Dear Anelda,
thank you for your reply. I think we have to go through the discussion
again about what should this lesson contain that you bring up. In nuce,
I think we should concentrate on 1-2 core principles behind cloud and
hpc. I agree that many scientists are confused by the availability of
computing resources today. Therefor, I had in mind teaching HPC
(classical simulations) versus HTC (a majority of bioinformatics, image
processing, deep learning, cloud stuff) based on a somewhat simple
example without diving too much into the details of the underlying APIs.
With this, I hope to convey a simple mental model that will help
participants not only to exploit HPC/cloud more effectively but also to
judge available technologies better.
I warmly invite you to repost your reply in a contrived manner here:
https://github.com/swcarpentry/hpc-novice/issues
Best,
Peter
On 06.12.2016 04:54, ane...@talarify.co.za wrote:
Dear Peter,
Seems like HPC Carpentry crops up at least every 2 months in
conversations around the world.
We recently had a great email chat between Mateusz Kuzak (Netherlands
eScience Centre), Christina Koch (University of Wisconsin-Madison),
Peter van Heusden (South African National Bioinformatics Institute)
and myself.
We would ideally like to see a lesson on research computing
infrastructure - explaining the different infrastructures that are
available - grid, cloud, hpc, other(?) and helping researchers to
understand when to use what - or rather what they need to address
their specific research computing needs. This is more than just HPC
novice, but with such a wide variety of infrastructures becoming very
accessible (even to researchers in Africa) it would be great if our
researchers understood why they should/could use the one or the other.
Once we have a lesson like this in place, an applied HPC lesson might
have much more impact? I'd love to see a HPC novice lesson developed
to complement that as our university (and I suppose many around South
Africa) could benefit greatly from it. Previously people have bumped
into the problem of HPCs which are differently configured, but I think
there is a sufficient level of abstraction possible where novices can
learn concepts applicable to HPC in the broad even if they were taught
the details of a specific system. Learning the new vocabulary of
queues, head nodes, schedulers, etc will already take them a long way
in being able to talk to their HPC service provider.
I'm not sure how far things have gone with the lesson I describe above
- I think Christina and possibly Mateusz was working on something to
this effect?
Happy to be involved in both.
By chance I'll be at the South African HPC conference this week and
could see if there is interest there to help develop?
Kind regards,
Anelda
_______________________________________________
Discuss mailing list
Discuss@lists.software-carpentry.org
http://lists.software-carpentry.org/listinfo/discuss
