Use Intel’s 20-Step Process to Choose a Management Solution for Your Cluster


By Ian Lumb | February 06, 2014 | HPC cluster management, HPC network, cluster management shell, Security, Software image, Cloud Computing, Cluster Management, HPC Cluster, Linux Cluster Management, Amazon EC2, GUI, adding nodes, HPC job schedulers, User authentication, access services, access control, Linux cluster scheduler, Workload managers, MPI, parallel programming and runtime environments



Intel cluster management evaluation guide20 criteria to identify your solution for cluster management? Absolutely!

How? Use the Intel Cluster Ready Program as your guide.

An Intel team established 20 evaluation criteria for the ICR Program, and produced a guide to help make the right choice. The guide is a great decision-making tool as the 20 criteria span the breadth and depth of the HPC components stack by considering:

  • Support - for Linux distros, middleware (e.g., workload managers), HPC tools (e.g., compilers) and libraries (e.g., MPI, math), cloud utilization as well as the management solution itself
  • Scalability - especially as constrained by the management solution
  • Provisioning - image vs. package vs. hybrid methods for disk-less as well as disk-based deployments
  • Configuration - to include interconnect fabrics and workload managers, plus persistent storage of configuration data
  • Customization - to easily incorporate and manage third-party software
  • UIs - to include both command-line interfaces as well as GUIs that provide parallel shells as a built-in capability
  • Monitoring - from the cluster itself, to jobs being executed by the workload-management system
  • AAA services - including user administration and hierarchical access control
  • Pricing - licensing and/or other criteria that detail financial aspects of the management solution

Each of the 20 criteria is assigned a `weight’ that reflects its importance resulting in a set of requirements for evaluating cluster-management solutions.

The Intel team used this framework to evaluate five candidate solutions for cluster management. You can use their report to set up your own formalized process – based upon weighted criteria – to evaluate candidate cluster management solutions.

High Performance Computing eBook