Redesigning Storage and Operating Systems for Modern Data Centers
Ricardo Macedo - INESC TEC
Nov. 21, 2025, 2:30 p.m. - Nov. 21, 2025, 3:30 p.m.
ENGMD 279
Hosted by: Oana Balmau
Emerging hardware technologies are challenging decades-old assumptions in the design of data-centric computer systems. Byte-addressable persistent memory delivers near-DRAM performance with durability. Compute Express Link (CXL) enables scalable memory expansion and efficient disaggregation with low latency. Modern NVMe devices achieve millions of IOPS at microsecond scale. Yet, data centers—from cloud platforms to HPC supercomputers—continue to rely on legacy software stacks (e.g., kernel-based file systems and general-purpose storage, memory, and network abstractions) that prevent them from fully exploiting these advances. Meanwhile, the exponential growth of I/O- and GPU-intensive workloads has driven massive expansions in data center scale, carbon footprint, and power consumption. Recent studies forecast that data centers may account for 8–13% of global electricity use by 2030, underscoring the urgency of rethinking efficiency and sustainability. In this talk, I will share our group’s research journey and ongoing efforts to tackle these challenges, centering on three pillars: (1) rethinking storage building blocks for new hardware paradigms, (2) maximizing performance and resource utilization in large-scale environments, and (3) driving energy-aware design for sustainable data centers.
Ricardo Macedo is a senior researcher at INESC TEC and an invited assistant professor at the University of Minho. Since 2023, he has led the research line “Storage and Operating Systems for Modern Data Centers” within the Distributed Storage team at INESC TEC. His work explores how emerging technologies—such as persistent memory, resource disaggregation, user-level storage systems, and dynamic power management—will shape the performance and sustainability of data-intensive applications in cloud computing and HPC infrastructures. Ricardo is an active member of the research community, with publications in top venues including VLDB, EuroSys, USENIX FAST, and ACM CSUR, and he frequently serves as a reviewer for conferences and journals in operating systems, storage, and HPC fields, such as ASPLOS, EuroPar, Journal of Supercomputing, and ACM TACO. He is also leading and participating in multiple international projects and collaborations with researchers from McGill University, UT Austin, TACC, AIST, Jepsen, among others.
