Home

Maestro is a cutting-edge, high-performance computing (HPC) cluster recently deployed and fully operated by the HPC Core platform. Designed to meet the growing demands of data-intensive and compute-heavy research, Maestro leverages the latest AMD Zen architecture, featuring servers equipped with either two 48-core CPUs or a single 96-core CPU per node, delivering exceptional parallel processing capabilities. These servers are interconnected via a high-performance, low-latency HDR100 InfiniBand network, enabling efficient and seamless direct internode communication—ideal for large-scale MPI-based applications and distributed computing workloads.

The cluster’s memory configuration is highly versatile: the majority of the 29 servers in the common partition are equipped with 512 GB of RAM, while a subset of specialized "bigmem" nodes feature an impressive 2 TB of RAM per server, catering to memory-intensive simulations and large dataset processing. In addition to the common partition, there are 52 dedicated servers—financed directly by individual research teams but available for interim use by all users across the institution—ensuring greater flexibility and scalability for high-priority projects.

Looking ahead, the HPC Core platform plans to expand Maestro significantly in 2026, with the addition of more compute nodes, including GPU-accelerated systems to support machine learning, AI workloads, and other GPU-optimized applications.

Maestro is directly connected to the Helix storage system and the Appa home/scratch servers, ensuring rapid data access and efficient I/O performance. This tight integration minimizes data transfer bottlenecks and accelerates workflows, making Maestro a powerful and responsive resource for researchers across disciplines.