<div dir="ltr">Regarding ROI for academia. My background is in computational molecular thermodynamics and in my ~10 yr academic research career I logged millions of core-hours and contributed to ~100 publications. During that time, I can definitely say that my time-to-publication or publication output was not significantly impacted by the HPC resources to which I had access. The largest factors was the speed at which my advisor could respond with his/her revisions. Second to that, was my success rate in acquiring grants.<div>
<br></div><div>Furthermore a few of my publications required nearly zero core-hours but earned a relatively large Impact Factor. On the other hand, some publications used tons of core-hours and were probably never even read by other researchers. Generally, I would use as much cluster time as I could get my hands on. The total mass of hardware to which I had access was directly related to the absolute problem size and number of data points generated, not the number or quality of my publications.<br>
<div><br></div><div style>FWIW, in our industry (HPC OnDemand), calculating ROI is almost trivial.</div><div style><br></div><div class="gmail_extra"><br>-- <br>Kevin Van Workum, PhD<br>Sabalcore Computing Inc.<br><a href="http://www.sabalcore.com" target="_blank">www.sabalcore.com</a><br>
<br></div></div></div>
<br>
<img src="http://www.sabalcore.com/images/Sabalcore_Computing.png">