11 February 2022 in The Journal of Supercomputing
24 October 2024 in IEEE Transactions on Parallel and Distributed Systems
It is challenging to scale parallel applications to modern supercomputers because of load imbalance, resource contention, and communications between processes. Profiling and tracing are two main performance analysis approaches for detecting these scalability bottlenecks. Profiling is low-cost but lacks detailed dependence for identifying root causes. Tracing records plentiful information but incurs significant overheads. To address these issues, we present ScalAna, which employs static analysis techniques to combine the benefits of profiling and tracing - it enables tracing's analyzability with overhead similar to profiling. ScalAna uses static analysis to capture program structures and data dependence of parallel applications, and leverages lightweight profiling approaches to record performance data during runtime. Then a parallel performance graph is generated with both static and dynamic data. Based on this graph, we design a backtracking detection approach to automatically pinpoint the root causes of scaling issues. We evaluate the efficacy and efficiency of ScalAna using several real applications with up to 704K lines of code and demonstrate that our approach can effectively pinpoint the root causes of scaling loss with an average overhead of 5.65% for up to 16,384 processes. By fixing the root causes detected by our tool, it achieves up to 33.01% performance improvement.
1 October 2019 in IEEE Access
Nowadays, the ocean numerical models are gradually developing towards multi-physical process and high resolution, with the increment of measured ocean data and more in-depth research in ocean field. Therefore, general computing capability is no longer able to meet these models’ needs. It is necessary to utilize more powerful hardware and parallel software to process the ocean numerical model programs. China has made great process in the research and development of homegrown high performance processors, and sunway sw26010 many-core processor is the most outstanding representative. This paper focuses the lag of the ocean numerical model software matched with homegrown processors, and makes parallel implementation and optimization to regional ocean modeling system (ROMS) based on sunway sw26010 many-core processor for the first time. Furthermore, three kinds of programming methods are utilized in this paper, including OpenACC*, athread with fortran and athread with C. The comparison among these programming methods has been made, from programming method, workload and execution efficiency, which has a practical guiding significance for the programmers that use sunway sw26010 many-core processors. The evaluation measures the execution times and speedups of model kernel and total ROMS with different optimizations, input datasets and numbers of computing processing elements (CPEs). The result shows that, to compare with original ROMS, the speedup of optimized hotspot program can be up to 3.69x.
14 October 2021 in IEEE Transactions on Parallel and Distributed Systems
Achieving microsecond-scale tail latency poses an extreme challenge to the conventional architecture of “NIC-OS-Application” in the face of high concurrent requests. Existing kernel-bypass network systems improve this situation significantly. Still, they cannot achieve load-aware in-server requests distribution, which in turn not only harms resource efficiency but, more importantly, beats the goal of squeezing tail latency. This paper proposes iBalancer, an in-server proactive load balancer for the kernel-bypass system, which aggressively handles NIC-side flow scheduling according to the load of threads on the processor-side. Furthermore, we propose a novel metric, “polling time interval (PTI),” to quantify the load of worker threads, which not only indicates utilization of the core bound to the worker thread but also reflects the differences in the processing time of different flows. By scheduling flows according to the metric PTI, iBalancer tends to average the queueing latencies of different flows, such as Set & Get operations for an in-memory key-value store. In addition, by decoupling flow scheduling from packet steering, iBalancer achieves a tail latency aware flow-to-core binding and preserves hardware-based request distribution among cores. The proposed system is evaluated and compared to mTCP and Shenango using two representative microsecond-scale network applications: Memcached KVS and a real-time deep-learning-based financial fraud identification application. Experimental results show that iBalancer can process up to 4.75 $ \times $ and 1.55 $ \times \ $ higher load over mTCP and Shenango under 500μs 99 th percentile tail latency limit on Memcached. For the financial fraud identification application, iBalancer is able to process 4.56 $ \times $ and 1.16 $ \times $ higher load than mTCP and Shenango considering 900μs tail latency.
26 April 2023 in The Journal of Supercomputing