Abstract: A download manger running on a computer system identifies an in-progress download of content by the computer system directly from a content system. The download manager causes the computer system to join a peer-to-peer network in which the content is being shared. The computer system starts to receive data blocks of the content from peer-to-peer network, as it continues the download from the content system. Based on the receipt of the content from the peer-to-peer network and from the content system, the download manager determines whether the computer system should rely primarily on the peer-to-peer network instead of the content system for receiving the content. If a determination is made to rely on the peer-to-peer network instead of the content system, the download manager terminates the download from the content system and continues receiving data blocks of the content from the peer-to-peer network.

Abstract: According to a peer-to-peer protocol, a peer-to-peer network includes multiple clubs and multiple peers that receive content distributed by a source of the network. Each of the multiple peers is part of at least two of the clubs. A joining peer identifies a club of the peer-to-peer network to join and receives identifying information for peers in the identified club. The joining peer randomly selects from the peers in the identified club at least a first peer and a second peer with which to establish connections. The random selection is proportionally weighted by the number of connections the first peer has with other peers and the number of connections the second peer has with other peers. The joining peer establishes a first connection with the first peer and a second connection with the second peer.

Abstract: A Peer-to-Peer protocol such as BitTorrent is used to assist streaming. Peers download streaming content from the P2P network while simultaneously playing the downloaded content. As the stream plays, an end system downloads any missing pieces directly from a server or other infrastructure node. This method roughly squares server capacity and can be refined to require on average 0(1) servers regardless of the number of concurrent users. Thus BitTorrent assisted streaming scales better than traditional server-client and other infrastructure-only solutions, each of which requires a number of infrastructure nodes that scale linearly as a function of the number of users.

Abstract: A peer in a peer-to-peer network joins an advertisement room having an associated identifier. A joining peer identifies a peer closest to the advertisement room. In various embodiments, the peer that is closest to the advertising room is the peer having a discreet hash table ID closest to the advertisement room identifier. The joining peer requests from the closest peer information about peers that the closest peer knows to be part of the advertisement room. The closest peer provides the joining peer with connection information for those peers, and the joining peer establishes connections with the peers. Peers in the advertisement room provide advertisements for chat rooms. To join a chat room, a peer determines which of its peers in the advertisement room is the closest peer to the chat room, based on the chat room identifier, and informs that peer that it is now part of the chat room.

Abstract: According to a peer-to-peer protocol, a peer-to-peer network includes multiple clubs and multiple peers that receive content distributed by a source of the network. Each of the plurality of peers is part of at least two of the clubs. As content is generated, the source divides the content into multiple data blocks and assigns each data block to a club. Each data block is transmitted to peers that belong to the club assigned to the data block. When a peer in a club receives a data block assigned to that club, the peer distributes the data block to other peers in the club. Additionally, the peer transmits the data block to peers in other clubs. Also, the peer receives data blocks assigned to other clubs from peers that are not members of the club.

Abstract: A download manger running on a computer system identifies an in-progress download of content by the computer system directly from a content system. The download manager causes the computer system to join a peer-to-peer network in which the content is being shared. The computer system starts to receive data blocks of the content from peer-to-peer network, as it continues the download from the content system. Based on the receipt of the content from the peer-to-peer network and from the content system, the download manager determines whether the computer system should rely primarily on the peer-to-peer network instead of the content system for receiving the content. If a determination is made to rely on the peer-to-peer network instead of the content system, the download manager terminates the download from the content system and continues receiving data blocks of the content from the peer-to-peer network.

Abstract: Transfer policy management methods in a peer-to-peer networking environment such as BitTorrent are provided. A download TPM maximizes the usage of available bandwidth, while focusing on as few torrents as possible. For a given list of torrents, a peer in accordance with the present invention determines how many download connections to maintain at a given time in order to maximize usage of bandwidth. If the variance of per-connection bandwidths exceeds a threshold, more connections are added until either variance does not exceed the threshold or a maximum number of connections are established. A seeding peer determines when to reduce the number of files being seeded by monitoring its average upload rate per torrent. If its average rate falls below a threshold, then the peer stops seeding a file. The file that the peer stops seeding is one to which the peer is making the smallest overall contribution.

Abstract: A system, method, and computer program product replace a failed node storing data relating to a portion of a data file. An indication of a new storage node to replace the failed node is received at each of a plurality of available storage nodes. The available storage nodes each contain a plurality of shares generated from a data file. These shares may have been generated based on pieces of the data file using erasure coding techniques. A replacement share is generated at each of the plurality of available storage nodes. The replacement shares are generated by creating a linear combination of the shares at each node using random coefficients. The generated replacement shares are then sent from the plurality of storage nodes to the indicated new storage node. These replacement shares may later be used to reconstruct the data file.

Abstract: According to a peer-to-peer protocol, a peer-to-peer network includes multiple clubs and multiple peers that receive content distributed by a source of the network. Each of the plurality of peers is part of at least two of the clubs. As content is generated, the source divides the content into multiple data blocks and assigns each data block to a club. Each data block is transmitted to peers that belong to the club assigned to the data block. When a peer in a club receives a data block assigned to that club, the peer distributes the data block to other peers in the club. Additionally, the peer transmits the data block to peers in other clubs. Also, the peer receives data blocks assigned to other clubs from peers that are not members of the club.

Abstract: Classification and verification of static file transfer protocols is provided. A network node sitting in the path of packets classifies traffic according to its content, and then statistically verifies that the content actually matches the initial classification. Classification and verification are standard building blocks in networks systems that provide quality-of-service. However, unlike traditional quality-of-service building blocks, a system of the present invention takes a trust-but-verify approach, and thus can rely on longer-timescale statistical online or offline verification to reduce computational requirements. This improves scalability and reduces the cost and complexity that in part has prevented prior service differentiation systems from succeeding in the Internet. Internet Service Providers (ISPs) can combine the classification and verification building blocks with a networked directory service to allow further classification as to whether particular content is for pay.

Abstract: Dynamic rate limiting of background traffic to alleviate congestion in the access network is enabled. ICMP echo round-trip times and ICMP losses to a nearby node outside the local area and just beyond the divergence in end-to-end paths are measured, allowing unambiguous discrimination of nearby from distant congestion points. Using round-trip time samples, either short-run delay or short-run variance in delay can be measured to estimate congestion. When combined with an appropriate control law, background traffic can be rapidly reduced to allow interactive traffic to traverse unhindered through the access network. The described system and methods can be implemented in the application-layer and without any additional support from the network.

Abstract: A peer in a peer-to-peer network joins an advertisement room having an associated identifier. A joining peer identifies a peer closest to the advertisement room. In various embodiments, the peer that is closest to the advertising room is the peer having a discreet hash table ID closest to the advertisement room identifier. The joining peer requests from the closest peer information about peers that the closest peer knows to be part of the advertisement room. The closest peer provides the joining peer with connection information for those peers, and the joining peer establishes connections with the peers. Peers in the advertisement room provide advertisements for chat rooms. To join a chat room, a peer determines which of its peers in the advertisement room is the closest peer to the chat room, based on the chat room identifier, and informs that peer that it is now part of the chat room.

Abstract: According to a peer-to-peer protocol, a peer-to-peer network includes multiple clubs and multiple peers that receive content distributed by a source of the network. Each of the plurality of peers is part of at least two of the clubs. As content is generated, the source divides the content into multiple data blocks and assigns each data block to a club. Each data block is transmitted to peers that belong to the club assigned to the data block. When a peer in a club receives a data block assigned to that club, the peer distributes the data block to other peers in the club. Additionally, the peer transmits the data block to peers in other clubs. Also, the peer receives data blocks assigned to other clubs from peers that are not members of the club.

Abstract: Classification and verification of static file transfer protocols is provided. A network node sitting in the path of packets classifies traffic according to its content, and then statistically verifies that the content actually matches the initial classification. Classification and verification are standard building blocks in networks systems that provide quality-of-service. However, unlike traditional quality-of-service building blocks, a system of the present invention takes a trust-but-verify approach, and thus can rely on longer-timescale statistical online or offline verification to reduce computational requirements. This improves scalability and reduces the cost and complexity that in part has prevented prior service differentiation systems from succeeding in the Internet. Internet Service Providers (ISPs) can combine the classification and verification building blocks with a networked directory service to allow further classification as to whether particular content is for pay.

Abstract: Classification and verification of static file transfer protocols is provided. A network node sitting in the path of packets classifies traffic according to its content, and then statistically verifies that the content actually matches the initial classification. Classification and verification are standard building blocks in networks systems that provide quality-of-service. However, unlike traditional quality-of-service building blocks, a system of the present invention takes a trust-but-verify approach, and thus can rely on longer-timescale statistical online or offline verification to reduce computational requirements. This improves scalability and reduces the cost and complexity that in part has prevented prior service differentiation systems from succeeding in the Internet. Internet Service Providers (ISPs) can combine the classification and verification building blocks with a networked directory service to allow further classification as to whether particular content is for pay.

Abstract: Dynamic rate limiting of background traffic to alleviate congestion in the access network is enabled. ICMP echo round-trip times and ICMP losses to a nearby node outside the local area and just beyond the divergence in end-to-end paths are measured, allowing unambiguous discrimination of nearby from distant congestion points. Using round-trip time samples, either short-run delay or short-run variance in delay can be measured to estimate congestion. When combined with an appropriate control law, background traffic can be rapidly reduced to allow interactive traffic to traverse unhindered through the access network. The described system and methods can be implemented in the application-layer and without any additional support from the network.

Abstract: A system, method, and computer program product replace a failed node storing data relating to a portion of a data file. An indication of a new storage node to replace the failed node is received at each of a plurality of available storage nodes. The available storage nodes each contain a plurality of shares generated from a data file. These shares may have been generated based on pieces of the data file using erasure coding techniques. A replacement share is generated at each of the plurality of available storage nodes. The replacement shares are generated by creating a linear combination of the shares at each node using random coefficients. The generated replacement shares are then sent from the plurality of storage nodes to the indicated new storage node. These replacement shares may later be used to reconstruct the data file.

Abstract: Dynamic rate limiting of background traffic to alleviate congestion in the access network is enabled. ICMP echo round-trip times and ICMP losses to a nearby node outside the local area and just beyond the divergence in end-to-end paths are measured, allowing unambiguous discrimination of nearby from distant congestion points. Using round-trip time samples, either short-run delay or short-run variance in delay can be measured to estimate congestion. When combined with an appropriate control law, background traffic can be rapidly reduced to allow interactive traffic to traverse unhindered through the access network. The described system and methods can be implemented in the application-layer and without any additional support from the network.