Towards Scalability in Unstructured P2P Networks
| Autor: | Tai-Hung Lai, 賴泰宏 |
|---|---|
| Rok vydání: | 2005 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 93 P2P (Peer-to-Peer) networks are popular distributed systems. They are generally classified as the structured P2P and the unstructured P2P. Because unstructured P2P networks use flooding mechanism to forward messages, the overhead of traffics is increased with the growth of the P2P members and thus limits the scalable capability. Unstructured P2P infrastructure is adopted by most of the P2P applications because of its feasibility and flexibility. In this thesis we study the scalability problems of Gnutella which is one of the most existing Unstructured P2P. Two methods are explored: (1) improve the search performance to avoid flooding (2) add the structural property to Gnutella overlay, and it lacks the packet-level analysis. We port Gnutellasim simulator into our ns simulation environments to study the scalability of Gnutella. The simulation results show that the overlay degrees increase with the number of peers thus incurring lots of messages. When the overlay degrees grow, the performance will get better if the peer nodes have powerful ability to handle these messages. Otherwise the message traffics will increase the overhead of the system and worsen the performance. We also find the background traffics make the system perform worse, but increase the overlay degrees. Our simulations reveal that the overlay degrees will affect the performance. By evaluating the scalability properties of Gnutella, it will be feasible to find how to increase the overlay degrees without deteriorating the system performance. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|