Modified QUIC Protocol with Congestion Control for Improved Network Performance

Abstract

In a computer communication network, transport layer is responsible for reliable data delivery with guaranteed Quality of Service (QoS). Congestion control mechanism is one of the prime features of transport layer and predominantly responsible for maintaining the performance of network. The main contributions of this thesis are to modify the existing handshaking mechanism of Quick UDP Internet Connections (QUIC) protocol, called modified QUIC (ModQUIC) to reduce control overhead and Congestion Window (cwnd) size update delay. The suggested modification fine tunes the window update mechanism with Acknowledgment (ACK), that results in smooth variation in cwnd growth. Further, it regulates the network traffic which in turn helps to control congestion. In the first phase, we have suggested modification of the existing QUIC protocol called the ModQUIC protocol and its performance has been evaluated with Chromium server-client model testbed. The results and analysis are presented in comparison to QUIC and TCP in terms of performance measures like throughput and delay. Here, the performance has been tested for limited and sufficient link bandwidth in presence of loss, whereas the validation of results have been carried out with the help of linear regression model. In the result analysis, it was evident that ModQUIC achieves a throughput improvement of 35.66% and 51.93% over QUIC and TCP respectively, whereas delay is also reduced to 3% and 5% in comparison to QUIC and TCP. In the second phase, ModQUIC performance is verified in emulated environment using the Mininet for transport layer and browser network. The experimental results of ModQUIC are compared with QUIC, TCP and TCP/HTTP2 based on throughput, delay and fairness. It has been observed that ModQUIC outperforms throughput with an increased speed by minimizing transmission delay and improvised fairness. In addition, specifically for lossy link and low bandwidth, ModQUIC performance is better compared to QUIC and TCP. One of the QUIC features is inbuilt congestion control mechanism, which is responsible for maintaining streaming data. If there is no congestion, regular window size update is carried out as a step by step approach. In next phase of research work, ModQUIC protocol is investigated with CUBIC and Botitleneck Bandwidth Round-trip-propagation (BBR) congestion control mechanisms and suggested use of a decreased value for the factor β = βTCP=n for n flows, which are competing to acquire bottleneck resources, where βTCP is the decrease factor used in TCP. The Chromium server-client testbed experiment results show that the ModQUIC with BBR giver better performance in terms of throughput, delay and datarate. The result analysis actually gives an improvement of 6.8%, 19.06% and 27.9% for the ModQUIC/BBR as compared to ModQUIC/CUBIC, QUIC/BBR and QUIC/CUBIC respectively in terms of throughput. Furthermore, the delay is reduced by 8.02%, 6.56% and 14.38% over ModQUIC/CUBIC, QUIC/BBR and QUIC/CUBIC respectively. The QUIC protocol versions are updating faster and new versions of the same are available for users. In the final contribution of the research work, the performance of QUIC protocol has been tested with respect to congestion control using India’s rapidly growing Internet Service Provider (ISP); Reliance Jio 4G (JioFi) network. This experimental study investigated ModQUIC performance for congestion control mechanisms CUBIC and BBR in JioFi. The experiment is conducted using a testbed, developed with JioFi and Raspberry Pi-3 wireless router along with network emulator; Netem. Furthermore, the QUIC performance is verified with respect to Throughput and Retransmission Ratio (RTR) in which it is observed that overall ModQUIC/BBR performance is better than ModQUIC/CUBIC in the current Internet. We observed that Reliance Jio is an economical solution for the highly populated country like India, but not a contemporary solution to fulfill India’s newly launched digitization project.