K1.1-B System Architectures [HPC Certification Forum]

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
# K1.1-B System Architectures
# Background

An HPC cluster is built out of a few up to many servers that are connected by a high-performance communication network.
The servers are called nodes.
HPC clusters use batch systems for running compute jobs.
Interactive access is usually limited to a few nodes.
A typical cluster contains of these parts:
  * various system-, hardware-, and I/O-architectures used for supercomputers, i.e. shared memory systems, distributed systems, and cluster systems
  * physical hardware; chassis/rack, computer system units, interconnect, power., compute node architecture with memory, local disk, and sockets
  * typical architecture of cluster systems consisting of nodes with different roles (e.g. so-called head, management, login, compute, interactive, visualization nodes, etc.)
Which are functioning as: 

* **Login or gateway nodes** where you are on after logging into the cluster and where you can work interactively
    * **Compute nodes** that are the workhorses of a cluster
    * **Admin or system nodes** that work in the background necessary for the operation of the cluster, e.g. for running the batch service, or starting and shutting down compute nodes
    * **Disk nodes** that provide global file systems, i.e. file systems that can be used on all other kinds of nodes
    * **Special nodes** e.g. for data movement, visualization, or pre- and post-processing of large data sets
    * **Head nodes** which can mean login or admin nodes used for management either by a user or an administrator
# Aim
  * To provide knowledge about the hardware components of an HPC cluster and their functions.

# Outcomes

  * Comprehend there are nodes with several functions
  * Comprehend that nodes are connected by a high-performance communication network
  * Comprehend that global file systems are available on all nodes of the cluster and are convenient because their files can be accessed directly on all nodes
    * Quantitatively, parallel file systems offer higher I/O performance than classic network file systems
    * Qualitatively, they allow several processes to write into the same file
  * Comprehend the two purposes of the communication network
    * Enabling high-speed data communication for parallel applications running on multiple nodes
    * Providing a high-speed connection to the disk systems in the cluster

# Subskills
  * [[skill-tree:k:1:1:1:b]]