[Beowulf] SC13 wrapup, please post your own

Lux, Jim (337C) james.p.lux at jpl.nasa.gov
Tue Nov 26 15:38:43 PST 2013

From: beowulf-bounces at beowulf.org [mailto:beowulf-bounces at beowulf.org] On Behalf Of Tim Cutts
Sent: Tuesday, November 26, 2013 8:26 AM
To: Peter St. John
Cc: Beowulf List
Subject: Re: [Beowulf] SC13 wrapup, please post your own

OK, yes, but that's still counting papers, just with a more sensible scoring matrix.  It has exactly the same problem in really demonstrating in a grant proposal or budget justification just what the ROI is going to be in those terms.

It's quite funny watching vendors flounder whenever they claim they can offer great ROI and you ask them to actually do so.  Usually their crude spreadsheets end up telling you're going to spend $millions more using their solution.


There are an amazing number of places where there are requests for ROI on an inherently fuzzy target.  For the last year I've been working on FINDER,  a radar to detect buried disaster victims.  Of course, I get asked several ROI type questions:

1)      What's your detection probability?

2)      What's the ROI for the investment in the development?

OK, the second one is sort of bizarre and hard to measure.. How much is "saving a life" worth? There's a variety of economic metrics for what a life is worth used in public policy discussions.. (do we weight by QUALYs? Saving the life of a 90 year old might not be "worth" as much as saving the life of a child, in terms of future economic impact..   Shades of Swift's "A modest proposal", as read on its face, as opposed to the political commentary which it as).  There's also the whole bias introduced by whether victims are likely to be there in the first place.   A 100 percent accurate detector doesn't do much good when there's nobody to detect.

The first one is even harder.  Is there a "standard rubble" (no, there is not).  Is there a "standard victim" (no, but I do have some anthropomorphic simulators).  Is there a valid standard of comparison?  Not really: there's no (valid) statistics on how effective search dogs are, for example. For many of the same reasons: not all rubble is the same, not all victims are the same, and victims are unevenly distributed.

But, to keep this all Beowulf oriented, we did use HPC to model the microwave propagation through simulated rubble.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.beowulf.org/pipermail/beowulf/attachments/20131126/a4d57390/attachment.html>

More information about the Beowulf mailing list