PhD Position – Computational Storage and Its Application to Scientific Computing

Type de contrat : CDD Contexte et atouts du poste Supervisory Team: • Jakob Luettgau, PhD (Inria, France) • Francios Tessier, PhD (Inria, France) • Michael Kuhn, Professor (OVGU, Germany) • Gabriel Antoniu, PhD (Inria, France) Location and Mobility The thesis will be hosted by the KerData team at the Inria research center of Rennes. Rennes is the capital city of Britanny, in the western part of France. It is easy to reach thanks to the high-speed train line to Paris. Rennes is a dynamic, lively city and a major center for higher education and research: 25% of its population are students. This thesis will include collaborations with international partners from Germany, thus research visits to and from the collaborator's teams are expected. The KerData team in a nutshell for candidates • KerData is a human-sized team currently comprising 5 permanent researchers, 2 engineers and 6 PhD students. You will work in a caring environment, offering a good work-life balance. • KerData is leading multiple projects in top-level national and international collaborative environments such as within the Joint-Laboratory on Extreme-Scale Computing: https://jlesc.github.io. Our team has active collaboration with high-profile academic institutions all around the world (including the USA, Spain, Germany or Japan) and with industry. • Our team strongly favors experimental research, validated by implementation and experimentation of software prototypes with real-world applications on real-world platforms including some of the most powerful supercomputers worldwide. • The KerData team is committed to personalized advising and coaching, to help PhD candidates train and grow in all directions that are critical in the process of becoming successful researchers. • Check our website for more about the KerData team here: https://team.inria.fr/kerdata/ Mission confiée Context and Overview Computational storage [15, 16]  is a promising technology to improve the efficiency and performance of various workloads, for example in scientific computing, machine learning (ML), and artificial intelligence (AI). As the volume of generated data continues to grow exponentially [5,9] conventional compute and storage architectures are increasingly constrained by large data movements between storage/memory and compute resources. Computational storage can eliminate many of these data movements by co-locating compute capabilities along with storage locations allowing to offload the typically much smaller (sub)programs [1, 12, 14]. While the concept is well established in literature, computational storage devices are not widely commercially available or deployed in data centers today. A key challenge is that computational storage encourages domain-specialization for highest efficiency while economic factors encourage commoditization of products catering to broad markets. A second challenge is that both for legacy applications as well as emerging applications such as ML/AI it remains an open research question how to program and orchestrate across distributed platforms with computational storage capabilities. Recent advancements in programming models and software portability on the one hand, and reconfigurable hardware and domain-specific hardware design on the other [5, 7, 11], suggest that a modular approach that identifies common building blocks across domain boundaries might hold the key to both aforementioned challenges. Computational storage research exists on accelerating specific workloads or applications [5, 11, 14] as well as on emulating computational storage devices but a systematic study focusing on scientific computing workloads and modelling of suitable architectures and data distribution strategies is missing.  Research Objectives This project aims to advance the research on computational storage for scientific computing and artificial intelligence applications. It will investigate mechanisms to formalize, capture, model and evaluate computational storage in distributed environments. The project is structured into three primary objectives: • Objective A: realize a survey and define ontologies and taxonomies for distributed computational storage systems from multiple angles across multiple domains: the application perspective (programming paradigm, data flows), the middleware perspective (workflow orchestration, resource management and data placement), and the system perspective (e.g., hardware, storage, compute) • Objective B:  establish the technical foundations to empirically assess hardware performance and decompose existing workflows and task characteristics on the one hand, as well as implement suitable computational storage testbeds (e.g. simulated [2,3,4], emulated [7, 10], and on realistic software stacks and physical hardware [6, 8, 9, 11]) • Objective C: develop a methodology to faithfully model performance extrapolations to large-scale deployment scenarios as a means to identify and validate common computational storage building