akougkas.io

[ io = input/output ]

avatar_profile.webp
Anthony Kougkas

assistant research professor


data management & storage

distributed & parallel systems

input/output for hpc & ai/ml



illinois tech

chicago, il

email

google scholar

github

linkedIn

i am an assistant research professor in computer science at illinois tech, where i serve as the deputy director of the gnosis research center. my work at illinois tech and as a guest research faculty at argonne national laboratory involves developing cutting-edge data management and storage solutions. with a strong focus on high-performance computing (hpc) and artificial intelligence (ai), i lead innovative projects and guide the next generation of computer scientists. always curious and ready to explore.

*for collaboration opportunities, please reach out via email or linkedin

interests

hpc storage & i/o
  • multi-tiered storage architectures, hierarchical data buffering & prefetching, distributed & parallel i/o optimization.
  • asynchronous i/o techniques, i/o scheduling strategies, bottleneck detection & optimization, efficient resource utilization.
data management for workflows
  • workflow execution optimizations, tiered data streaming engines, context-aware active storage, data prefetching algorithms.
  • data management for large datasets, prevention of i/o interference, advanced data compression techniques.
systems for ai & scientific ml
  • advanced data management for ai/ml frameworks, ai-driven i/o optimizations, exascale-ready storage.
  • integration of ai/ml with hpc systems, context-aware active storage, task-driven frameworks & data labeling.

projects

NEW iowarp: a state-of-the-art i/o buffering platform for hpc environments, utilizing hdf5 data structures to optimize data placement across memory and storage hierarchies, including nvram and ssds.

hermes: a state-of-the-art i/o buffering platform for hpc environments, utilizing hdf5 data structures to optimize data placement across memory and storage hierarchies, including nvram and ssds.

iris: a unified data access framework merging hpc and analytics and ai, enabling integrated data and metadata management, and intelligent data placement.

chronolog: a high-performance, distributed log storage system that manages activity and log workloads using physical time for event ordering, reducing contention, and enabling elastic scaling across storage tiers. supports plugins, sql-like engines, and tensorflow.

coeus: a framework that accelerates scientific insights through enriched metadata management, optimizing queries by leveraging ai/ml to balance computation and storage, and enhancing query performance with tiered data placement and staging.

dtio: a task-driven i/o framework for the hyperconvergence of hpc, ai, and cloud, offering scalable, distributed i/o optimization, enhanced data movement, and improved fault tolerance.

labios: an innovative, energy-efficient label-based i/o system for multi-tiered hpc environments, providing solutions such as multi-tenancy, resource scheduling, and seamless integration with diverse storage pools.

deepio: a scalable i/o runtime for ai workflows, optimizing dnn model updates between training and inference tasks. it balances inference quality, training duration, and throughput using novel caching, versioning, and asynchronous data transfer techniques.

contributions

publications: advancing the fields of HPC data management and storage with over 50 peer-reviewed publications.

patents: US patent for the “Label-Based Data Representation I/O Process and System”, US 2021/0374152 A1.

funding: strong track record in funding, successfully securing over $10M of federal grants for my research.

mentoring: co-advising ten doctoral students in data streaming, AI-powered I/O, and programmable storage.

recognition: Best Paper Awards at CCGrid’21 for DLIO and at HPDC’19 for LABIOS papers.

software: deployed open-source software like Hermes, ChronoLog, and DLIO (now in Apache MLPerf suite).

teaching: active learning, engaging teaching environment, emphasizing hands-on, lab-focused experiences.

selected publications

  1. LabStor: A modular and extensible platform for developing high-performance, customized I/O stacks in userspace
    Luke Logan, Jaime Cernuda Garcia , Jay Lofstead , Xian–He Sun, and Anthony Kougkas
    In International Conference for High Performance Computing, Networking, Storage and Analysis , Nov 2022
  2. DLIO: A data-centric benchmark for scientific deep learning applications
    Hariharan Devarajan , Huihuo Zheng, Anthony Kougkas , Xian-He Sun , and Venkatram Vishwanath
    In 21st International Symposium on Cluster, Cloud and Internet ComputingBest Paper Award ║ , May 2021
  3. Labios: A distributed label-based i/o system
    Anthony Kougkas , Hariharan Devarajan , Jay Lofstead , and Xian-He Sun
    In 28th International Symposium on High-Performance Parallel and Distributed Computing.Best Paper Award ║ , Jun 2019
  4. Hermes: a heterogeneous-aware multi-tiered distributed I/O buffering system
    Anthony Kougkas , Hariharan Devarajan , and Xian-He Sun
    In 27th International Symposium on High-Performance Parallel and Distributed Computing , Jun 2018
  5. Leveraging burst buffer coordination to prevent I/O interference
    Anthony Kougkas , Matthieu Dorier , Rob Latham , Rob Ross , and Xian-He Sun
    In 12th International Conference on e-Science , Jun 2016