NEW YORK CITY – December 17, 2025 – VAST Information immediately introduced that SciNet on the College of Toronto, and SHARCNET on the College of Waterloo have adopted VAST as their knowledge and AI working system, constructing on beforehand introduced deployments with Canadian-based infrastructure suppliers which might be utilizing VAST to energy large-scale GPU providers, based on VAST.
Collectively, these organizations signify a cross-section of Canada’s AI ecosystem – from nationwide tutorial compute websites and world-renowned AI analysis institutes to cloud infrastructure suppliers – converging on a standard requirement: a scalable, resilient knowledge and AI working system that may hold tempo with GPU-era workloads whereas remaining easy sufficient to function with lean groups.
SciNet, considered one of Canada’s 5 nationwide tutorial HPC websites and residential to the Trillium supercomputer, is working its third-generation system with roughly 240,000 CPU cores and 250 GPUs. Fairly than repeat a legacy mannequin, SciNet has consolidated on the VAST AI OS infrastructure, designed to serve conventional HPC codes and rising AI workloads aspect by aspect.
“We run a rare mixture of workloads by researchers throughout Canada, and our new system pushes properly past the dimensions of our earlier era,” mentioned Daniel Gruner, CTO, SciNet. “With VAST, we not must juggle totally different storage tiers or bolt on burst buffers simply to maintain up with demanding I/O. We will help each conventional HPC and new AI workloads on VAST’s AI OS that our customers hammer day by day – and it merely retains up.”
SHARCNET, one other of Canada’s nationwide host websites and the group behind the brand new Nibi cluster on the College of Waterloo, has equally chosen the VAST AI OS as the information basis for its refreshed infrastructure. Nibi combines greater than 135,000 CPU cores with 288 GPUs and superior immersion cooling, and serves a “lengthy tail” of analysis disciplines – from bioinformatics and physics to economics and the humanities.
“We help hundreds of researchers with wildly totally different I/O patterns, from billion-file bioinformatics datasets to college students studying HPC for the primary time,” mentioned John Morton, Director of Know-how, SHARCNET. “With our earlier system, a single misbehaving job may affect your complete cluster. One of the best factor we will say about VAST is that it has largely disappeared into the background – it has quietly absorbed billions of recordsdata and a few extraordinarily demanding workloads with out ever changing into an issue we now have to debug.”
-
Unify HPC and AI on One Platform: VAST AI OS permits organizations to run tightly coupled simulations, knowledge analytics, AI coaching and inference on a single, unified platform – as a substitute of managing a number of tiers, burst buffers, and database and compute providers from disparate infrastructure.
-
Maintain GPUs Totally Utilized: VAST’s Disaggregated Shared-All the pieces (DASE) structure delivers constant efficiency for small-file, high-IOPS and metadata-heavy workloads in order that CPU and GPU sources stay compute-bound, not I/O-bound.
-
Simplify Operations for Lean Groups: VAST offers groups with a contemporary administration expertise, built-in snapshots and catalog providers, and automatic resilience options that scale back the necessity for guide tuning and firefighting.
-
Allow Safe, Multi-Tenant Environments: The VAST DataSpace delivers a world namespace with multi-tenancy and knowledge providers to help safe, compliant environments for delicate knowledge and multi-organization analysis collaborations.
-
Put together for Multi-Web site and Multi-Cloud Futures: VAST allows organizations to construct towards a standard knowledge basis that may span nationwide analysis websites, Canadian infrastructure suppliers and public cloud environments, enabling replication, shared datasets and constant governance.
“Canada has made daring, strategic investments in AI – from nationwide analysis facilities to cloud infrastructure designed for large-scale GPU computing,” mentioned Pezhman Sharifi, Director at VAST Information Canada Inc. “What unites SciNet, SHARCNET and different Canadian infrastructure companions is a shared want for an working system that retains GPUs and researchers fed with knowledge with out including operational complexity. The VAST AI OS delivers a single, world platform for Canada’s innovation economic system – from tutorial labs to manufacturing AI providers.”
