RTP is a connection-oriented, full-duplex protocol designed to support data in high-speed networks. RTP connections are established within an HPR subnet and are used to carry session traffic. These connections can be thought of as transport pipes over which sessions are carried.

RTP connections can carry data at very high speeds by using low-level intermediate routing and minimizing the number of flows over the links for error recovery and flow control.

The RTP functions include:

• Nondisruptive path switching

  An RTP connection's physical path can be switched automatically to reroute sessions around a failure in the network. The RTP connection is reestablished over a new physical path that bypasses the failing link or node, and the session's traffic flow is resumed on the RTP connection non-disruptively. Any data that was in the network at the time of the failure is recovered automatically using RTP′s end-to-end error recovery.

• End-to-end error recovery

  In basic APPN, error recovery is done on every link in a network. To address the emerging high-speed lines with much lower bit error rates, HPR removed the requirement to do link-level error recovery and instead does error recovery on an end-to-end basis. This improves performance by reducing the number of flows required to do the link-level error recovery on every link. RTP also supports selective retransmission, where only missing or corrupted packets are resent, and not all packets since the failure occurred.

• Re-sequencing packets

  A major observable fact in a multilink transmission group (MLTG) is that packets may arrive at the endpoint out of sequence. The RTP endpoints re-sequence the data in this case.

• End-to-end flow control and congestion control

  Flow control is the mechanism that controls the pace at which data is sent into the network to prevent flooding the resources along the route and to prevent the endpoint from being congested. In an APPN network, flow control is done on each stage of the session by using adaptive session-level pacing. This method provided excellent performance for networks with low speed lines and poor quality. For high-speed networks, adaptive session-level pacing was found inadequate due to the amount of processing required in each node.

  HPR introduced a protocol that is suited for high-speed routing called adaptive rate based (ARB) flow/congestion control. It regulates the flow of traffic over an RTP connection by adaptively changing the sender′s rate based on feedback from the receiver. This protocol allows for high link utilization and prevents congestion before it occurs, rather than recovering from congestion after it occurs.

Figure 1. RTP connection

Figure 1 shows an RTP connection that is carrying multiple sessions. Traffic from many sessions requesting the same class of service can be routed over the same RTP connection.