Computing for large problems.
WHAT IT IS
Aggregating computing power to solve problems too large or slow for a single machine.
WHAT IT IS USED FOR
Weather and climate modelling Computational fluid dynamics Molecular and materials simulation Structural analysis Seismic processing Large-scale machine learning training
WHAT A CLUSTER CONSISTS OF
Compute nodes A high-speed interconnect Shared storage Login nodes, where users prepare work A scheduler, allocating resources
WHY THE INTERCONNECT MATTERS SO MUCH
Many workloads communicate constantly between nodes, and latency dominates.
WHAT A SCHEDULER DOES
Queues jobs, allocates nodes, and enforces fair use.
WHAT A JOB SCRIPT DECLARES
How many nodes and cores How long What memory What to run
WHY ACCURATE ESTIMATES MATTER
Overstated requirements wait longer in the queue; understated ones are terminated.
WHAT DISTINGUISHES THIS FROM CLOUD COMPUTING
Tight coupling, specialised interconnects, and shared scheduled access.
WHAT IT SHARES
The underlying principle of using many machines.
