Biography
I am an associate professor at the Institute of Computing Technology,
Chinese Academy of Sciences since 2021 and an assistant professor from 2019 to 2021.
My research focuses on big data and AI benchmarking, workload characterization, computer architecture, and proxy benchmarks for simulation.
I received my Ph.D. degree in 2019 from Institute of Computing Technology,
Chinese Academy of Sciences, and University of Chinese Academy of Sciences.
Prof. Jianfeng Zhan is my advisor.
I received my B.S. degree in 2012 from Huazhong University of Science and Technology in China. |
Research
My research interests focus on the following points of Data center computing and benchmarking:
- Big data and AI benchmarking.
- Workload Characterization.
- Computer Architecture.
- Profiling and tracing in data centers.
- Proxy benchmarks for simulation.
- ScenarioBench: Modeling, Characterizing, and Optimizing Ultra-scale Real-world or Future Applications and Systems Using Benchmarks. We propose a scenario-distilling methodology that formalizes a real-world or future application scenario as a Directed Acyclic Graph-based model (in short, DAG model). We design and implement a reusing benchmark framework. Based on the framework, we implement the scenario benchmarks---AIBench Scenario, including E-commerce Search Intelligence and Online Translation Intelligence. AIBench Scenario is a benchmarks suite modeling AI-augmented Internet service scenarios, including E-commerce Search Intelligence and Online Translation Intelligence. Each scenario benchmark models the critical paths of a real-world application scenario as a permutation of the AI and non-AI modules.
- AIBench Training: Balanced AI Training Benchmarks. We present a balanced methodology considering comprehensiveness, representativeness, affordability, and portability. Our methodology widely investigates AI tasks and models and covers the algorithm-level, system-level, and microarchitectural-level factors space to the most considerable extent. We provide nineteen real-world AI workloads to achieve comprehensiveness and representativeness. Besides, we propose two AIBench Training subset: RPR and WC subsets to achieve affordability. We give the hotspot functions as microbenchmarks (AIBench Micro) to achieve portability for simulator-based research after profiling.
- AIBench Inference: Comprehensive AI Inference Benchmarks. We present a comprehensive AI inference benchmark suite including nineteen real-world AI workloads and covering text processing, image processing, audio processing, video processing, and 3D data processing.
- HPC AI500: A Benchmark Suite for HPC AI Systems. HPC AI500 presents a comprehensive methodology, tools, Roofline performance models, and innovative metrics for benchmarking, optimizing, and ranking HPC AI systems. We abstract the HPC AI system into nine independent layers, and present explicit benchmarking rules and procedures to assure equivalence of each layer, repeatability, and replicability.
- Data Motifs: A Lens towards Fully Understanding Big Data and AI Workloads. Identifying abstractions of time-consuming units of computation is an important step toward domain-specific hardware and software co-design. Straightforwardly, we can tailor the architecture to characteristics of an application, several applications, or even a domain of applications. Data motif is a new approach to modelling and characterizing big data and AI workloads. We consider each big data and AI workload as a pipeline of one or more classes of unit of computation performed on different initial or intermediate data inputs, each of which captures the common requirements while being reasonably divorced from individual implementations. We call this abstraction a data motif. Significantly different from the traditional kernels, a data motif’s behaviors are affected by the sizes, patterns, types, and sources of different data inputs; Moreover, it reflects not only computation patterns, memory access patterns, but also disk and network I/O patterns.
- BigDataBench 4.0 project. BigDataBench 4.0 adopts a scalable data motif based benchmarking methodology and contains 13 representative real-world data sets and 47 benchmarks. The latest version of BigDataBench is available from http://www.benchcouncil.org/BigDataBench/index.html.
- Proxy Benchmark Generating for Big Data and AI. We propose a data motif-based proxy benchmark generating methodology which combines data motifs with different weights to mimic the big data and AI workloads. The proxy benchmarks shorten the execution time by 100s times on real systems while maintaining the average system and micro-architecture performance data accuracy above 90%, even changing the input data sets or cluster configurations. Moreover, the generated proxy benchmarks reflect consistent performance trends across different architectures.
- High Throughput Computers project, Institute of Computing Technology, Chinese Academy of Sciences, Beijing. System and micro-architectural level analysis of high throughput workloads.
Publications
-
AIBench Scenario: Scenario-distilling AI Benchmarking. [pdf].The 30th International Conference on Parallel Architectures and Compilation Techniques (PACT 2021).
-
AIBench Training: Balanced Industry-Standard AI Training Benchmarking. [pdf].2021 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS 2021).
-
HPC AI500 V2.0: The Methodology, Tools, and Metrics for Benchmarking HPC AI Systems. [pdf].2021 IEEE International Conference on Cluster Computing (CLUSTER 2021).
-
WPC: Whole-Picture Workload Characterization Across Intermediate Representation, ISA, and Microarchitecture. [pdf].IEEE Computer Architecture Letters (2021).
-
AI-oriented Workload Allocation for Cloud-Edge Computing. [pdf].The 21th IEEE/ACM International Symposium on Cluster, Cloud and Internet Computing (CCGrid 2021).
-
AIBench: Towards Scalable and Comprehensive Datacenter AI Benchmarking. [pdf].2018 BenchCouncil International Symposium on Benchmarking, Measuring and Optimizing (Bench18).
-
Data Motifs: A Lens Towards Fully Understanding Big Data and AI Workloads. [pdf].The 27th International Conference on Parallel Architectures and Compilation Techniques (PACT18).
-
Data Motif-based Proxy Benchmarks for Big Data and AI Workloads. [pdf].IEEE International Symposium on Workload Characterization (IISWC 2018).
-
BigDataBench: a Data Motif-based Big Data and AI Benchmark Suite [pdf].arXiv preprint arXiv:1802.08254.
-
BigDataBench: a Big Data Benchmark Suite from Web Search Engines [pdf].Third Workshop on Architectures and Systems for Big Data (ASBD 2013) in conjunction with The 40th International Symposium on Computer Architecture, May 2013.
-
BigDataBench:An Open Source Big Data Benchmark Suite [pdf].CHINESE JOURNAL OF COMPUTERS, 2016.
-
Landscape of big medical data: a pragmatic survey on prioritized tasks. [pdf].IEEE Access, 2019.
-
Understanding Processors Design Decisions for Data Analytics in Homogeneous Data Centers. [pdf].IEEE Transactions on Big Data, 2017.
-
CVR: Efficient Vectorization of SpMV on X86 Processors. [pdf].CGO 2018.
-
BigDataBench: a Big Data Benchmark Suite from Internet Services. [pdf].HPCA 2014, Industry Session.
-
Understanding Big Data Analytics Workloads on Modern Processors. [pdf].TPDS 2016.
-
Trends on Methods for Prediction of Tandem Mass Spectra of Peptides. [pdf].Progress in Biochemistry and Biophysics, 2018.
Invited Talks & Tutorials
Invited Talks
- AIBench Scenario: Scenario-distilling AI Benchmarking
Keynote on Benchmarking in the Data Center: Expanding to the Cloud workshop held in conjunction with PPoPP 2021. - BenchCouncil AI and Big Data Benchmarks
The European Big Data Value Forum (EBDVF 2020). - Data Motif: A Benchmark Proposal for Big Data and AI
2018 BenchCouncil International Symposium on Benchmarking, Measuring and Optimizing (Bench18). - BigDataBench: A Dwarf-based Big Data and AI Benchmark Suite
BPOE-9: The ninth workshop on Big data benchmarks, Performance, Optimization, and Emerging hardware, in conjunction with Architectural Support for Programming Languages and Operating Systems (ASPLOS 2018) - Big Data Dwarfs: Methodology, Dwarf Library and Simulation Benchmarks
BPOE-8: The eighth workshop on Big data benchmarks, Performance, Optimization, and Emerging hardware.
- AIBench Tutorial in conjunction with The International Symposium on Computer Architecture (ISCA 2021)
- AIBench Tutorial in conjunction with Architectural Support for Programming Languages and Operating Systems (ASPLOS 2021)
- AIBench Tutorial in conjunction with Architectural Support for Programming Languages and Operating Systems (ASPLOS 2020)
- AIBench Tutorial in conjunction with High Performance Computer Architecture (HPCA 2020)
- BigDataBench Tutorial on 2018 BenchCouncil International Symposium on Benchmarking, Measuring and Optimizing (Bench'18)
- BigDataBench Tutorial in conjunction with Architectural Support for Programming Languages and Operating Systems (ASPLOS 2018).
- High Volume Computing: The Motivations, Metrics, and Benchmarks Suites for Data Center Computer Systems.
in conjunction with the 19th IEEE International Symposium on High Performance Computer Architecture (HPCA 2013)
Professional Activities
- General Co-Chair of Bench 2022
- Performance Committee Member of SC22
- Poster Co-chair of IEEE Cluster 2022
- Performance Committee Member of ICPP 2021
- Technical Program Committee Member of Bench 2021
- Program Co-Chair of Bench 2020
- Program Co-Chair of HotDC 2020
- AI Challenge Co-Chair & Publicity Co-Chair of Bench 2019
- Competition Committee Co-chair of 2019 International Symposium on Intelligent Computers
- Publicity Co-Chair of Bench 2018
- Web and Publicity Co-Chair of BPOE-9
- Web Chair of SDBA'18
- Web and Publicity Chair of BPOE-8
Honors
- 2021.11 2021 BenchCouncil Rising Star Award
- 2019.1 Outstanding Graduates of UNIVERSITY OF CHINESE ACADEMY OF SCIENCE
- 2018.6 Merit Student of UNIVERSITY OF CHINESE ACADEMY OF SCIENCE
- 2017.10 National Scholarship for Ph.D.
- 2015.12 Doctoral scholarship, ICT, CAS, Beijing