Gregory Ganger Professor, Affiliated Faculty Office 2208 Mehrabian Collaborative Innovation Center Email ganger@ece.cmu.edu Phone (412) 268-1297 Department CIT - Electrical and Computer Engineering Computer Science Department: Affiliated Administrative Support Person Karen Lindenfelser Research Interests Systems Data-Intensive and Cloud Computing Distributed Systems Advisees Daiyaan Arfeen Sanjith Athlur Suhas Jayaram Subramanya Timothy Kim Sara McAllister Hojin Park Ziyue Qiu Daniel L.-k. Wong Theo Gregersen CSD Courses Taught 15719 - Spring, 2025 15746 - Fall, 2024 15719 - Spring, 2024 I have broad research interests in computer systems, including cloud computing, storage/file systems, operating systems and distributed systems. I am involved in several ongoing projects in such areas as systems for large-scale ML, cloud/cluster resource scheduling, and exploitation of new storage/NVM technologies. Big-learning systems for Big Data Modern data analytics often relies on statistical machine learning (ML) to parameterize models that fit observation data, for use in making predictions, correlating causes with effects, etc. Growth in data and desired model precision dictate parallel execution of ML algorithms on clusters, with the corresponding work distribution, synchronization, and data consistency challenges. The big-learning group is exploring powerful new approaches for efficient, scalable, and robust big-learning on Big Data. Cloud Computing We are exploring software systems challenges in efficiently supporting and exploiting cloud computing, such as resource allocation/scheduling and exploiting elasticity for stateful services (e.g., storage) and long-running computations (e.g., large-scale ML). Parallel Data Lab (PDL) As Director of the Parallel Data Lab, I lead and collaborate on a number of storage-related projects in areas such as storage system architecture, file systems, and Big Data systems. For example, in addition to the activities discussed above, we are exploring how system software should change to accommodate new storage technologies like non-volatile RAM (e.g., PCM) and best exploit Flash. Publications Conference Baleen: ML Admission & Prefetching for Flash Caches 2024 • PROCEEDINGS OF THE 21ST USENIX SYMPOSIUM ON NETWORKED SYSTEMS DESIGN AND IMPLEMENTATION, NSDI 24 • 347-371 Wong DL-K, Wu H, Molder C, Gunasekar S, Lu J, Khandkar S, Sharma A, Berger DS, Beckmann N, Ganger GR Conference FairyWREN: A Sustainable Cache for Emerging Write-Read-Erase Flash Interfaces 2024 • PROCEEDINGS OF THE 18TH USENIX SYMPOSIUM ON OPERATING SYSTEMS DESIGN AND IMPLEMENTATION, OSDI 2024 • 745-764 McAllister S, Wang YS, Berg B, Berger DS, Amvrosiadis G, Beckmann N, Ganger GR Preprint GraphPipe: Improving Performance and Scalability of DNN Training with Graph Pipeline Parallelism 2024 Jeon B, Wu M, Cao S, Kim S, Park S, Aggarwal N, Unger C, Arfeen D, Liao P, Miao X, Alizadeh M, Ganger GR, Chen T, Jia Z Preprint PipeFill: Using GPUs During Bubbles in Pipeline-parallel LLM Training 2024 Arfeen D, Zhang Z, Fu X, Ganger GR, Wang Y Journal Article Extending and Programming the NVMe I/O Determinism Interface for Flash Arrays 2023 • ACM Transactions on Storage • 19(1): Li H, Putra ML, Shi R, Kurnia FI, Lin X, Do J, Kistijantoro AI, Ganger GR, Gunawi HS
Conference Baleen: ML Admission & Prefetching for Flash Caches 2024 • PROCEEDINGS OF THE 21ST USENIX SYMPOSIUM ON NETWORKED SYSTEMS DESIGN AND IMPLEMENTATION, NSDI 24 • 347-371 Wong DL-K, Wu H, Molder C, Gunasekar S, Lu J, Khandkar S, Sharma A, Berger DS, Beckmann N, Ganger GR
Conference FairyWREN: A Sustainable Cache for Emerging Write-Read-Erase Flash Interfaces 2024 • PROCEEDINGS OF THE 18TH USENIX SYMPOSIUM ON OPERATING SYSTEMS DESIGN AND IMPLEMENTATION, OSDI 2024 • 745-764 McAllister S, Wang YS, Berg B, Berger DS, Amvrosiadis G, Beckmann N, Ganger GR
Preprint GraphPipe: Improving Performance and Scalability of DNN Training with Graph Pipeline Parallelism 2024 Jeon B, Wu M, Cao S, Kim S, Park S, Aggarwal N, Unger C, Arfeen D, Liao P, Miao X, Alizadeh M, Ganger GR, Chen T, Jia Z
Preprint PipeFill: Using GPUs During Bubbles in Pipeline-parallel LLM Training 2024 Arfeen D, Zhang Z, Fu X, Ganger GR, Wang Y
Journal Article Extending and Programming the NVMe I/O Determinism Interface for Flash Arrays 2023 • ACM Transactions on Storage • 19(1): Li H, Putra ML, Shi R, Kurnia FI, Lin X, Do J, Kistijantoro AI, Ganger GR, Gunawi HS
