Goals, Key Features and Protocol Overview

The objectives of VTP include the establishment and release of an end-to-end con- nection, reliable delivery of data packets, in-sequence byte stream delivery of data to the application, flow and congestion control. The major design goals are to maximize the throughput of a connection, thereby preserving fairness among con- tending flows, and adapting the throughput of the different flows according to the minimum available bandwidth along the path, as it will be described in the follow- ing.

In order to achieve these goals, VTP must consider the characteristics of the highly dynamic, wireless vehicular environment, such as packet loss rate, end-to- end delay, delay jitter and reordering impact the transport layer [82]. VTP consid- ers these metrics in its design choices.

In order to provide the services, as described above, VTP relies on the follow- ing key features:

• Rate-based transmission,

• Decoupling of error and congestion control,

• Congestion control via explicit signaling in the packet header,

• Selective acknowledgments that are sent in periodic intervals depending on the current transmission rate and the source-destination distance, and • Use of statistical knowledge (e.g., expected communication/disruption dura-

tion) for rate calculation, error and congestion control. The remainder of this section provides an overview of VTP.

VTP establishes a connection between a sender and a receiver. This connec- tion is full-duplex, but the following description focuses on uni-directional traffic to simplify the explanation. A VTP connection can be either in a connected or disrupted state, indicating whether multi-hop connectivity exists or not.

In absence of acknowledgments, the VTP sender assesses whether a source- destination path exists or a disruption has occurred. The sender estimates the expected remaining communication duration based on the statistical knowledge of [82]. In case the statistical mean is below a threshold, the sender switches to disrupted state.

In disrupted state (e.g., the network is partitioned), the VTP sender only trans- mits periodic probe packets without data to detect path recovery. The rate of the probe packets is adjusted according to the statistically expected disruption duration for the given source-destination distance.

The arrival of acknowledgments indicates a connected state. In this state, the VTP sender steadily transmits packets at the maximum allowed data rate. This rate is determined by the feedback about available bandwidth along the path from intermediate nodes.

The congestion control of VTP uses explicit signaling of available bandwidth in the header of each data packet to adjust the transmission rate of the sender and avoid congestion. This mechanism decouples congestion control from error and flow control. The VTP sender inserts the minimum of its locally available or de- sired bandwidth in the VTP header of each packet. Intermediate nodes may reduce this value in the header: Each node measures and maintains the available band- width in its vicinity, as explained in Section 3.5.3.3. Furthermore, each interme- diate node periodically collects the number of connections on which it forwards and their respective desired bandwidths. This allows an allocation of bandwidth to each flow according to the max-min fairness algorithm [12]. An intermediate node reduces the header bandwidth field in case it exceeds its share of the available bandwidth.

When the VTP packet arrives at the receiver, it contains the minimum of the available bandwidth along the multi-hop path. The VTP receiver maintains this information in a weighted average function to smooth fluctuations. It includes this weighted average of the available bandwidth in the acknowledgments.

The VTP receiver uses selective acknowledgments (SACKs) in order to effi- ciently report blocks of lost packets. Furthermore, the receiver transmits SACKs in dynamic intervals in order to wait for delayed or reordered packets and to reduce the contention on the wireless channel. The dynamic calculation of the SACK in- terval considers the current transmission rate and the source-destination distance.

VTP provides reliability via retransmission of lost packets whereas the SACKs inform the sender about received and lost packets. The VTP sender maintains a retransmission timer per connection. When packets are not acknowledged before the retransmission timer expires, the packets are considered lost and scheduled for retransmission. However, the retransmission timeout calculation cannot rely on the typical round trip time (RTT) measurements because of the extreme fluctuation of RTTs in VANETs [82]. Therefore, the VTP sender calculates the retransmission timeout out of the expected SACK interval, the current transmission rate and the source-destination distance.

One of the main objectives of VTP is the aggressive convergence to the max- imum possible transmission rate of the connection (while preserving fairness to contending data traffic) to exploit even short connectivity periods. Therefore, VTP uses a quick start mechanism on connection establishment or after a disruption: The VTP sender transmits a syn packet to establish a connection or probe packets during a disruption to check when connectivity resumes. Upon reception of one of

3.5. VEHICULAR TRANSPORT PROTOCOL SPECIFICATION 65 these specific packets, the receiver replies immediately with an acknowledgment which contains the available bandwidth as collected by the syn or probe packet. Consequently, the VTP sender can initiate its transmission at the maximum pos- sible rate after one RTT. Note that the syn packet already contains data to exploit particularly short connection periods.

Section 3.5.3 explains in detail the respective transport mechanisms of VTP from sender, source and intermediate node perspectives. Before, Section 3.5.1 summarizes the basic assumptions of the ad hoc communication system and the vehicular environment.