All
I came here as part of my effort to give feedback to the research software portal design. If I understand it correctly, this portal is meant to target three large groups of people:
1) software users (researchers, educators, and students),
2) software developers, and
3) operators/integrators that make software available to their users (such as Gateway and HPC operators).
As I perceive it, the website is definitely helpful for category 3 above (operators/integrators), and somewhat useful for category 2 (developers). Now here are some details (for category 1-2):
Wearing the glasses of a “software user”, I would have hard time understanding what things are presented in this portal. As software users, I would like to get:
1) an orientation of the entire portal (like, what is this portal? What is it supposed to provide? What can I expect from it?)
2) a list of software which I may be interested in (say, software in genomics; or software in computational atomic physics; or software in combustion—something like these)
3) a map of relationship between these software packages. Some sort of “mind map” would be great, although not easy to make, and require the domain knowledge. For example, for AMBER, there may be some tools out there that preprocess the input, and other tools that postprocess the output (e.g. trajectory analysis, …).
4) an information about where (computing site) this software is available, how to access it, how to run it, etc.
5) sample codes and tutorials on using the software, which will be especially useful for newcomers.
These are the things that are high on my mind. As a novice “software user”, I would be lost in this portal trying to look for the stuff that are of my interest. For example, on this page:
https://software-dev.xsede.org/community/software-users
under “Discover Software”, “Software available on HPC resources”, it led me to here:
where there were two software packages (“OSG GSI OpenSSH login service” and “Xstream GSI OpenSSH login service”). But I was expecting to find, for example: Abinit, Quantum Espresso, and CP2K are available on SDSC Comet; FFTW3 available at TACC, Comet, Bridges; and so on. So there is disconnect between the wording used in the site and what came up.
As a “software developer”, the information provided in the portal is somewhat useful, but I would still be confused. I would need to know a little bit more on the XSEDE way of providing and integrating software. For a software programmer who only know how to write the program and run the program on a bare-metal HPC environment (like Bridges or Comet), I would have no idea about gateways and other ways of executing software (e.g. Open Science Grid). There are also documents that dated back in 2013 or even earlier, such as the ones in this page:
https://software-dev.xsede.org/xcsr/use-case-registry
They are so specific to XSEDE. My question is, what is the relevance of these documents to HPC software developers in general?
In brief, if there are some pages dedicated to orienting people to the portal, it would be really helpful. It would be nice to provide introductory pages which orient the uninitialized people (like me) to the whole effort, its goal, what it is meant to provide, etc. They would really increase the usability of the portal. Perhaps some of these contents are well beyond the original scope of the firstly envisioned plan; in that case, take this input as a suggestion for the future improvement.
Thanks,
Wirawan
Wirawan,
>> 1) an orientation of the entire portal (like, what is this portal? What is it supposed to provide? What can I expect from it?)
We're working on an overview/about page that will provide that Portal orientation you suggest.
>> 2) a list of software which I may be interested in (say, software in genomics; or software in computational atomic physics; or software in combustion—something like these)
We are providing keyword/topic based software search. As we mentioned in another thread, we need to identify repositories that we can load software information from, or software owners that would want to enter their software information by hand. The easiest and best approach is for us to ingest information from other repositories.
>> 3) a map of relationship between these software packages. Some sort of “mind map” would be great, although not easy to make, and require the domain knowledge. For example, for AMBER, there may be some tools out there that preprocess the input, and other tools that postprocess the output (e.g. trajectory analysis, …).
Interesting idea. This is something we won't be able to address right away, but we'll keep track of the idea.
>> 4) an information about where (computing site) this software is available, how to access it, how to run it, etc.
Good idea. We have that information already for XSEDE service providers and are making it more searchable thru this portal.
>> 5) sample codes and tutorials on using the software, which will be especially useful for newcomers.
Our goal is to enable basic software discovery and to point users at existing online resources about that software. You can see an example here:
https://software-dev.xsede.org/operational-software-and-service-componen...
>> There are also documents that dated back in 2013 or even earlier, such as the ones in this page:
>> https://software-dev.xsede.org/xcsr/use-case-registry
>> They are so specific to XSEDE. My question is, what is the relevance of these documents to HPC software developers in general?
One of the important features of the portal is to enable researchers, educators, students, and operators to discover what software does, which is what we call use cases. We have over 100 such use cases originally documented by XSEDE but that should be relevant outside XSEDE. We do need to rewrite many of these use cases to be non-XSEDE specific, because they should apply elsewhere too. We want other research infrastructures and projects to be able to document their use cases and link them to the software they use. This is our idea for how to make research software infrastructure more transparent. We have some ways to go to make it easy to do thru this portal.
Thanks for your detailed feedback.
JP