To share at least a little knowledge with the world, I documented the basics of discrete event simulation for reasoning about #HPC system design. The example I wrote shows how to calculate MTTDL for RAID arrays of differing sizes, parity disks, and drive MTBFs.

glennklockwood.com/garden/discr...

I hadn't heard of hardened libc++ before - libcxx.llvm.org/Hardening.html. Obviously, Google's code (lots of it being infra code) is different from scientific software, but it's still interesting. Anyone played with this?

And success stories are the best advocacy: www.researchcomputingteams.org/newsletter_i...

Some other things I've written on this basic topic - scientific judgement is part of our job: www.researchcomputingteams.org/newsletter_i...

If a VPR or a funder has to decide between two similar centres, one can demonstrate opening up new research directions, good careers for trainees, and spinoffs, and the best the other can offer is 89% utilization or fully checked off worthiness lists, how do you think that decision is going to go?

Without us making the case for our work and our researcher client's work, how can those in charge possibly be fully informed for their next funding decision?

No one else is going to give them that information they need.

Advocating for the work we do means *qualitatively* showing decision makers how our work and our researcher client's work directly supports their priorities and missions. How the impact we're having is the impact they want to see.

We owe it to our teams, we owe it to the researchers whose life's work we support, to powerfully advocate for the work we do.

So allocation of resources for supporting research is decided based on human research judgement. Yes, messy, flawed, biased, human research judgement.

There are too many diverse *kinds* of research and scholarship for them to be able to be compared against each other in any kind of quantitative or checklist-algorithmic way.

There are more useful, worthy things to spend research funding on then there is research funding. That would be true even if research funding doubled tomorrow.

How many units of "impact in qualitative social sciences" are there in one unit of "impact in quantitative computational biology"?

But *even if it was quantifiable*, research funders and institutional decision makers have to decide how to allocate scarce resources between incommensurate things. How many units of "reusable research software" equals one unit of "well-used HPC cluster" equals one unit of "hire more postdocs?"

The worthiness or impact of the work we support is basically unquantifiable in the short term. The work we do to *support* that work, doubly so.

No one would actually *say* "Our work is demonstrably worthy - we've shown 85% utilization | 13 out of 14(!!) of the FAIR4RS principles met - so our work is done. If the funders don't fund us, it's on them". But...