Image based cluster management is ...
Node images are complete, functional snapshots of the compute node operating systems, typically generated and stored on the head node.
When the Y-HPC cluster is being "built", node images are copied from the server to each compute node as it reformats and rebuilds itself,
receiving boot directions from the selected template.
Each node image contains the applications, libraries, and data required for specific tasks. One node image may be used for multiple functions,
even by multiple end-users. Also, Y-HPC enables you to Copy, Edit, Remove, and Create new node images in order that you may change the function
of your cluster, switching between operating systems or from computational fluid dynamics to bioinformatics by simply selecting a different node
image from the GUI and then rebooting the nodes.
|
... flexible, scalable, & fast.
As your cluster is updated and grows over time, Y-HPC may be used to quickly, easily denote, archive, and retrieve older node images for
comparison or for reversion to a prior, working state. As each node image may be anywhere from 500MB to several, even hundreds of gigabytes, you
are limited only by the storage capacity on your head node, the Y-HPC server.
The template settings enable you to modify the function of each node image without having to modify the actual OS delivered to the node. A
Template enables you to change the boot parameters from a full installation local to the compute node drive to a partial netboot to a full NFS
root.
|
