`
`https://cvw.cac.cornell.edu/parallel/hpc/nodes
`
`Parallel Programming
`Concepts and High
`Performance Computing
`
`Introduction to Parallel
`Computation
`
`Terminology: Threads and
`Processes
`
`High Performance Computing
`High Performance Computing
`Taxonomy of Parallel
`Computers
`Nodes
`Clusters
`HPC Ecosystem
`
`Memory Access
`
`Program Design
`
`Data Communication
`
`Efficiency
`
`Thinking Big
`
`Quiz
`
`Nodes
`Parallel Programming Concepts and High
`Performance Computing > High Performance
`Computing > Nodes
`
`A node is a standalone physical computer unit with a
`network connection that typically runs its own copy
`of the operating system. Thus, a workstation in your
`office or your laptop computer is a node.
`Supercomputer clusters are composed of nodes
`connected by a communications network. Just as
`there is a major difference in packaging between a
`workstation and a laptop, the nodes in a cluster like
`Stampede2 are packaged into small units that can be
`mounted in a dense configuration that provides
`appropriate power, cooling, and network
`connections.
`
`Modern nodes often contain several processor chips
`that share access to local memory and storage.
`Furthermore, modern processor chips contain
`multiple processing elements called cores. Each core
`is capable of running an independent thread of
`execution, so each core on the processor chip can
`simultaneously process a separate stream of
`instructions.
`
`Thus, at any instant, a node can be actively
`processing as many streams of instruction as it has
`cores. The number of cores in a node is the sum of
`the number of cores on each of the node's processor
`chips. Parallel programming allows a program to use
`more than one core at a time. Depending on how
`
`1 of 2
`
`11/19/2024, 6:11 PM
`
`PETITIONERS - AMERICAN/SOUTHWEST, Exhibit 1025
`Page 1 of 2
`
`
`
`Cornell Virtual Workshop > Parallel Programming Concepts and High P...
`
`https://cvw.cac.cornell.edu/parallel/hpc/nodes
`
`the program is designed, it might run on multiple
`cores within the same node or run on multiple cores
`across multiple nodes.
`
`A conceptual depiction of a modern node with
`four eight-core processors that share a common
`memory pool. A node typically also handles local
`storage, network connectivity, and power. It is
`increasingly common for nodes to include special
`accelerator hardware like GPUs or TPUs. This node
`has a GPU.
`
`Resize Image
`
`Back
`
`Next
`
`© 2024 | Cornell University |
`Center for Advanced Computing |
`Copyright Statement | Inclusivity Statement
`
`2 of 2
`
`11/19/2024, 6:11 PM
`
`PETITIONERS - AMERICAN/SOUTHWEST, Exhibit 1025
`Page 2 of 2
`
`



