The HPC is not really one computer; it’s three that are acting like one. The three computers function together as a cluster. You can think of a cluster as multiple computers working together as one computer. Clustering is the primary method for constructing supercomputers today, but it has its trade-offs. In order for all the computers to work together seamlessly and act like one logical computer, a lot of information has to be shared among the systems. Essentially, they each need the same files and user accounts. HPC is configured with a distribution of the Linux operating system known as Rocks that is specialized for cluster environments. When all goes according to plan, Rocks has tools that keep all the information shared between each node (computer) in the cluster.
We have been doing a lot of configuration work so far on the HPC: setting up user accounts, giving appropriate permissions, installing necessary programs, etc. Something went amiss and Zachary’s user account found its way to Limbo. We aren’t exactly sure how it happened, but best guess is that his user account was not properly synced throughout the cluster. Little anomalies such as this one kind of give an idea of the necessary complexity that comes with using a cluster.
Even though the HPC doesn’t like Zachary, we still think he’s an OK guy.
Till We Meet Again,