Usability Huddle 1

• RP USABILITY

  • huddle on improving RP usability
    • motivation: tutorial, summer school (500 people -> mongodb instances)
    • polar science group: 250k long running jobs
      • -> reliability
    • reliability: can be quantified
    • usability: less so, but can prioritized
    • docs: faq, error messages, what-if, how-to
    • Ming/Shawn: less info: tell me what to put where, press 'run'
    • Antons : have MongoDB installed as part of RP?
      • TODO AM : investigate options
      • use an mlab DB only for getting_started.py
      • TODO MT: investigate an automated mlab setup
    • level of knowledge user needs to know in order to use RP
      • there is no normative user :/
    • error messages
      • errors are in wrong place (remote logs)
      • same holds for other information -> post mortem
      • write a root_cause_error.txt
      • raise errors early
      • error should contain suggestion on how to fix / what to do --> but they are mostly bugs :/
      • example: #1040
        • LRMS has no nodes left to run units
        • LRMS has no nodes left to run units: please increase pilot size
      • we should collect user experiences and feedback
      • Shawn : play the novice user: what message / guidance do I want?
      • Antons: play the average user: what message do I want?
      • Vivek : play the advanced user: same...
    • getting started: not only how, but importantly why (TODO Ming, Iannis)

• for now: focus on "Getting Started"

  • Tickets for documentation
  • suggest improvements for error messages
  • links to Installation and Quickstart in Introduction

• Quick Start Guide

  • https://docs.google.com/document/d/17cAhUyd0uV7IPtQ52U1kKOAJSFc10NVKSIfrMbEAOXA/edit
  • important to get an overview -- not only for end users
  • interactivity is important
  • real interactive runs are hard though

• Getting Feedback, judging quality

  • level i: https://en.wikipedia.org/wiki/System_usability_scale
    • works mostly in interactive settings, not for websites
  • level ii: single ease question (SEQ):
    • "how easy was task-x?"
    • can be extented with 'why?' / 'why not?'
  • level iii: https://en.wikipedia.org/wiki/Technology_acceptance_model
  • is usually worthwhile to do
  • limited effort
  • TODO MT: start with level i

• MongoDB deployment / usage:

  • binary distributions exist
  • hard to make foolproof, only works for local examples: but might be useful in this scope
  • MLab:
    • create account: well documented
    • automate account creation: hard
    • Heroku: can create mlab on the fly, so there must be a possibility
  • TODO AM: follow up
    • is local mode worthwhile introducing?
    • if yes: are there better approaches than local MongoDB install?

• Error Handling:

  • TODO AM, Antons, Sean, Vivek, MS: suggestions
  • Monday, July 1st: 5pm