We developed a PYNQ cluster that consists of economical Zynq boards, called M-KUBOS, that are interconnected through low-cost high-performance GTH serial links. For the software environment, we employed the PYNQ open-source software platform. The PYNQ cluster is anticipated to be a multi-access edge computing (MEC) server for 5G mobile networks. We implemented the ResNet-50 inference accelerator on the PYNQ cluster for image recognition of MEC applications. By estimating the execution time of each ResNet-50 layer, layers of ResNet-50 were divided into multiple boards so that the execution time of each board would be as equal as possible for efficient pipeline processing. Owing to the PYNQ cluster in which FPGAs were directly connected by high-speed serial links, stream processing without network bottlenecks and pipeline processing between boards were readily realized. The implementation on 4 boards achieved 292 GOPS performance, 75.1 FPS throughput, and 7.81 GOPS/W power efficiency. It achieved 17 times faster speed and 130 times more power efficiency compared to the implementation on the CPU, and 5.8 times more power efficiency compared to the implementation on the GPU.

抄録全体を表示

In this work, we propose the Network Performance Collector (NPC) workflow for automated network performance characterization. The workflow is based on the collection, processing as well as visualization of network performance metrics such as throughput and latency and can be used for analysis with different network performance models. Depending on the chosen model, benchmark tools such as iperf or sockperf as well as microbenchmarks specific to parallel programming models can be automated and orchestrated for data collection with the NPC. The data obtained can then be used by NPC, for example, to validate and characterize the performance of the underlying network or to analyze the system boundaries for a particular application. We provide a prototype implementation of the proposed workflow and demonstrate its effectiveness by automating an extended Roofline model analysis.

抄録全体を表示