Computer Architecture 101 for Scientists

Topic: Computer Architecture 101 for Scientists, and what it means for your cluster jobs

Abstract:

Not knowing how a supercomputer works can literally cost you the better part of your precious resource allocation. We introduce the basic concepts of parallel computer architecture and how they impact the way cluster users should think about their compute jobs. Cores, sockets, caches, memory, networking, and data storage all play a role for performance; as a user, knowing your way around the supercomputer’s architecture can help you to more “bang” for your CPU/GPU time “buck.” If you are intrigued (or even intimidated) by all this “complicated” hardware stuff but always wanted to know more, this event is for you. Turns out, it’s not that complicated at all.

Links mentioned in the talk: