TY - JOUR
T1 - Peering by exploiting peer heterogeneity
AU - Hsiao, Hung-Chang
AU - King, Chung Ta
AU - Gao, Shih Yen
PY - 2005/5/1
Y1 - 2005/5/1
N2 - Resource discovery in peer-to-peer (P2P) systems have been extensively studied. Unfortunately, most of the systems studied are not designed to take advantage of the heterogeneity in peer nodes. In this paper, we propose a novel P2P overlay called RATTAN, which serves as an underlay of a Gnutella-like network. RATTAN exploits the heterogeneity of peer nodes by structuring capable nodes as the core of the overlay. Using a tree-like structure, RATTAN can maximize the search scope with a minimal number of query messages. We evaluate RATTAN with simulation. The experiments show the following interesting results. First, RATTAN is robust by exploiting redundant overlay links. Second, the maximum bandwidth demand for processing the protocol of a single RATTAN overlay is nearly 1M bits/sec. However, around 80% of the nodes merely take 66 bits/sec. One implication is that we can use a small number of relatively capable peers (e.g., stable machines with a 100M bits/sec network interface) to process the 1M bits/sec protocol overhead and serve other peers that only need to spend 66 bits/sec for processing protocol overhead.
AB - Resource discovery in peer-to-peer (P2P) systems have been extensively studied. Unfortunately, most of the systems studied are not designed to take advantage of the heterogeneity in peer nodes. In this paper, we propose a novel P2P overlay called RATTAN, which serves as an underlay of a Gnutella-like network. RATTAN exploits the heterogeneity of peer nodes by structuring capable nodes as the core of the overlay. Using a tree-like structure, RATTAN can maximize the search scope with a minimal number of query messages. We evaluate RATTAN with simulation. The experiments show the following interesting results. First, RATTAN is robust by exploiting redundant overlay links. Second, the maximum bandwidth demand for processing the protocol of a single RATTAN overlay is nearly 1M bits/sec. However, around 80% of the nodes merely take 66 bits/sec. One implication is that we can use a small number of relatively capable peers (e.g., stable machines with a 100M bits/sec network interface) to process the 1M bits/sec protocol overhead and serve other peers that only need to spend 66 bits/sec for processing protocol overhead.
UR - http://www.scopus.com/inward/record.url?scp=84893455731&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84893455731&partnerID=8YFLogxK
U2 - 10.1108/17427370580000116
DO - 10.1108/17427370580000116
M3 - Article
AN - SCOPUS:84893455731
VL - 1
SP - 101
EP - 114
JO - International Journal of Pervasive Computing and Communications
JF - International Journal of Pervasive Computing and Communications
SN - 1742-7371
IS - 2
ER -