2018/11/11 by Walid Aljoby, Aljoby, Walid, Xin Wang +5
Computer Science · #Cloud Computing and Resource Management #FOS: Computer and information sciences #Networking and Internet Architecture (cs.NI) #Software System Performance and Reliability #Software-Defined Networks and 5G
paper · pdf · doi:10.48550/arxiv.1811.04377
openalex publication_date 2018/11/11 · openalex created_date 2022/08/02 · openalex updated_date 2026/07/28
Data communication in cloud-based distributed stream data analytics often\ninvolves a collection of parallel and pipelined TCP flows. As the standard TCP\ncongestion control mechanism is designed for achieving "fairness" among\ncompeting flows and is agnostic to the application layer contexts, the\nbandwidth allocation among a set of TCP flows traversing bottleneck links often\nleads to sub-optimal application-layer performance measures, e.g., stream\nprocessing throughput or average tuple complete latency.\n Motivated by this and enabled by the rapid development of the\nSoftware-Defined Networking (SDN) techniques, in this paper, we re-investigate\nthe design space of the bandwidth allocation problem and propose a cross-layer\nframework which utilizes the additional information obtained from the\napplication layer and provides on-the-fly and dynamic bandwidth adjustment\nalgorithms for helping the stream analytics applications achieving better\nperformance during the runtime.\n We implement a prototype cross-layer bandwidth allocation framework based on\na popular open-source distributed stream processing platform, Apache Storm,\ntogether with the OpenDaylight controller, and carry out extensive experiments\nwith real-world analytical workloads on top of a local cluster consisting of 10\nworkstations interconnected by a SDN-enabled switch. The experiment results\nclearly validate the effectiveness and efficiency of our proposed framework and\nalgorithms.\n