Advanced Storage Solutions - with STEAM Academy

by Gianmaria Del Monte (CERN)

Monday 3 Aug 2026, 14:00 → 16:00 Europe/Zurich

31/3-004 - IT Amphitheatre (CERN)

Find more information on this indico page

Abstract

Last year, CERN's storage systems wrote 1.2 exabytes of data and served 8.2 exabytes back, with peaks of one terabyte per second, and not a single byte of LHC physics is allowed to get lost along the way. This lecture explains how that is possible, starting from surprisingly small questions: why is saving a single file to disk harder than it looks? Why does CERN still archive its data on magnetic tape, a kilometre of it per cartridge? What breaks when your "computer" becomes a hundred thousand disks spread across thousands of machines?

Starting from the physical media (hard drives, tape, flash) and the failure modes of ordinary filesystems, we build up the key concepts of distributed storage (replication, erasure coding, consistency models, the CAP theorem) and then see how they come together in the systems CERN develops and operates in production: EOS, Ceph, and the CERN Tape Archive. No prior storage knowledge is assumed; by the end, the exabytes should feel almost reasonable.

Bio

Gianmaria is a Software Engineer at CERN in the IT Storage group, where he works on the development and maintenance of EOS, the large-scale distributed storage system that manages the massive data volumes produced by the LHC experiments. His work spans the full life of exabyte-scale storage, from the design of new features to keeping a system serving petabytes a day reliable in production.

Please note that pictures and videos might be taken during the event. The pictures and videos might be used for communication about the event. By joining the lecture, you are agreeing to being featured in these communication actions. 

slides.pdf