2

Does this proposal define a new site beyond Super User, Server Fault, Computational Science, Linux, and our other computing sites?

Research Computing (proposal link)
for research computing facilitators, data center operators, XSEDE campus champions, ACI-REFS, and other users/supporters of advanced research computing

The purpose of Area 51 (our site-creation process) is to create sites in new subject areas that cannot be asked elsewhere. There's a discussion to decide if this proposal is mostly "general computing" questions that can be asked elsewhere. We generally do not create sites simply to give special-interest groups their own space unless the subjects are very specialized.

What do you think?

Top Example Questions

Have a look at the "top 40" questions above. Ignoring questions that may be "too broad" or too subjective, do you see a preponderance of questions that cannot be asked on either Server Fault, Super User, or our other computing sites?

Top 10 Marquee Questions for Research Computing

  1. What are cgroups and are they useful for cluster administration?
  2. How do I transfer large files to and from a remote system?
  3. My ssh terminal sessions keep timing out. What should I do to correct this situation?
  4. How do I find out what software is installed on an XSEDE resource, and how to link to it or use it?
  5. I would like to use Singularity or Shifter to run a program that requires a custom library under the /opt protected directory? How do I do this?
  6. How can I use passwordless ssh?
  7. How do you handle emergency resource requests, while still maintaining fair access to all users?
  8. What is OpenHPC, how and in what ways are people finding it useful?
  9. What annual conferences in the US pacific northwest specifically cover cluster operations?
  10. What are the problems with host Docker instances on a shared cluster?

Your help would be appreciated. Thank you.

7

I've commented my feelings on each of these questions.... Overall I don't see the need for a 'research' based site, especially if said research is on linux/unix systems (which it generally is).

  1. What are cgroups and are they useful for cluster administration? -> Linux/Unix

  2. How do I transfer large files to and from a remote system? -> too broad, either superuser (for windows or linux) or linux/unix. Definitely not a 'research' question.

  3. My ssh terminal sessions keep timing out. What should I do to correct this situation? -> Linux/Unix
  4. How do I find out what software is installed on an XSEDE resource, and how to link to it or use it? -> No idea what this is so i can't comment
  5. I would like to use Singularity or Shifter to run a program that requires a custom library under the /opt protected directory? How do I do this? -> Very much sounds like Linux/Unix
  6. How can I use passwordless ssh? -> Linux/Unix
  7. How do you handle emergency resource requests, while still maintaining fair access to all users? -> Linux/Unix would cover both schedulers, and nicing of processes which I think this would come down to?
  8. What is OpenHPC, how and in what ways are people finding it useful? -> Opinion based, bad question over all.
  9. What annual conferences in the US pacific northwest specifically cover cluster operations? -> Bad question, could ask on Linux/Unix
  10. What are the problems with host Docker instances on a shared cluster? -> Linux/Unix
3

My reaction was pretty similar to djsmiley2k's. The vast majority of questions in the list would either fit well within existing sites' scopes or are bad questions for the SE model.

The best rationale for a new site may not be lack of coverage, but lack of expertise elsewhere. The value would be in creating a pool of experts for the things that are more uniquely limited to research, like software packages that nobody else uses. They may technically fit under another site, but too few people there can answer the questions or competently judge answers.

2

Thats... unusual. I don't ever recall this ever happening.

With the current SE site model, overlap happens so its probably as much "can this site survive" as "will questions get asked elsewhere".

The answer to the question you haven't asked is "The site might work as something for the Research Computing community". Arguably a good chunk of these folks would be sysadmins and programmers - an equivalent of the site of many names , aka software engineering would be handy.


What are cgroups and are they useful for cluster administration?

Serverfault or U&L would fit. SU might be possible for the first part, though this feels like a "define a thing I saw in a book" question

How do I transfer large files to and from a remote system?

Has probably already been asked on SF, SU and U&L

My ssh terminal sessions keep timing out. What should I do to correct this situation?

Depending on the client/server combination, SU or U&L.

How do I find out what software is installed on an XSEDE resource, and how to link to it or use it?

ehh, Maybe. Supercomputers might come under SF, but sadly I've never worked in that field.

I would like to use Singularity or Shifter to run a program that requires a custom library under the /opt protected directory? How do I do this?

U&L, SU, SF

How can I use passwordless ssh?

Depending on the platforms - SU, AU, SE, U&L maybe even SO

How do you handle emergency resource requests, while still maintaining fair access to all users?

SF presumably. Its a classic 'soft' systemadministration question

What is OpenHPC, how and in what ways are people finding it useful?

First part lacks research. Second part is broad and listy. Bad question IMO

What annual conferences in the US pacific northwest specifically cover cluster operations?

Not really a good fit for a mainly technical community like a traditional QA site. If you're scoping for community resources it might work. Almost a shopping question

What are the problems with host Docker instances on a shared cluster?

Too broad. Expects Prescience. Would/should be closed on most sites.


On the whole I don't see these questions filling a unique niche that the existing sites don't serve very well.

  • This happens... just about every day. Folks propose sites in subjects we already cover. Folks want a site for Java, or they want one for cats or Minecraft or camping. But if you removed all the questions that can already be answered elsewhere, you should have enough leftover to justify building a new site. We don't want to optimize for simply re-shuffling existing communities to repurposed sites. Any overlap has to be incidental. – Robert Cartaino Nov 24 '17 at 2:47

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .