Method of multicast file distribution and synchronization

DETAILED DESCRIPTION This invention presents a method that delivers arbitrary data from a single source node, known as a server, to a large number of receivers using a novel method based on the selection of an active receiver.
The active receiver selection is done using a novel token granting mechanism.
In addition, the protocol introduces methods for efficiently transmitting and retransmitting data segments.
The active receiver is responsible for providing flow control feedback to the server during data transfer as well as generating requests for retransmission.
All receivers are offered a chance to become an active receiver in a controlled manner, and only one receiver can be an active receiver at any given time.
This process continues until there are no receivers in the group with any missing segments.
The second task of the protocol provides synchronization of the file version to ensure that all receivers have the last distributed version of the data file.
The inventive method eliminates the acknowledgment implosion problem associated with multicast transport protocols by making only one receiver responsible for generating acknowledgments and also requesting retransmissions.
In addition, it provides flow control, avoids duplicate transmissions and makes effective use of unicast and multicast communication.
The invented method is especially suitable for networks where the server and receivers are connected through long delay media such as satellite links.
None of the earlier disclosed mechanisms can work efficiently on satellite links as they depend on random time-outs at receivers.
The current invention is also suitable for wide area networks with low latencies.
In addition to being independent of topology constraints, the invented method does not require support of IP multicast routing.
It can work both in networks that support multicast routing and in those that do not yet support multicast routing.