Maxime GONTHIER

Contact and information

I am currently working as a Postdoc at the Globus Labs (University of Chicago) and at the Argonne National Laboratory. My work is under the supervision of Ian Foster and Kyle Chard.

I completed my PhD at ENS Lyon in the field of Computer Science. I worked in the ROMA team at the LIP laboratory under the supervision of Loris Marchal (ENS Lyon, CNRS) and in the STORM team at INRIA Bordeaux with the supervision of Samuel Thibault (Université de Bordeaux, Inria).

Email:

mgonthier (at) uchicago (dot) edu

Research interests

My research focuses on scheduling problems. I study more specifically the issue of scheduling task sharing data under memory constraint. To solve this problem I create algorithms and implement them in the StarPU runtime.

I also had the chance to collaborate with the Division of Scientific Computing at Uppsala University. During this collaboration I worked on the subject of batch scheduling of jobs needing large input files.

PhD defense

My PhD defense took place the September 25^th, 2023 in the LaBRI in Bordeaux, France. The subject is Scheduling Under Memory Constraint in Task-based Runtime Systems.
Abstract: Hardware accelerators, such as GPUs, now provide a large part of the computational power used for scientific simulations. GPUs come with their own limited memory and are connected to the main memory of the machine via a bus with limited bandwidth. Scientific simulations often operate on very large data, to the point of not fitting in the limited GPU memory. In this case, one has to turn to out-of-core computing, where data movement quickly becomes a performance bottleneck. During this thesis, we worked on the problem of scheduling for a task-based runtime to improve data locality in an out-of-core setting, in order to reduce data movements. We designed strategies for both task scheduling and data eviction from limited memories. We implemented them in the StarPU runtime and compared them to existing scheduling techniques. Our strategies achieves significantly better performance when scheduling tasks on multiple GPUs with limited memory, as well as on multiple CPU cores with limited main memory.

My PhD manuscrit is available here. A capture of my PhD defense is available:

Publications

Authors	Link	Year	Conference/Journal	Type of publication
				Peer reviewed conference
J. Gregory PAULOSKI, Valerie HAYOT-SASSON, Maxime GONTHIER, Nathaniel HUDSON, Haochen PAN, Sicheng ZHOU, Ian FOSTER, Kyle CHARD	TaPS: A Performance Evaluation Suite for Task-based Execution Frameworks	2024	20th IEEE International Conference on e-Science	Peer reviewed conference
André BAUER, Maxime GONTHIER, Ryan CHARD, Haochen PAN, Daniel GRZENDA, Martin STRAESSER, J. Gregory PAULOSKI, Alok KAMATAR, Matthew E. BAUGHMAN, Nathaniel HUDSON, Ian FOSTER, Kyle CHARD	An Empirical Investigation of Container Building Strategies and Warm Times to Reduce Cold Starts in Scientific Computing Serverless Functions	2024	20th IEEE International Conference on e-Science	Peer reviewed conference
Maxime GONTHIER, Elisabeth LARSSON, Loris MARCHAL, Carl NETTELBLAD, Samuel THIBAULT	Data-Driven Locality-Aware Batch Scheduling	2024	APDCM 2024 - 26th Workshop on Advances in Parallel and Distributed Computational Models, 38th IEEE International Parallel and Distributed Processing Symposium	Peer reviewed conference
Bogdan NICOLAE, Justin M. WOZNIAK, Tekin BICER, Hai NGUYEN, Parth PATEL, Haochen PAN, Amal GUERROUDJI, Maxime GONTHIER, Valerie HAYOT-SASSON, Eliu HUERTZ, Kyle CHARD, Ryan CHARD, Matthieu DORIER, Nageswara S. V. RAO, Anees AL-NAJJAR, Alessandra CORSI, Ian FOSTER	Diaspora: Resilience-Enabling Services for Real-Time Distributed Workflows	2024	The 1st international workshop on Near Real-time Data Processing for Interconnected Scientific Instruments	Peer reviewed conference
Haochen PAN, Ryan CHART, Sicheng ZHOU, Alok KAMATAR, Rafael VESCOVI, Valérie HAYOT-SASSON, André BAUER, Maxime GONTHIER, Kyle CHARD, Ian FOSTER	Octopus: Experiences with a Hybrid Event-Driven Architecture for Distributed Scientific Computing	2024	Fault Tolerance for HPC at eXtreme Scales (FTXS) Workshop	Peer reviewed conference
Maxime GONTHIER	Exploiting data locality to maximize the performance of data-sharing tasksets	2023	ComPAS - Conférence francophone d’informatique en Parallélisme, Architecture and Système,	National conference
Maxime GONTHIER, Loris MARCHAL, Samuel THIBAULT	Taming data locality for task scheduling under memory constraint in runtime systems	2023	Future Generation Computer Systems (FGCS)	Journal
Maxime GONTHIER, Loris MARCHAL, Samuel THIBAULT	A generic scheduler to foster data locality for GPU and out-of-core task-based applications	2023	~	Research report
Maxime GONTHIER, Loris MARCHAL, Samuel THIBAULT	Memory-Aware Scheduling Of Tasks Sharing Data On Multiple GPUs	2023	ISC	Poster
Maxime GONTHIER, Loris MARCHAL, Samuel THIBAULT, Elisabeth LARSSON, Carl NETTELBLAD	Locality-aware batch scheduling of I/O intensive workloads	2023	~	Research report
Maxime GONTHIER, Loris MARCHAL, Samuel THIBAULT	Taming data locality for task scheduling under memory constraint in runtime systems	2023	Future Generation Computer Systems	Journal
Maxime GONTHIER, Loris MARCHAL, Samuel THIBAULT	Locality-Aware Scheduling of Independant Tasks for Runtime Systems	2022	~	Research report
Maxime GONTHIER, Loris MARCHAL, Samuel THIBAULT	Memory-Aware Scheduling of Tasks Sharing Data on Multiple GPUs with Dynamic Runtime Systems	2022	IPDPS	Conference
Maxime GONTHIER, Loris MARCHAL, Samuel THIBAULT	Locality-Aware Scheduling of Independent Tasks for Runtime Systems (Extended Version)	2021	~	Research report
Maxime GONTHIER, Loris MARCHAL, Samuel THIBAULT	Locality-Aware Scheduling of Independent Tasks for Runtime System	2021	EuroPar	Workshop

Programming projects

Most of our research papers are associated with a source code in order to reproduce the presented results.

For instance, you can find our first algorithm developed to statically schedule task under memory constraint here and our second algorithm that dynamically schedule tasks here.

The code I developed during my research collaboration in Uppsala is a batch simulator as well as 4 schedulers and is available here.

I implemented into the StarPU runtime 5 new schedulers and a new eviction policy.

Teaching