. J. Communications, Network and System Sciences, 2010, 3, 119-125 doi:10.4236/ijcns.2010.32018 Published Online February 2010 (http://www.scirp.org/journal/ijcns/). A Topology-Aware Relay Lookup Scheme

More information

Network Topology and Traceroutes

Network Topology and Traceroutes A Distributed Approach to End-to-End Network Topology Inference Xing Jin Qiuyan Xia S.-H. Gary Chan Department of Computer Science and Engineering The Hong Kong University of Science and Technology Clear

More information

Fault-Tolerant Framework for Load Balancing System

Fault-Tolerant Framework for Load Balancing System Fault-Tolerant Framework for Load Balancing System Y. K. LIU, L.M. CHENG, L.L.CHENG Department of Electronic Engineering City University of Hong Kong Tat Chee Avenue, Kowloon, Hong Kong SAR HONG KONG Abstract:

More information

AN OVERVIEW OF QUALITY OF SERVICE COMPUTER NETWORK

AN OVERVIEW OF QUALITY OF SERVICE COMPUTER NETWORK Abstract AN OVERVIEW OF QUALITY OF SERVICE COMPUTER NETWORK Mrs. Amandeep Kaur, Assistant Professor, Department of Computer Application, Apeejay Institute of Management, Ramamandi, Jalandhar-144001, Punjab,

More information

The Feasibility of Supporting Large-Scale Live Streaming Applications with Dynamic Application End-Points

The Feasibility of Supporting Large-Scale Live Streaming Applications with Dynamic Application End-Points The Feasibility of Supporting Large-Scale Live Streaming Applications with Dynamic Application End-Points Kay Sripanidkulchai, Aditya Ganjam, Bruce Maggs, and Hui Zhang Instructor: Fabian Bustamante Presented

More information

Design and Deployment of Locality-aware Overlay Multicast Protocol for Live Streaming Services

Design and Deployment of Locality-aware Overlay Multicast Protocol for Live Streaming Services Design and Deployment of Locality-aware Overlay Multicast Protocol for Live Streaming Services Xuping Tu, Hai Jin, Dafu Deng, Chao Zhang, and Quan Yuan Cluster and Grid Computing Lab Huazhong University

More information

Stability of QOS. Avinash Varadarajan, Subhransu Maji {avinash,smaji}@cs.berkeley.edu

Stability of QOS. Avinash Varadarajan, Subhransu Maji {avinash,smaji}@cs.berkeley.edu Stability of QOS Avinash Varadarajan, Subhransu Maji {avinash,smaji}@cs.berkeley.edu Abstract Given a choice between two services, rest of the things being equal, it is natural to prefer the one with more

More information

CROSS LAYER BASED MULTIPATH ROUTING FOR LOAD BALANCING

CROSS LAYER BASED MULTIPATH ROUTING FOR LOAD BALANCING CHAPTER 6 CROSS LAYER BASED MULTIPATH ROUTING FOR LOAD BALANCING 6.1 INTRODUCTION The technical challenges in WMNs are load balancing, optimal routing, fairness, network auto-configuration and mobility

More information

A NETWORK CONSTRUCTION METHOD FOR A SCALABLE P2P VIDEO CONFERENCING SYSTEM

A NETWORK CONSTRUCTION METHOD FOR A SCALABLE P2P VIDEO CONFERENCING SYSTEM A NETWORK CONSTRUCTION METHOD FOR A SCALABLE P2P VIDEO CONFERENCING SYSTEM Hideto Horiuchi, Naoki Wakamiya and Masayuki Murata Graduate School of Information Science and Technology, Osaka University 1

More information

MOVING P2P LIVE STREAMING TO MOBILE AND UBIQUITOUS ENVIRONMENT. Xiaofei Liao, Hai Jin, Wenbin Jiang

MOVING P2P LIVE STREAMING TO MOBILE AND UBIQUITOUS ENVIRONMENT. Xiaofei Liao, Hai Jin, Wenbin Jiang Computing and Informatics, Vol. 28, 2007, 823 835 MOVING P2P LIVE STREAMING TO MOBILE AND UBIQUITOUS ENVIRONMENT Xiaofei Liao, Hai Jin, Wenbin Jiang Cluster and Grid Computing Lab. Services Computing Technology

More information

Efficient Data Retrieving in Distributed Datastreaming

Efficient Data Retrieving in Distributed Datastreaming Efficient Data Retrieving in Distributed Datastreaming Environments Yunhao Liu, Jun Miao, Lionel M. Ni, and Jinsong Han Department of Computer Science Hong Kong University of Science and Technology Clear

More information

HPAM: Hybrid Protocol for Application Level Multicast. Yeo Chai Kiat

HPAM: Hybrid Protocol for Application Level Multicast. Yeo Chai Kiat HPAM: Hybrid Protocol for Application Level Multicast Yeo Chai Kiat Scope 1. Introduction 2. Hybrid Protocol for Application Level Multicast (HPAM) 3. Features of HPAM 4. Conclusion 1. Introduction Video

More information

Influence of Load Balancing on Quality of Real Time Data Transmission*

Influence of Load Balancing on Quality of Real Time Data Transmission* SERBIAN JOURNAL OF ELECTRICAL ENGINEERING Vol. 6, No. 3, December 2009, 515-524 UDK: 004.738.2 Influence of Load Balancing on Quality of Real Time Data Transmission* Nataša Maksić 1,a, Petar Knežević 2,

More information

Chapter 4. VoIP Metric based Traffic Engineering to Support the Service Quality over the Internet (Inter-domain IP network)

Chapter 4. VoIP Metric based Traffic Engineering to Support the Service Quality over the Internet (Inter-domain IP network) Chapter 4 VoIP Metric based Traffic Engineering to Support the Service Quality over the Internet (Inter-domain IP network) 4.1 Introduction Traffic Engineering can be defined as a task of mapping traffic

More information

Scalable Application Layer Multicast

Scalable Application Layer Multicast Scalable Application Layer Multicast Suman Banerjee, Bobby Bhattacharjee, Christopher Kommareddy Department of Computer Science,University of Maryland, College Park, MD 2742, USA suman,bobby,kcr @cs.umd.edu

More information

Comparative Analysis of Congestion Control Algorithms Using ns-2

Comparative Analysis of Congestion Control Algorithms Using ns-2 www.ijcsi.org 89 Comparative Analysis of Congestion Control Algorithms Using ns-2 Sanjeev Patel 1, P. K. Gupta 2, Arjun Garg 3, Prateek Mehrotra 4 and Manish Chhabra 5 1 Deptt. of Computer Sc. & Engg,

More information

Study of Different Types of Attacks on Multicast in Mobile Ad Hoc Networks

Study of Different Types of Attacks on Multicast in Mobile Ad Hoc Networks Study of Different Types of Attacks on Multicast in Mobile Ad Hoc Networks Hoang Lan Nguyen and Uyen Trang Nguyen Department of Computer Science and Engineering, York University 47 Keele Street, Toronto,

More information

Performance Evaluation of AODV, OLSR Routing Protocol in VOIP Over Ad Hoc

Performance Evaluation of AODV, OLSR Routing Protocol in VOIP Over Ad Hoc (International Journal of Computer Science & Management Studies) Vol. 17, Issue 01 Performance Evaluation of AODV, OLSR Routing Protocol in VOIP Over Ad Hoc Dr. Khalid Hamid Bilal Khartoum, Sudan dr.khalidbilal@hotmail.com

More information

Computer Network. Interconnected collection of autonomous computers that are able to exchange information

Computer Network. Interconnected collection of autonomous computers that are able to exchange information Introduction Computer Network. Interconnected collection of autonomous computers that are able to exchange information No master/slave relationship between the computers in the network Data Communications.

More information

WITH THE RAPID growth of the Internet, overlay networks

WITH THE RAPID growth of the Internet, overlay networks 2182 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 24, NO. 12, DECEMBER 2006 Network Topology Inference Based on End-to-End Measurements Xing Jin, Student Member, IEEE, W.-P. Ken Yiu, Student

More information

EECS 489 Winter 2010 Midterm Exam

EECS 489 Winter 2010 Midterm Exam EECS 489 Winter 2010 Midterm Exam Name: This is an open-book, open-resources exam. Explain or show your work for each question. Your grade will be severely deducted if you don t show your work, even if

More information

Quality of Service Routing Network and Performance Evaluation*

Quality of Service Routing Network and Performance Evaluation* Quality of Service Routing Network and Performance Evaluation* Shen Lin, Cui Yong, Xu Ming-wei, and Xu Ke Department of Computer Science, Tsinghua University, Beijing, P.R.China, 100084 {shenlin, cy, xmw,

More information

Introduction to LAN/WAN. Network Layer

Introduction to LAN/WAN. Network Layer Introduction to LAN/WAN Network Layer Topics Introduction (5-5.1) Routing (5.2) (The core) Internetworking (5.5) Congestion Control (5.3) Network Layer Design Isues Store-and-Forward Packet Switching Services

More information

How To Provide Qos Based Routing In The Internet

How To Provide Qos Based Routing In The Internet CHAPTER 2 QoS ROUTING AND ITS ROLE IN QOS PARADIGM 22 QoS ROUTING AND ITS ROLE IN QOS PARADIGM 2.1 INTRODUCTION As the main emphasis of the present research work is on achieving QoS in routing, hence this

More information

Two Dimensional Array Based Overlay Network for Balancing Load of Peer-to-Peer Live Video Streaming

Two Dimensional Array Based Overlay Network for Balancing Load of Peer-to-Peer Live Video Streaming Two Dimensional Array Based Overlay Network for Balancing Load of Peer-to-Peer Live Video Streaming Abdullah Faruq Ibn Ibrahimy, Islam Md. Rafiqul, Farhat Anwar and Muhammad Ibn Ibrahimy Department of

More information

QUALITY OF SERVICE METRICS FOR DATA TRANSMISSION IN MESH TOPOLOGIES

QUALITY OF SERVICE METRICS FOR DATA TRANSMISSION IN MESH TOPOLOGIES QUALITY OF SERVICE METRICS FOR DATA TRANSMISSION IN MESH TOPOLOGIES SWATHI NANDURI * ZAHOOR-UL-HUQ * Master of Technology, Associate Professor, G. Pulla Reddy Engineering College, G. Pulla Reddy Engineering

More information

Consecutive Geographic Multicasting Protocol in Large-Scale Wireless Sensor Networks

Consecutive Geographic Multicasting Protocol in Large-Scale Wireless Sensor Networks 21st Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications Consecutive Geographic Multicasting Protocol in Large-Scale Wireless Sensor Networks Jeongcheol Lee, Euisin

More information

Definition. A Historical Example

Definition. A Historical Example Overlay Networks This lecture contains slides created by Ion Stoica (UC Berkeley). Slides used with permission from author. All rights remain with author. Definition Network defines addressing, routing,

More information

other. A B AP wired network

other. A B AP wired network 1 Routing and Channel Assignment in Multi-Channel Multi-Hop Wireless Networks with Single-NIC Devices Jungmin So + Nitin H. Vaidya Department of Computer Science +, Department of Electrical and Computer

More information

AN INITIAL PEER CONFIGURATION ALGORITHM

AN INITIAL PEER CONFIGURATION ALGORITHM AN INITIAL PEER CONFIGURATION ALGORITHM FOR MULTI-STREAMING PEER-TO-PEER NETWORKS Tomoyuki Ishii and Atsushi Inoie * Department of Network Engineering, Kanagawa Institute of Technology, Atsugi-city, Japan

More information

Adapting Distributed Hash Tables for Mobile Ad Hoc Networks

Adapting Distributed Hash Tables for Mobile Ad Hoc Networks University of Tübingen Chair for Computer Networks and Internet Adapting Distributed Hash Tables for Mobile Ad Hoc Networks Tobias Heer, Stefan Götz, Simon Rieche, Klaus Wehrle Protocol Engineering and

More information

A Novel Load Balancing Optimization Algorithm Based on Peer-to-Peer

A Novel Load Balancing Optimization Algorithm Based on Peer-to-Peer A Novel Load Balancing Optimization Algorithm Based on Peer-to-Peer Technology in Streaming Media College of Computer Science, South-Central University for Nationalities, Wuhan 430074, China shuwanneng@yahoo.com.cn

More information

Architectural Framework for Large- Scale Multicast in Mobile Ad Hoc Networks

Architectural Framework for Large- Scale Multicast in Mobile Ad Hoc Networks Architectural Framework for Large- Scale Multicast in Mobile Ad Hoc Networks Ahmed Helmy Electrical Engineering Department University of Southern California (USC) helmy@usc.edu http://ceng.usc.edu/~helmy

More information

An Efficient QoS Routing Protocol for Mobile Ad-Hoc Networks *

An Efficient QoS Routing Protocol for Mobile Ad-Hoc Networks * An Efficient QoS Routing Protocol for Mobile Ad-Hoc Networks * Inwhee Joe College of Information and Communications Hanyang University Seoul, Korea iwj oeshanyang.ac.kr Abstract. To satisfy the user requirements

More information

A Comparative Study of Tree-based and Mesh-based Overlay P2P Media Streaming

A Comparative Study of Tree-based and Mesh-based Overlay P2P Media Streaming A Comparative Study of Tree-based and Mesh-based Overlay P2P Media Streaming Chin Yong Goh 1,Hui Shyong Yeo 1, Hyotaek Lim 1 1 Dongseo University Busan, 617-716, South Korea cgnicky@gmail.com, hui_shyong@hotmail.com,

More information

OSPF Version 2 (RFC 2328) Describes Autonomous Systems (AS) topology. Propagated by flooding: Link State Advertisements (LSAs).

OSPF Version 2 (RFC 2328) Describes Autonomous Systems (AS) topology. Propagated by flooding: Link State Advertisements (LSAs). OSPF Version 2 (RFC 2328) Interior gateway protocol (IGP). Routers maintain link-state database. Describes Autonomous Systems (AS) topology. Propagated by flooding: Link State Advertisements (LSAs). Router

More information

Content Delivery Networks. Shaxun Chen April 21, 2009

Content Delivery Networks. Shaxun Chen April 21, 2009 Content Delivery Networks Shaxun Chen April 21, 2009 Outline Introduction to CDN An Industry Example: Akamai A Research Example: CDN over Mobile Networks Conclusion Outline Introduction to CDN An Industry

More information

A NOVEL RESOURCE EFFICIENT DMMS APPROACH

A NOVEL RESOURCE EFFICIENT DMMS APPROACH A NOVEL RESOURCE EFFICIENT DMMS APPROACH FOR NETWORK MONITORING AND CONTROLLING FUNCTIONS Golam R. Khan 1, Sharmistha Khan 2, Dhadesugoor R. Vaman 3, and Suxia Cui 4 Department of Electrical and Computer

More information

A Well-organized Dynamic Bandwidth Allocation Algorithm for MANET

A Well-organized Dynamic Bandwidth Allocation Algorithm for MANET A Well-organized Dynamic Bandwidth Allocation Algorithm for MANET S.Suganya Sr.Lecturer, Dept. of Computer Applications, TamilNadu College of Engineering, Coimbatore, India Dr.S.Palaniammal Prof.& Head,

More information

Faculty of Engineering Computer Engineering Department Islamic University of Gaza 2012. Network Chapter# 19 INTERNETWORK OPERATION

Faculty of Engineering Computer Engineering Department Islamic University of Gaza 2012. Network Chapter# 19 INTERNETWORK OPERATION Faculty of Engineering Computer Engineering Department Islamic University of Gaza 2012 Network Chapter# 19 INTERNETWORK OPERATION Review Questions ٢ Network Chapter# 19 INTERNETWORK OPERATION 19.1 List

More information

Computer Networks - CS132/EECS148 - Spring 2013 ------------------------------------------------------------------------------

Computer Networks - CS132/EECS148 - Spring 2013 ------------------------------------------------------------------------------ Computer Networks - CS132/EECS148 - Spring 2013 Instructor: Karim El Defrawy Assignment 2 Deadline : April 25 th 9:30pm (hard and soft copies required) ------------------------------------------------------------------------------

More information

Source vs Data-driven Approach for Live P2P Streaming

Source vs Data-driven Approach for Live P2P Streaming Source vs Data-driven Approach for Live P2P Streaming Thomas Silverston and Olivier Fourmaux Université Pierre et Marie Curie-Paris6 Laboratoire d Informatique de Paris 6 8 rue du Capitaine Scott 7515

More information

Parallel overlays for high data-rate multicast data transfer

Parallel overlays for high data-rate multicast data transfer Computer Networks 51 (27) 31 42 www.elsevier.com/locate/comnet Parallel overlays for high data-rate multicast data transfer K.K. To, Jack Y.B. Lee * Department of Information Engineering, The Chinese University

More information

An Adaptive Approach to Optimize Tree-Based Overlay for P2P Streaming

An Adaptive Approach to Optimize Tree-Based Overlay for P2P Streaming An Adaptive Approach to Optimize Tree-Based Overlay for P2P Streaming Ming-Hung Hsu Tai-Hua Hsiao and Yu-Ben Miao tree-based overlay content data are pushed from an upstream node to a downstream node with

More information

An Efficient Load Balancing Technology in CDN

An Efficient Load Balancing Technology in CDN Issue 2, Volume 1, 2007 92 An Efficient Load Balancing Technology in CDN YUN BAI 1, BO JIA 2, JIXIANG ZHANG 3, QIANGGUO PU 1, NIKOS MASTORAKIS 4 1 College of Information and Electronic Engineering, University

More information

Figure 1. The Example of ZigBee AODV Algorithm

Figure 1. The Example of ZigBee AODV Algorithm TELKOMNIKA Indonesian Journal of Electrical Engineering Vol.12, No.2, February 2014, pp. 1528 ~ 1535 DOI: http://dx.doi.org/10.11591/telkomnika.v12i2.3576 1528 Improving ZigBee AODV Mesh Routing Algorithm

More information

Disjoint Path Algorithm for Load Balancing in MPLS network

Disjoint Path Algorithm for Load Balancing in MPLS network International Journal of Innovation and Scientific Research ISSN 2351-8014 Vol. 13 No. 1 Jan. 2015, pp. 193-199 2015 Innovative Space of Scientific Research Journals http://www.ijisr.issr-journals.org/

More information

Intelligent Content Delivery Network (CDN) The New Generation of High-Quality Network

Intelligent Content Delivery Network (CDN) The New Generation of High-Quality Network White paper Intelligent Content Delivery Network (CDN) The New Generation of High-Quality Network July 2001 Executive Summary Rich media content like audio and video streaming over the Internet is becoming

More information

First Midterm for ECE374 02/25/15 Solution!!

First Midterm for ECE374 02/25/15 Solution!! 1 First Midterm for ECE374 02/25/15 Solution!! Instructions: Put your name and student number on each sheet of paper! The exam is closed book. You have 90 minutes to complete the exam. Be a smart exam

More information

A Passive Method for Estimating End-to-End TCP Packet Loss

A Passive Method for Estimating End-to-End TCP Packet Loss A Passive Method for Estimating End-to-End TCP Packet Loss Peter Benko and Andras Veres Traffic Analysis and Network Performance Laboratory, Ericsson Research, Budapest, Hungary {Peter.Benko, Andras.Veres}@eth.ericsson.se

More information

A Power Efficient QoS Provisioning Architecture for Wireless Ad Hoc Networks

A Power Efficient QoS Provisioning Architecture for Wireless Ad Hoc Networks A Power Efficient QoS Provisioning Architecture for Wireless Ad Hoc Networks Didem Gozupek 1,Symeon Papavassiliou 2, Nirwan Ansari 1, and Jie Yang 1 1 Department of Electrical and Computer Engineering

More information

Improving the Performance of TCP Using Window Adjustment Procedure and Bandwidth Estimation

Improving the Performance of TCP Using Window Adjustment Procedure and Bandwidth Estimation Improving the Performance of TCP Using Window Adjustment Procedure and Bandwidth Estimation R.Navaneethakrishnan Assistant Professor (SG) Bharathiyar College of Engineering and Technology, Karaikal, India.

More information

Datagram-based network layer: forwarding; routing. Additional function of VCbased network layer: call setup.

Datagram-based network layer: forwarding; routing. Additional function of VCbased network layer: call setup. CEN 007C Computer Networks Fundamentals Instructor: Prof. A. Helmy Homework : Network Layer Assigned: Nov. 28 th, 2011. Due Date: Dec 8 th, 2011 (to the TA) 1. ( points) What are the 2 most important network-layer

More information

Computer Networks Homework 1

Computer Networks Homework 1 Computer Networks Homework 1 Reference Solution 1. (15%) Suppose users share a 1 Mbps link. Also suppose each user requires 100 kbps when transmitting, but each user transmits only 10 percent of the time.

More information

A Utility-Aware Middleware Architecture for Decentralized Group Communication Applications

A Utility-Aware Middleware Architecture for Decentralized Group Communication Applications A Utility-Aware Middleware Architecture for Decentralized Group Communication Applications Jianjun Zhang 1, Ling Liu 1, Lakshmish Ramaswamy 2, Gong Zhang 1, and Calton Pu 1 1 College of Computing, Georgia

More information

Route Discovery Protocols

Route Discovery Protocols Route Discovery Protocols Columbus, OH 43210 Jain@cse.ohio-State.Edu http://www.cse.ohio-state.edu/~jain/ 1 Overview Building Routing Tables Routing Information Protocol Version 1 (RIP V1) RIP V2 OSPF

More information

Implementation of a Lightweight Service Advertisement and Discovery Protocol for Mobile Ad hoc Networks

Implementation of a Lightweight Service Advertisement and Discovery Protocol for Mobile Ad hoc Networks Implementation of a Lightweight Advertisement and Discovery Protocol for Mobile Ad hoc Networks Wenbin Ma * Department of Electrical and Computer Engineering 19 Memorial Drive West, Lehigh University Bethlehem,

More information

An Efficient Hybrid P2P MMOG Cloud Architecture for Dynamic Load Management. Ginhung Wang, Kuochen Wang

An Efficient Hybrid P2P MMOG Cloud Architecture for Dynamic Load Management. Ginhung Wang, Kuochen Wang 1 An Efficient Hybrid MMOG Cloud Architecture for Dynamic Load Management Ginhung Wang, Kuochen Wang Abstract- In recent years, massively multiplayer online games (MMOGs) become more and more popular.

More information

Proxy-Assisted Periodic Broadcast for Video Streaming with Multiple Servers

Proxy-Assisted Periodic Broadcast for Video Streaming with Multiple Servers 1 Proxy-Assisted Periodic Broadcast for Video Streaming with Multiple Servers Ewa Kusmierek and David H.C. Du Digital Technology Center and Department of Computer Science and Engineering University of

More information

A Fast Path Recovery Mechanism for MPLS Networks

A Fast Path Recovery Mechanism for MPLS Networks A Fast Path Recovery Mechanism for MPLS Networks Jenhui Chen, Chung-Ching Chiou, and Shih-Lin Wu Department of Computer Science and Information Engineering Chang Gung University, Taoyuan, Taiwan, R.O.C.

More information

IMPROVED PROXIMITY AWARE LOAD BALANCING FOR HETEROGENEOUS NODES

IMPROVED PROXIMITY AWARE LOAD BALANCING FOR HETEROGENEOUS NODES www.ijecs.in International Journal Of Engineering And Computer Science ISSN:2319-7242 Volume 2 Issue 6 June, 2013 Page No. 1914-1919 IMPROVED PROXIMITY AWARE LOAD BALANCING FOR HETEROGENEOUS NODES Ms.

More information

Final for ECE374 05/06/13 Solution!!

Final for ECE374 05/06/13 Solution!! 1 Final for ECE374 05/06/13 Solution!! Instructions: Put your name and student number on each sheet of paper! The exam is closed book. You have 90 minutes to complete the exam. Be a smart exam taker -

More information

Communications and Computer Networks

Communications and Computer Networks SFWR 4C03: Computer Networks and Computer Security January 5-8 2004 Lecturer: Kartik Krishnan Lectures 1-3 Communications and Computer Networks The fundamental purpose of a communication system is the

More information

DAG based In-Network Aggregation for Sensor Network Monitoring

DAG based In-Network Aggregation for Sensor Network Monitoring DAG based In-Network Aggregation for Sensor Network Monitoring Shinji Motegi, Kiyohito Yoshihara and Hiroki Horiuchi KDDI R&D Laboratories Inc. {motegi, yosshy, hr-horiuchi}@kddilabs.jp Abstract Wireless

More information

Scaling 10Gb/s Clustering at Wire-Speed

Scaling 10Gb/s Clustering at Wire-Speed Scaling 10Gb/s Clustering at Wire-Speed InfiniBand offers cost-effective wire-speed scaling with deterministic performance Mellanox Technologies Inc. 2900 Stender Way, Santa Clara, CA 95054 Tel: 408-970-3400

More information

Efficient and low cost Internet backup to Primary Video lines

Efficient and low cost Internet backup to Primary Video lines Efficient and low cost Internet backup to Primary Video lines By Adi Rozenberg, CTO Table of Contents Chapter 1. Introduction... 1 Chapter 2. The DVP100 solution... 2 Chapter 3. VideoFlow 3V Technology...

More information

Optimizing Enterprise Network Bandwidth For Security Applications. Improving Performance Using Antaira s Management Features

Optimizing Enterprise Network Bandwidth For Security Applications. Improving Performance Using Antaira s Management Features Optimizing Enterprise Network Bandwidth For Security Applications Improving Performance Using Antaira s Management Features By: Brian Roth, Product Marketing Engineer April 1, 2014 April 2014 Optimizing

More information

CS 5480/6480: Computer Networks Spring 2012 Homework 1 Solutions Due by 9:00 AM MT on January 31 st 2012

CS 5480/6480: Computer Networks Spring 2012 Homework 1 Solutions Due by 9:00 AM MT on January 31 st 2012 CS 5480/6480: Computer Networks Spring 2012 Homework 1 Solutions Due by 9:00 AM MT on January 31 st 2012 Important: No cheating will be tolerated. No extension. CS 5480 total points = 32 CS 6480 total

More information

2004 Networks UK Publishers. Reprinted with permission.

2004 Networks UK Publishers. Reprinted with permission. Riikka Susitaival and Samuli Aalto. Adaptive load balancing with OSPF. In Proceedings of the Second International Working Conference on Performance Modelling and Evaluation of Heterogeneous Networks (HET

More information

Shared Backup Network Provision for Virtual Network Embedding

Shared Backup Network Provision for Virtual Network Embedding Shared Backup Network Provision for Virtual Network Embedding Tao Guo, Ning Wang, Klaus Moessner, and Rahim Tafazolli Centre for Communication Systems Research, University of Surrey Guildford, GU2 7XH,

More information

packet retransmitting based on dynamic route table technology, as shown in fig. 2 and 3.

packet retransmitting based on dynamic route table technology, as shown in fig. 2 and 3. Implementation of an Emulation Environment for Large Scale Network Security Experiments Cui Yimin, Liu Li, Jin Qi, Kuang Xiaohui National Key Laboratory of Science and Technology on Information System

More information

1. Comments on reviews a. Need to avoid just summarizing web page asks you for:

1. Comments on reviews a. Need to avoid just summarizing web page asks you for: 1. Comments on reviews a. Need to avoid just summarizing web page asks you for: i. A one or two sentence summary of the paper ii. A description of the problem they were trying to solve iii. A summary of

More information

Dynamic Load Balancing and Node Migration in a Continuous Media Network

Dynamic Load Balancing and Node Migration in a Continuous Media Network Dynamic Load Balancing and Node Migration in a Continuous Media Network Anthony J. Howe Supervisor: Dr. Mantis Cheng University of Victoria Draft: April 9, 2001 Abstract This report examines current technologies

More information

Experimentation with the YouTube Content Delivery Network (CDN)

Experimentation with the YouTube Content Delivery Network (CDN) Experimentation with the YouTube Content Delivery Network (CDN) Siddharth Rao Department of Computer Science Aalto University, Finland siddharth.rao@aalto.fi Sami Karvonen Department of Computer Science

More information

Routing with OSPF. Introduction

Routing with OSPF. Introduction Routing with OSPF Introduction The capabilities of an internet are largely determined by its routing protocol. An internet's scalability, its ability to quickly route around failures, and the consumption

More information

P2P VoIP for Today s Premium Voice Service 1

P2P VoIP for Today s Premium Voice Service 1 1 P2P VoIP for Today s Premium Voice Service 1 Ayaskant Rath, Stevan Leiden, Yong Liu, Shivendra S. Panwar, Keith W. Ross ARath01@students.poly.edu, {YongLiu, Panwar, Ross}@poly.edu, Steve.Leiden@verizon.com

More information

Concept of Cache in web proxies

Concept of Cache in web proxies Concept of Cache in web proxies Chan Kit Wai and Somasundaram Meiyappan 1. Introduction Caching is an effective performance enhancing technique that has been used in computer systems for decades. However,

More information

Per-Flow Queuing Allot's Approach to Bandwidth Management

Per-Flow Queuing Allot's Approach to Bandwidth Management White Paper Per-Flow Queuing Allot's Approach to Bandwidth Management Allot Communications, July 2006. All Rights Reserved. Table of Contents Executive Overview... 3 Understanding TCP/IP... 4 What is Bandwidth

More information

Optimum Path Computation Algorithms for Multimedia and Real Time Services over IP-Networks

Optimum Path Computation Algorithms for Multimedia and Real Time Services over IP-Networks Optimum Path Computation Algorithms for Multimedia and Real Time Services over IP-Networks 1 Optimum Path Computation Algorithms for Multimedia and Real Time Services over IP-Networks Rajiv kumar, Yuvraj

More information

QoS Strategy in DiffServ aware MPLS environment

QoS Strategy in DiffServ aware MPLS environment QoS Strategy in DiffServ aware MPLS environment Teerapat Sanguankotchakorn, D.Eng. Telecommunications Program, School of Advanced Technologies Asian Institute of Technology P.O.Box 4, Klong Luang, Pathumthani,

More information

Distributed Computing over Communication Networks: Topology. (with an excursion to P2P)

Distributed Computing over Communication Networks: Topology. (with an excursion to P2P) Distributed Computing over Communication Networks: Topology (with an excursion to P2P) Some administrative comments... There will be a Skript for this part of the lecture. (Same as slides, except for today...

More information

Internet Infrastructure Measurement: Challenges and Tools

Internet Infrastructure Measurement: Challenges and Tools Internet Infrastructure Measurement: Challenges and Tools Internet Infrastructure Measurement: Challenges and Tools Outline Motivation Challenges Tools Conclusion Why Measure? Why Measure? Internet, with

More information

A Comparison of General Approaches to Multiprocessor Scheduling

A Comparison of General Approaches to Multiprocessor Scheduling A Comparison of General Approaches to Multiprocessor Scheduling Jing-Chiou Liou AT&T Laboratories Middletown, NJ 0778, USA jing@jolt.mt.att.com Michael A. Palis Department of Computer Science Rutgers University

More information

Objectives. The Role of Redundancy in a Switched Network. Layer 2 Loops. Broadcast Storms. More problems with Layer 2 loops

Objectives. The Role of Redundancy in a Switched Network. Layer 2 Loops. Broadcast Storms. More problems with Layer 2 loops ITE I Chapter 6 2006 Cisco Systems, Inc. All rights reserved. Cisco Public 1 Objectives Implement Spanning Tree Protocols LAN Switching and Wireless Chapter 5 Explain the role of redundancy in a converged

More information

2. Research and Development on the Autonomic Operation. Control Infrastructure Technologies in the Cloud Computing Environment

2. Research and Development on the Autonomic Operation. Control Infrastructure Technologies in the Cloud Computing Environment R&D supporting future cloud computing infrastructure technologies Research and Development on Autonomic Operation Control Infrastructure Technologies in the Cloud Computing Environment DEMPO Hiroshi, KAMI

More information

Adaptive MAP Selection with Load Balancing Mechanism for the Hierarchical Mobile IPv6

Adaptive MAP Selection with Load Balancing Mechanism for the Hierarchical Mobile IPv6 Tamkang Journal of Science and Engineering, Vol. 12, No. 4, pp. 481 487 (2009) 481 Adaptive MAP Selection with Load Balancing Mechanism for the Hierarchical Mobile IPv6 Ying-Hong Wang, Chih-Peng Hsu* and

More information

Assignment #3 Routing and Network Analysis. CIS3210 Computer Networks. University of Guelph

Assignment #3 Routing and Network Analysis. CIS3210 Computer Networks. University of Guelph Assignment #3 Routing and Network Analysis CIS3210 Computer Networks University of Guelph Part I Written (50%): 1. Given the network graph diagram above where the nodes represent routers and the weights

More information

Congestion Control Overview

Congestion Control Overview Congestion Control Overview Problem: When too many packets are transmitted through a network, congestion occurs t very high traffic, performance collapses completely, and almost no packets are delivered

More information

Adaptive Hybrid Multicast with Partial Network Support

Adaptive Hybrid Multicast with Partial Network Support Adaptive Hybrid Multicast with Partial Network Support Huan Luo and Khaled Harfoush Department of Computer Science North Carolina State University Raleigh, NC,USA 27695 hluo,harfoush@cs.ncsu.edu Abstract

More information

Differenciated Bandwidth Allocation in P2P Layered Streaming

Differenciated Bandwidth Allocation in P2P Layered Streaming Differenciated Bandwidth Allocation in P2P Layered Streaming Abbas Bradai, Toufik Ahmed CNRS-LaBRI University of Bordeaux- 5, Cours de la libération. Talence, 45 {bradai, tad} @labri.fr Abstract There

More information

OSPF Routing Protocol

OSPF Routing Protocol OSPF Routing Protocol Contents Introduction Network Architecture Campus Design Architecture Building Block Design Server Farm Design Core Block Design WAN Design Architecture Protocol Design Campus Design

More information

Two-Stage Forking for SIP-based VoIP Services

Two-Stage Forking for SIP-based VoIP Services Two-Stage Forking for SIP-based VoIP Services Tsan-Pin Wang National Taichung University An-Chi Chen Providence University Li-Hsing Yen National University of Kaohsiung Abstract SIP (Session Initiation

More information

Enabling Conferencing Applications on the Internet using an Overlay Multicast Architecture

Enabling Conferencing Applications on the Internet using an Overlay Multicast Architecture Enabling Conferencing Applications on the Internet using an Overlay Multicast Architecture Yang-hua Chu, Sanjay G. Rao, Srinivasan Seshan and Hui Zhang Carnegie Mellon University {yhchu, sanjay, srini+,

More information

Path Selection Analysis in MPLS Network Based on QoS

Path Selection Analysis in MPLS Network Based on QoS Cumhuriyet Üniversitesi Fen Fakültesi Fen Bilimleri Dergisi (CFD), Cilt:36, No: 6 Özel Sayı (2015) ISSN: 1300-1949 Cumhuriyet University Faculty of Science Science Journal (CSJ), Vol. 36, No: 6 Special

More information

An Active Packet can be classified as

An Active Packet can be classified as Mobile Agents for Active Network Management By Rumeel Kazi and Patricia Morreale Stevens Institute of Technology Contact: rkazi,pat@ati.stevens-tech.edu Abstract-Traditionally, network management systems

More information

Path Optimization in Computer Networks

Path Optimization in Computer Networks Path Optimization in Computer Networks Roman Ciloci Abstract. The main idea behind path optimization is to find a path that will take the shortest amount of time to transmit data from a host A to a host

More information