A data sender in an IP-based network is only capable of efficiently using a network path if it knows the packet size limit of the path, i.e., the Path Maximum Transmission Unit (PMTU). An IP node that cannot forward and must drop an oversized packet notifies the sender with a Packet Too Big (PTB) message, which usually contains the maximum packet size it can forward. Classical PMTU Discovery (PMTUD) relies on receiving a PTB message when sending a packet that is too big for the path. However, in the Internet, scenarios exist where a sender does not receive a PTB message in such a case. Hence, a PMTUD framework for transport protocols has been specified that works without a PTB message.

One such transport protocol is QUIC. QUIC was only recently standardized, but as the transport protocol for the newest version of HTTP (HTTP/3), it is already widely in use in today’s Internet. The specification of QUIC recommends the use of the PMTUD framework. However, many QUIC implementations do not have PMTUD integrated. Instead, they restrict outgoing packets to a size that is smaller than the PMTU of most paths. Sending packets smaller than possible means splitting data into smaller parts and, therefore, sending more packets than necessary. Since devices in the network spend most of their computing power per packet rather than per byte, this wastes energy and may increase hardware costs (e.g., if, in consequence, a computer serving multiple QUIC endpoints needs more or more powerful hardware resources).

This dissertation presents a modified version of the specified PMTUD framework and describes how to integrate it into QUIC. In particular, it presents a PMTU search algorithm and a PTB detection algorithm, which are required to complete the framework into a full PMTUD solution. A QUIC simulation model has been developed to evaluate the PMTUD with these algorithms in QUIC. One result from simulations, for example, shows how many fewer packets a QUIC endpoint needs to send to transfer a message when using this PMTUD.