Citation Context

...ths are 1.5 Mb/s, the traffic sources are modeled as a sixlayer CBR stream at rates 32 \Theta 2 m kb/s; m = 0 : : : 5, and the start-time of each receiver is randomly chosen uniformly on the interval =-=[30; 120]-=- seconds. The protocol constants from Table 1 have the following values: ff = 2, fi = 2=3, k 1 = 1, k 2 = 2, g 1 = 0:25, g 2 = 0:25, T min J = 5 sec, T max J = 600 sec. Each join-timer interval is cho...

Citation Context

...yer. Under this scheme, a receiver searches for the optimal level of subscription much as a TCP source searches for the bottleneck transmission rate with the slow-start congestion avoidance algorithm =-=[21]-=-. The receiver adds layers until congesR 1 R 3 R S 2 1 R 2 R 3 128 kb/s 10Mb/s 512 kb/s 10Mb/s 10Mb/s R S Figure 3: End-to-end adaptation. tion occurs and backs off to an operating point below this bo...

Citation Context

... TCP congestion control works well in isolation but in aggregation can be unfair [13]. As an optimization, network mechanism can be introduced to make TCP perform better: Random Early Detection (RED) =-=[14]-=- gateways or Fair Queuing (FQ) [10] routers minimize the interaction between connections to improve fairness. Similarly, we can design RLM to behave relatively well in a loosely controlled, drop-tail,...

Citation Context

...n to the overall group size. In other words, we can fix the aggregate joinexperiment rate independent of session size much as RTCP scales back its control message rate in proportion to the group size =-=[28]. However,-=- reducing the experiment rate in this manner decreases the learning rate. For large groups, the algorithm will take too long to converge. Our solution is "shared learning": Before a receiver...

Citation Context

... for evaluating the protocol 's performance, but the simulator provided feedback that was critical to the design process. Ns is an eventdriven packet-level simulator controlled and configured via Tcl =-=[27]-=-. Shortest-path routes are computed for the input topology and multicast packets are routed via reverse-path forwarding. A flooding algorithm similar to Dense-mode Protocol Independent Multicast (PIM)...

Citation Context

...in isolation but in aggregation can be unfair [13]. As an optimization, network mechanism can be introduced to make TCP perform better: Random Early Detection (RED) [14] gateways or Fair Queuing (FQ) =-=[10]-=- routers minimize the interaction between connections to improve fairness. Similarly, we can design RLM to behave relatively well in a loosely controlled, drop-tail, besteffort network, and as an opti...

Citation Context

...g RLM. This approach has the desirable side effect that RLM is shielded from interactions with other protocols. The RLM framework could be combined with the Scalable Reliable Multicast (SRM) protocol =-=[17]-=- in the LBNL whiteboard,swb, to optimize the latency of rate-controlled transmissions. Because SRM uses a token-bucket rate-controller, it has the same limitations that single-layer video has in heter...

Citation Context

...long paths with potentially high variability in bandwidth. The simplest solution to this problem is to distribute a uniform representation of the signal to all interested receivers using IP Multicast =-=[8]-=-. Unfortunately, this is suboptimal --- low-capacity regions of the network suffer congestion while high-capacity regions are underutilized. The problems posed by heterogeneity are not just theoretica...

Citation Context

...wo techniques for adaptation. Their first technique is a negotiation algorithm run by each receiver that obtains the highest available quality of service explicitly from the network (e.g., using RSVP =-=[35]-=-). Their second approach uses layered multicast with an aggressive adaptation scheme where a new receiver subscribes to all the layers in the distribution and drops layers until the quality of the del...

Citation Context

...mission system, ensuring that the pieces of the design interact well with each other. To this end, our system is based on Clark and Tennenhouse's Application Level Framing (ALF) protocol architecture =-=[5]. While AL-=-F says that an application's semantics should be reflected in the design of its network protocol, we further believe that the "network's semantics" should also be reflected in the applicatio...

Citation Context

... hand, if the granularity of resource management were not as fine-grained, then RLM adaptation within an integrated services environment might still make sense. For example, Class Based Queuing (CBQ) =-=[16] could be -=-used to provide an "adaptive-rate video" traffic class with some specified bandwidth. Then within this CBQ class, video sessions could contend for the aggregate class bandwidth using RLM. Th...

Citation Context

...ero-mean noise process to model variable coding delays (fN k g is i.i.d. uniform on [\Gamma\Delta=2; \Delta=2]). Unfortunately, this simple model fails to capture the burstiness of real video streams =-=[18]-=-. Because convergence in RLM relies on matching the layered rates to available capacity, smooth sources are well-behaved and this traffic model is overly optimistic. On the other hand, a bursty source...

Citation Context

... Shortest-path routes are computed for the input topology and multicast packets are routed via reverse-path forwarding. A flooding algorithm similar to Dense-mode Protocol Independent Multicast (PIM) =-=[7]-=- handles forwarding and pruning of multicast flows. Hierarchical sources are modeled as a set of constant-bit rate (CBR) streams with fixed packet sizes. Packets are generated at times defined by the ...

Citation Context

...ernet. As depicted in Figure 1, a video application is run on a "seminar host" that sources a single-rate signal at 128 kb/s, the nominal rate for video over the Internet Multicast Backbone,=-= or MBone [11]-=-. However, a number of users on the local campus network have high bandwidth connectivity and would prefer to receive higherrate, higher-quality video. At the other bandwidth extreme, many users have ...

Citation Context

...lready widely deployed in the Internet, we can field our system by building it into an application. We are currently implementing RLM and our layered codec in the UCB/LBNL video conferencing tool vic =-=[25]-=-. Vic's network transport is based on the Real-time Transport Protocol (RTP) [28], an application level protocol for multimedia transport standardized by the Internet Engineering Task Force. Although ...

Citation Context

...he source simply cannot meet the conflicting requirements of a set of heterogeneous receivers. An alternative approach is to combine a layered compression algorithm with a layered transmission scheme =-=[29, 32]-=-. In this approach, a signal is encoded into a number of layers that can be incrementally combined to provide progresSIGCOMM -- August 1996 -- Stanford, CA 2 64 kb/s Gateway H Ethernets H Router MBone...

Citation Context

...nging network conditions. A better approach is to adjust the transmission rate to match the available capacity in the network, i.e., to react to congestion. Pioneering research in rate-adaptive video =-=[1, 19, 23]-=- has shown that this is feasible, but unfortunately, in the context of multicast, the notion of network capacity is ill defined. A control scheme that adjusts the rate of a single stream at the source...

Citation Context

...son for the transition, either packet loss or a timeout. Actions associated with a transition are indicated in parentheses. Join-timers (T J ) are randomized to avoid protocol synchronization effects =-=[15]-=-, while detection-timers (T D ) are set to a scaled value of the detection-time estimator. The add action implies that we subscribe to the next layer in the multicast group hierarchy, while the drop a...

Citation Context

...added to the network, RLM should work effectively in concert with it. Similar circumstance surrounds the design of TCP. TCP congestion control works well in isolation but in aggregation can be unfair =-=[13]-=-. As an optimization, network mechanism can be introduced to make TCP perform better: Random Early Detection (RED) [14] gateways or Fair Queuing (FQ) [10] routers minimize the interaction between conn...

Citation Context

...ceivers to correlate joinexperiments with resulting congestion periods, the network must instantiate a new flow expediently. In the case of IP Multicast, the Internet Group Management Protocol (IGMP) =-=[12]-=- carries out both of these operations on reasonable time scales. When a receiver joins a new group, the host immediately informs the next-hop router, which in turn, immediately propagates a graft mess...

Citation Context

...c. RLM alone does not provide fairness. In general it is not possible to achieve a "fair" allocation of bandwidth without some additional machinery in the network, even if all the end-nodes =-=cooperate [22]-=-. But, if machinery for fairness is added to the network, RLM should work effectively in concert with it. Similar circumstance surrounds the design of TCP. TCP congestion control works well in isolati...

Citation Context

...nging network conditions. A better approach is to adjust the transmission rate to match the available capacity in the network, i.e., to react to congestion. Pioneering research in rate-adaptive video =-=[1, 19, 23]-=- has shown that this is feasible, but unfortunately, in the context of multicast, the notion of network capacity is ill defined. A control scheme that adjusts the rate of a single stream at the source...

Citation Context

...he source simply cannot meet the conflicting requirements of a set of heterogeneous receivers. An alternative approach is to combine a layered compression algorithm with a layered transmission scheme =-=[29, 32]-=-. In this approach, a signal is encoded into a number of layers that can be incrementally combined to provide progresSIGCOMM -- August 1996 -- Stanford, CA 2 64 kb/s Gateway H Ethernets H Router MBone...

Citation Context

...ution. While much of the previous work leaves this as an implementation detail, a novel mechanism based on IP Multicast was suggested by Deering [6] and elaborated on and/or independently reported in =-=[4, 9, 20, 26, 33]-=-. In this approach, the different layers of the hierarchical signal are striped across multiple multicast groups and receivers adapt to congestion by adding and dropping layers (i.e., joining and leav...

Citation Context

...interactions that could result in oscillatory behavior. We will assess the impact of such interactions in future work. We implemented the RLM protocol described above in the LBNL network simulator ns =-=[24]-=-. Not only did this implementation serve as a framework for evaluating the protocol 's performance, but the simulator provided feedback that was critical to the design process. Ns is an eventdriven pa...

Citation Context

...ution. While much of the previous work leaves this as an implementation detail, a novel mechanism based on IP Multicast was suggested by Deering [6] and elaborated on and/or independently reported in =-=[4, 9, 20, 26, 33]-=-. In this approach, the different layers of the hierarchical signal are striped across multiple multicast groups and receivers adapt to congestion by adding and dropping layers (i.e., joining and leav...

Citation Context

...ution. While much of the previous work leaves this as an implementation detail, a novel mechanism based on IP Multicast was suggested by Deering [6] and elaborated on and/or independently reported in =-=[4, 9, 20, 26, 33]-=-. In this approach, the different layers of the hierarchical signal are striped across multiple multicast groups and receivers adapt to congestion by adding and dropping layers (i.e., joining and leav...

Citation Context

...nging network conditions. A better approach is to adjust the transmission rate to match the available capacity in the network, i.e., to react to congestion. Pioneering research in rate-adaptive video =-=[1, 19, 23]-=- has shown that this is feasible, but unfortunately, in the context of multicast, the notion of network capacity is ill defined. A control scheme that adjusts the rate of a single stream at the source...

Citation Context

...ution. While much of the previous work leaves this as an implementation detail, a novel mechanism based on IP Multicast was suggested by Deering [6] and elaborated on and/or independently reported in =-=[4, 9, 20, 26, 33]-=-. In this approach, the different layers of the hierarchical signal are striped across multiple multicast groups and receivers adapt to congestion by adding and dropping layers (i.e., joining and leav...

Citation Context

...forwarding of subflows along all the links in a distribution. While much of the previous work leaves this as an implementation detail, a novel mechanism based on IP Multicast was suggested by Deering =-=[6]-=- and elaborated on and/or independently reported in [4, 9, 20, 26, 33]. In this approach, the different layers of the hierarchical signal are striped across multiple multicast groups and receivers ada...

Citation Context

...output rate of the source coder, allowing the source to react to queue buildup before packet loss occurs. These source-based rate-adaptation schemes are poorly matched to multicast environments. QMTP =-=[34]-=- and the IVS congestion control scheme [1] adapt by soliciting feedback from the receivers in a scalable fashion, but these schemes do not cope well with bandwidth heterogeneity. Either lowcapacity re...

Citation Context

...n for interoperable real-time audio/video applications, it was designed without any explicit notion of a layered signal representation. In joint work, we have extended RTP for layered stream delivery =-=[31] and are c-=-urrently implementing our proposed changes in vic. Since the RLM protocol processing is not in the "fast path", run-time performance is not critical. In fact, our prototype is implemented al...

Citation Context

...ution. While much of the previous work leaves this as an implementation detail, a novel mechanism based on IP Multicast was suggested by Deering [6] and elaborated on and/or independently reported in =-=[4, 9, 20, 26, 33]-=-. In this approach, the different layers of the hierarchical signal are striped across multiple multicast groups and receivers adapt to congestion by adding and dropping layers (i.e., joining and leav...

Citation Context

...a multi-resolution extension to the CU-SeeMe video conferencing system where IP Multicast receivers can subscribe to either a 160x120 or a 320x240 stream by joining either one or two multicast groups =-=[2]-=-. Receivers drop down to the 160x120 resolution when they detect high packet loss rates. Concurrent with our work, Hoffman and Speer have built a similar system based on the layered multicast architec...