Feng Li

Feng Li earned his PhD degree in Computer Science from Purdue University under the direction of Dr. Fengguang Song . His research interests are in areas of Distributed Systems, High Performance Computing, and scientific workflows.


  • [2023-07-15] Feng successfully defended his dissertation work “Efficient in-situ workflows for time-critical applications on heterogeneous ecosystems”.

  • [2023-07-07] “Efficient In-situ Workflow Planning for Geographically Distributed Heterogeneous Environments” is accepted as a journal paper in Future Generation Computer Systems, Elsevier.

  • [2023-04-29] “INSTANT: a runtime framework to orchestrate in-situ workflows” is accepted in the Proceedings of the 27th International European Conference on Parallel and Distributed Computing (EuroPar’23).

  • [2023-03-06] Feng started a new position at Eli Lilly.

Contact Me

You can reach me through li2251 at purdue dot edu.


  • [2015-2023] Purdue University, IN

    PhD in Computer Science. Dissertation: Efficient in-situ workflows for time-critical applications on heterogeneous ecosystems

  • [2011~2015] Huazhong University of Science & Technology, Wuhan, China

    Bachelor in Computer Science and Technology.

Research Projects

Computer Science Department at Purdue University

Jan 2020 ~ March 2023, cross-environment in-situ workflow management

_images/xcomposer-highlevel.jpeg _images/xcomposer-wind.png
  • Design, and prototype a framework that can be used to launch in-situ workflows across HPC and Cloud systems.

  • Formalize the scheduling problem for in-situ workflows, design and evaluate heuristic-based algorithms to improve workflow metrics such as throughput and latency.

  • A paper was accepted by PASC’21.

  • A paper about resource planning for in-situ workflow resource planning.

  • A paper about generic runtime support for in-situ workflows across sites.

Jan 2020 ~ Aug 2022, on demand HPC/Cloud access for CyberWater project

_images/cyberwater-launchagent.jpeg _images/launchagent-workflow.png
  • Design/test/improve the ‘on demand’ HPC/Cloud acceleration feature for Cyberwater project – a community-driven, multi-institutional earth-science project.

  • Enhance local development environment with Airavata SciGap based gateway APIs, so that computationally expensive/data-intensive operations are offloaded to XSEDE HPCs and public Clouds.

  • Test, refine, and optimize hydrology simulation models in modern HPC environments.

  • A paper was accepted by e-science’21.

Jan 2019~ Dec 2019, Architecture-aware Neural Network

Optimize Neural Networks from the system perspective:

  • Co-locate data in deep/heterogeneous memory hierarchy, so that communication overhead between different components can be minimized.

  • Intelligent scheduling algorithm for computation kernels to enable efficient and large-scale parallelism in modern SIMD and NUMA architectures.

  • github repo.

Nov 2017~ Mar 2018, Memory-aware Lattice-Boltzmann Method

Prototype and optimize a memory-aware Lattice-Boltzmann Method (LBM, a computational fluid dynamics approach to simulating complex fluid flows), which can enhance data reuse across multiple time steps.

  • A paper was accepted by SBAC-PAD’18.

April 2017~ Nov 2017, Performance Analysis of In-situ Methods in HPC

In-depth performance evaluation of various high-performance transport methods (ADIOS, flexpath, DataSpaces, Decaf, DIMEs) in multiple HPC systems.

_images/hpdc_perf_comp.png _images/hpdc_dimes_traces.png

Aug 2016 ~ Mar 2017, Machine Learning Workflow in HPC

Working on s software framework for scientific workflows where RDMA technique is used to couple numerical simulation, data analysis and real-time visualization application together.

  • A distributed and optimized anomaly detection method is used to detect vortex and other special patterns from turbulence flows.

  • A paper was accepted by PEARC’17 (Best Student Paper Award).

Aug 2015 ~ Apr 2016, KSSR

Working on a kernelized sparse self-representation model(KSSR) and a novel Kernelized Fast Iterative Soft-Thresholding Algorithm(K-FISTA), to recover the underlying nonlinear structure among data.

  • My work mainly includes the implementation, evaluation of the KSSR model and K-FISTA algorithm.

  • A paper was accepted in SDM’16.

Storage system group at IBM Research, Almaden

May 2019~ Aug 2019, research internship

  • Designed and implemented a unified file system interface (KVFS) for multiple key-value store backends, so that file operations are translated into key-value store put/get operations.

  • Used FUSE to implement KVFS, and designed mechanisms to handle the mappings between file abstractions and data objects.

  • Code base in IBM Comanche: https://github.com/IBM/comanche/tree/unstable/src/fuse.

May 2018~ Aug 2018, research internship

  • Worked on a high-performance key-value store, which uses NVMe SSD as data storage and keeps critical metadata in the persistent memory (pmem).

  • By utilizing the advantages of persistent memory, the access to metadata such as block allocation and key-value mappings can be both hardened and fast.

May 2017~ Aug 2017, research internship

  • Designed and implemented an NVMe-backed light-weight memory service — CO-PAGER (Collaborative Paging).

  • CO-PAGER captures virtual memory page faults and performs paging operations on NVMe SSDs using fast userspace I/O.

  • A paper is accepted in HP3C’19.


I use those tools intensively in my research workflow:

  • Programming languages (C/C++/Java/Scala/Python)

  • Performance analysis tools (Intel Vtune, Linux Perf, TAU)

  • Mathematical-modelling tool (CPLEX)

  • Big Data/Deep learning frameworks (Tensorflow/Pytorch/Apache Spark)

  • Storage related: redis, spdk, dpdk, pmdk, fuse, fio

  • Cloud/container solutions(Google Cloud Platform, Amazon AWS, Openstack, Docker, k8s)

  • CI/Build tools(cmake, Apache Maven, travis CI, Google Gtest)


  • Feng Li, and Fengguang Song. 2023. INSTANT: A Runtime Framework to Orchestrate In-Situ Workflows. To appear in Proceedings of the 27th International European Conference on Parallel and Distributed Computing (EuroPar’23), Springer.


  • My spare time

I like hiking, jogging. I have travelled to many great places in China, and I am planning to visit more places in the US! In my spare time I like playing pingpong or tennis with friends.

  • TODO

I should probably do some tech blogs..