CHAOS: Linux from Livermore
Cluster Administration Tools
Managing a cluster of many compute nodes often involves performing the same task on every node, or comparing the status or behavior of all nodes, or partitioning the nodes using planned configuration differences and then taking account of those differences for all future software updates. As the number of nodes grows (from clusters of 10 to 100 to over 1000 nodes), the difficulty of cluster management grows too and serious inefficiencies can appear.
The first subsection below summarizes special CHAOS-supported tools for system administrators. The second subsection introduces user tools to help anyone manage "NUMA (nonuniform memory access) nodes" on some LC Linux/CHAOS clusters.Links to subtopics in this section:
Navigation Links: [ Document List ] [ HPC Home ] [ Previous ] [ Next ]