Linux 6.0 Continues Plumbing For Compute Express Link (CXL)

[phoronix]

Phoronix: Linux 6.0 Continues Plumbing For Compute Express Link (CXL)

The Linux kernel continues making preparations around the very exciting Compute Express Link (CXL) thanks to the work of Intel engineers...

https://www.phoronix.com/news/Linux-6.0-CXL

[DrYak]

And all the HPC scientific software dev on this forum rejoiced!

(Though probably most of us are stuck with the admin running some ancient Red Hat or CentOS with completely out of date 3.10 kernel, so it will take a while until those benefits trickle down)

[chuckula]

And all the HPC scientific software dev on this forum rejoiced!

(Though probably most of us are stuck with the admin running some ancient Red Hat or CentOS with completely out of date 3.10 kernel, so it will take a while until those benefits trickle down)

Out of curiosity, how complex are the OS installations on nodes in an HPC cluster? At least on the biggest systems that I know "run Linux", I have always heard that the actual compute nodes are executing extremely stripped-down software stacks that are there just to run the current job and then get partially/completely wiped between jobs. Obviously the "management nodes" have beefier installations to do job scheduling & monitor system health, but the thousands of compute nodes were very plain. Has that changed?