August 19-23, 2002 | Suman Banerjee, Bobby Bhattacharjee, Christopher Kommareddy
The paper introduces a new scalable application-layer multicast protocol, NICE, designed for low-bandwidth, data streaming applications with large receiver sets. NICE is based on hierarchical clustering of peers and supports multiple data delivery trees with desirable properties. Extensive simulations and a wide-area testbed implementation are conducted to evaluate the protocol's performance. The results show that NICE achieves lower link stress (by about 25%), improved or similar end-to-end latencies, and similar failure recovery properties compared to the Narada protocol. Additionally, NICE uses significantly less control traffic, making it more efficient for large groups. The wide-area implementation demonstrates that NICE can handle groups of 32-100 members distributed across 8 different sites, with average control overheads below 1 Kbps for groups of size 100. The protocol's scalability and efficiency are highlighted, particularly in handling large receiver sets and low-bandwidth data streams.The paper introduces a new scalable application-layer multicast protocol, NICE, designed for low-bandwidth, data streaming applications with large receiver sets. NICE is based on hierarchical clustering of peers and supports multiple data delivery trees with desirable properties. Extensive simulations and a wide-area testbed implementation are conducted to evaluate the protocol's performance. The results show that NICE achieves lower link stress (by about 25%), improved or similar end-to-end latencies, and similar failure recovery properties compared to the Narada protocol. Additionally, NICE uses significantly less control traffic, making it more efficient for large groups. The wide-area implementation demonstrates that NICE can handle groups of 32-100 members distributed across 8 different sites, with average control overheads below 1 Kbps for groups of size 100. The protocol's scalability and efficiency are highlighted, particularly in handling large receiver sets and low-bandwidth data streams.