29 March 2018 in IEEE Access
Satellite networks are multilayered, and the number of satellites in a single constellation is also increasing. These characteristics make satellite network more suitable for multipath transmission like multipath transmission control protocol (MPTCP). With the use of MPTCP, the bandwidth of different satellite channels can be aggregated, and the mobility performance of users can be improved. Furthermore, software defined networking (SDN) is introduced to the MPTCP to solve the shared bottleneck problem. However, the performance of existing scheme is still far from optimal. The main problems include: (1) static number setting of MPTCP subflows on a per host basis and (2) unaware of the traffic load during the subflow route selection. The shared bottleneck problem is more serious in the satellite network with lattice-like topology. To solve the above problems, we propose an sdn cooperated MPTCP (scMPTCP) architecture and its related algorithms. We extend TCP options to piggyback the relevant control information to flexibly support the communication between the subflows of transport layer and SDN controller. Then, we propose a load and shared bottleneck aware subflow route selection algorithm and adjust algorithm. These two algorithms select routes for new subflows based on the available bandwidth of each route and avoid the bottleneck of other subflows, and also can adapt to the changes of network load. We implement the scMPTCP and its algorithms. The evaluation results show that compared with searching over the non-overlapping paths or shortest paths schemes, our scheme can achieve much higher total system throughput. Moreover, by adjusting the subflows which share bottlenecks, the total aggregated throughput of that connection is also improved greatly.
25 November 2024 in Proceedings of the ACM on Networking
Together with the construction of RDMA networks for data center applications, the RDMA-coupled DCQCN dominates the RDMA Congestion Control (CC). However, DCQCN suffers severe performance problems in high-speed RDMA networks with modern high-performance distributed applications such as machine learning training. This paper presents RECC, inspired by both the latest emerging programmability of RDMA NICs (RNICs) and limitations in existing RDMA congestion control mechanisms. RECC comprehensively leverages RTT and ECN events from RNICs to handle congestion timely and precisely, along with a History-aware Burst Smooth mechanism to avoid wrong rate decisions under various traffic patterns. We implement RECC completely based on commercial RNICs without any modifications to switches, RDMA protocol stack, and applications. The results of microbenchmark testbed experiments and real Machine Learning (ML) workload experiments with hundreds of 200G RNICs show that RECC can significantly reduce network tail latency and pause duration by up to 64.4% and 95%, respectively, compared with DCQCN. In addition, large-scale simulations with realistic workloads demonstrate that RECC achieves comparable performance with HPCC.