Research On Coupled Congestion Control And Data Scheduling Mechanism In MPTCP-based Multi-path Data Transmission

With the development of various access technologies,a communication device usually has multiple network interfaces.As a new bandwidth aggregation technology at the transport layer,MPTCP(Multipath TCP)has got much attention due to its support for concurrent use of multiple interfaces and compatibility with applications.It aims at providing a fully reliable and bandwidth aggregation service while keeping fairness with regular single-path TCP.In recent years,domestic and foreign institutions have made a lot of research to improve the performance of MPTCP.However,there are still have some problems in MPTCP congestion control algorithm,packet scheduling algorithm,and integration with practical applications.First of all,the existing congestion control algorithms of MPTCP are based on Net-work Fairness by default.Network Fairness severely limits the performance of MPTCP subflows at non-bottleneck,resulting in inefficient use of link resources.We should de-sign the congestion control algorithm based on the principle of Bottleneck Fairness,and only couple control subflows at the bottleneck to achieve fairness to single-path TCP.At the non-bottleneck position,the subflow should be decoupled to make it operate as a single-path TCP to improve the overall throughput.Secondly,random wireless packet loss often occurs in wireless heterogeneous networks.The performance of the existing MPTCP congestion control algorithm drops significantly in this scenario,we need to propose a new ongestion control algorithm to solve this problem and further give the corresponding scheduling algorithm.Finally,due to the transparency to upper-layer ap-plications,the existing MPTCP transmission scheme does not consider the requirements of upper-layer applications such as VR(Virtual Reality)when performing data schedul-ing,and blindly performs data scheduling for different paths.Therefore,a cross-layer optimization strategy is needed to improve the entire transmission process.In order to solve the above problems,this dissertation proposes the following three solutions:(1)This dissertation proposes a new Bottleneck Fairness based multipath trans-mission scheme,which consists of two main parts,a new Bottleneck Fairness based congestion control algorithm,and a new Bottleneck Fairness based packet schedul-ing algorithm.The Bottleneck Fairness based congestion control algorithm uses ECN(Explicit Congestion Notification)mechanism to detect shared bottlenecks among sub-flows.Meanwhile,it implements the judgement with two stages and makes sure that two judgement decisions are consistent so as to ensure a high detection accuracy.After the shared bottleneck detection is completed,the Bottleneck Fairness based congestion control algorithm decouples of the subflows outside the bottleneck set,and realizes the coupled congestion control based on the subflow congestion degree inside the bottle-neck set.Besides,the Bottleneck Fairness based congestion control algorithm realizes load balancing among subflows by dynamically adjusting the CWND(Congestion Win-dow),while ensuring the fairness with single-path TCP within the bottleneck set.Dif-ferent from the previous packet scheduling algorithms that ignore the coupled changes in the subflows' CWND when estimating the number of data packets allocated to each subflow,the new Bottleneck Fairness based packet scheduling algorithm can be more acurrate.Based on the congestion degree of each subflow and the bottleneck set it be-longs to,the new scheduling algorithm models the CWND of each subflow.It improves the accuracy of estimation,and further improves the overall transmission performance.(2)The BBR(Bottleneck Bandwidth and Round-trip propagation time)algorithm has a strong adaptability to the packet loss environment,it can maintain high throughput when random packet loss occurs.In order to improve the performance of MPTCP in wireless heterogeneous network environment,this dissertation proposes a BBR-based multipath transmission scheme.It consists of a new BBR-based congestion control al-gorithm and a new packet scheduling algorithm,which is implemented on the basis of our proposed BBR-based congestion control algorithm.The BBR-based congestion control algorithm adaptively adjusts the sending rate of each subflow according to the real probing rate rather than the loss information,to enhance the goodput while en-suring bottleneck fairness with single-path TCP.Then,considering that the different transmission rates of subflows in heterogeneous wireless networks will lead to out-of-order problems,the new scheduling algorithm uses two phases scheduling to ensure that packets arrive at the receiver in order to reduce completion time:Initially,redundant transmission is realized on different subflows,which can shorten the completion time of latency-sensitive mouse flows through more bandwidth consumption.After reaching a fixed statistical threshold,in the second phase,a forward prediction packet schedul-ing algorithm is run to accurately estimate the completion time of each packet on each subflow based on the values of and RTT obtained by the proposed congestion control algorithm,and schedules the packets to the subflow with the shortest completion time.We evaluate the performance of our proposed scheme through NS-3 and real network tests.The results show that the shceme outperforms existing MPTCP schemes in wire-less wireless heterogeneous environment.(3)Aiming at solving the problem of insufficient bandwidth and unguaranteed de-lay in the transmission of 360-degree VR video,this dissertation proposes an MPTCP-based adaptive streaming scheme for 360-degree Virtual Reality videos.Firstly,the scheme dynamically selects the appropriate bitrate of tile according to the bandwidth and the transmission delay of different subflows.Then,it finely schedules the video segments to different subflows so that the video packets can reach the receiver's play-back buffer on time and in order to improve the user's QoE(Quality of Experience).We verify the performance of our proposed scheme via both the testbed built in the labora-tory and NS-3.Evaluation results show that our proposed scheme outperforms existing tile-based strategies when network fluctuations or errors in FoV predictions occur and can provide more reliable video transmission performance.