P2P Definition

P2P Home


The hybrid P2P is regarded as centralized, but not in the sense of a general-purpose server. It has router-terminals that are used for holding a catalog of addresses. The addresses are referenced by a set of indexes, which determines the appropriate address set. This router-terminal could be dynamic or static.

If it is dynamic then the router-terminal could be characterized as a peer group manager. However, we could diminish the functions of the centralized terminal until it is nothing more than a router database controlled by the peer interface.

As a router database, all powers of management remain with the peers. This eliminates the centralized computer (router-terminal set) as a peer group manager. All management functions are then held in the peers.

All the peers in a peer group have equal and similar status. The more management power we take out of a peer (and therefore out of all of them since all peers have equal management powers) the more we degrade the P2P architecture.

This implies that even if the routers can dynamically update their routing catalogs (router-terminals functioning as agents), that the use of these catalogs is within the management scope of every peer.

It is feasible that other than dynamic updates done by the router-terminals (possible agent-driven) that each peer could have the power for updating the database, at least in regard to what they are willing to store and upload.

Router-terminals, as catalogs of addresses, are referenced by indexes. Independent of upgrades, there are two general ways for hybrid P2P to utilize the router terminals.

1. The peer in client mode sends request to the router-terminal. The router-terminal memory is scanned for an open address of a peer that holds the requested object and forwards the request to the selected peer-server (peer in server mode) the peer-server downloads directly to peer-client (peer in client mode).

This could be mapped as
peer-client ---> router terminal
router-terminal ---> peer-server (receives peer-client address for download)
peer-server ---> peer-client (receives download)

2. The peer in client mode sends request to router, the router returns the required address set, the peer in client mode scans for an open address (peer node that is on) and sends download request to the open address. The receiver of the request functions in peer-server mode and downloads the request.

This could be mapped as
peer-client ---> router terminal
router-terminal ---> peer-client (receives required address, possibly a set)
peer-client ---> peer-server (receives request)
peer-server ---> peer-client (download to peer-client)

In the second mapping, the process burden for the router is less then in the first mapping. In other words, in the second mapping, the processing can be put in possession of the peers where in P2P it likely should be. For example, in the event of a set of addresses, the program scanning for an open address of a peer containing the requested item can be operated in the peer that is in peer-client mode. This also minimizes congestion at the router in terms of the query processing time at the router. This is because the router downloads are quite small, relatively.

It is now understood that even in a central catalog structure that the router is not a server in the proper sense of the term. It functions more like a peer-group database and the database management system is maintained in the peers.

Note that the P2P structure also suggests the possibility that the router itself can be distributed. With dynamic distribution (to maximize equality) this opens up the possibility of a pure P2P architecture that uses distributed catalogs.

copyright 2001 by Ross Lee Graham

P2P Definition

Pure P2P

P2P Home

Ross Lee Graham
Last modified: Thu Apr 19 19:52:30 MET DST 2001