Sidharth Kumar

Sidharth Kumar Assistant Professor
email
Campbell Hall 126
(205) 934-0650

Research and Teaching Interests: High performance computing, visualization, machine learning, MPI

Office Hours: By appointment

Education:

  • Ph.D., Computing, University of Utah
  • B.Tech, Information and communication technology, DAIICT, Gandhinagar, India

As an undergraduate at DAIICT, I was fascinated by computer graphics, ultimately leading me to join The University of Utah. While at the U, I spent a summer at Argonne National Lab, where I was exposed to the world of supercomputing.

Personal Website

I was thrilled by the idea of using supercomputers to solve computationally massive problems. Ever since, I have worked at the intersection of HPC, analytics and visualization, helping scientists, engineers, and decision makers extract knowledge from massive amounts of complex data.

 

Sidharth’s primary area of research is high performance computing, focusing on analysis and visualization of large scale data. In particular, he has interests in bringing interactivity to data-intensive processing tasks, requiring effective use of a range of techniques encompassing analytics, high performance computing, visualization and machine learning. More broadly, his research interests include large data processing, parallel I/O, storage and file systems, in-situ analytics and visualization, scalable algorithms, large data file formats, progressive processing, scientific data, interactive techniques, and scientific visualization.

Sidharth’s research provides a unified solution to facilitate key requirements of large-scale data management including parallel I/O, out-of-core processing, data streaming and remote visualization. He is also keen about using machine learning techniques for performance modeling and autotuning of HPC and big data systems. He has publications at top venues in data management and HPC. He has also scaled his data movement framework to 768K cores and have deployed it on some of the fastest supercomputers of the world (Mira, Titan, Edison, Shaheen).

  • Sidharth Kumar, Duong Hoang, Steve Petruzza, John Edwards, Valerio Pascucci. Reducing network congestion and global communication bottlenecks during aggregation on Torus and Dragonfly topologies for writing hierarchical data. IEEE Conference On High Performance Computing, Data, and Analytics. HiPC 2017, acceptance rate - 23% (42/184).
  • Aaditya Landge, Ivan Rodero, Sidharth Kumar, Manish Parasar, Valerio Pascucci, Peer-T. Bremer. Evaluation of In-Situ Analysis Strategies at Scale for Power Efficiency and Scalability. IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing. CCGRID 2016, acceptance rate - 20% (40/200)
  • Sidharth Kumar, John Edwards, Peer-T. Bremer, Aaron Knoll, Cameron Christensen, Venkatram Vishwanath, Phil Carns, John Schmidt, Valerio Pascucci. Efficient I/O and storage of adaptive resolution data. IEEE Conference On High Performance Computing Networking, Storage And Analysis. SC 2014, acceptance rate - 20% (82/394)
  • Sidharth Kumar, Cameron Christensen, John Schmidt, Peer-T. Bremer, Eric Brugger, Venkatram Vishwanath, Philip Carns, Giorgio Scorzelli, Hemanth Kolla, Ray Grout, Jacqueline Chen, and Valerio Pascucci. Fast multi-resolution reads of massive simulation datasets. The International Supercomputing Conference. ISC 2014, acceptance rate - 50%
  • Sidharth Kumar, Avishek Saha, Venkatram Vishwanath, Philip Carns, John Schmidt, Giorgio Scorzelli, Hemanth Kolla, Ray Grout, Robert Latham, Robert Ross, Michael E. Papka, Jacqueline Chen, and Valerio Pascucci. Characterization and modeling of PIDX parallel I/O for performance optimization. IEEE Conference On High Performance Computing Networking, Storage And Analysis. SC 2013, acceptance rate - 20% (92/457)
  • Sidharth Kumar, Venkatram Vishwanath, Phil Carns, Joshua A. Levine, Robert Latham, Giorgio Scorzelli, Robert Ross, Hemanth Kolla, Ray Grout, Jackie Chen, Michael E. Papka, Valerio Pascucci. Efficient data restructuring and aggregation for I/O acceleration in PIDX. IEEE Conference On High Performance Computing Networking, Storage And Analysis. SC 2012, acceptance rate - 21% (100/472)
  • Sidharth Kumar, Venkatram Vishwanath, Phil Carns, Brian Summa, Giorgio Scorzelli, Valerio Pascucci, Robert Ross, Jackie Chen, Hemanth Kolla. PIDX: efficient parallel I/O for multi-resolution multidimensional scientific datasets. IEEE International Conference on Cluster Computing. Cluster 2011, acceptance rate - 27% (39/140)