Roadmap Components

1
Roadmap Components

• Data access and discovery, storage and management
• Grid enabled technologies and infrastructures
• Technical expertise
• High performance computing
• High capacity communication networks

2
PfC Elements
one desktop
collections
workflow
communities
3
PfC perspective on Characterisation

• New, enhanced or altered capability funded by
  characterisation can impinge on PfC in a number
  of ways
• Real-time vs asynchronous access
• Data rates and retention periods
• Metadata and provenance
• Computing requirements
• Meta resource scheduling
• Grid/Web/ standards and interfaces (two-way
  street!)

4
Characterisation perspective on PfC

• New, enhanced or altered capability funded by PfC
  can impinge on characterisation in a number of
  ways
• Aarnet 3 and beyond
• Significant (PB) regional storage systems
• At least one peak system 100 Tflop/s and many
  mid range systems 10 Tflop/s
• Possible eResearch centres, successors to APAC
  partners, expertise in every state
• Future FRODO/MERRI like activities providing
  expertise and (maybe) software

5
Some starting points

• Assume a 1 Gbit shared network exists between
  facilities and users
• Assume permanent archiving of data is out of
  scope for characterisation (and maybe out of
  scope for PfC, ie NCRIS)
• Assume remote interactivity needs to be provided,
  but keep it simple noting that the end points
  will be multi-purpose
• Note that facilities need robust and complex
  software but face common problems, dont green
  field it

6
My thoughts

• Data quality is a requirement so provenance and
  metadata support is a facility requirement (the
  community needs to determine what that means,
  however Archer et al can help)
• Data should be kept for a defined period in its
  capture environment with at least asynchronous
  access supported, hence some data store is a
  facility requirement
• Facilities may need to strengthen their dedicated
  software teams, software is mission critical,
  even integration is hard

7
Case by case

• X-Ray Techniques
• Data volumes appear to justify a plan for
  dedicated storage on or near-site
• Need to check dimensions for real-time data
  streaming on a national scale, it may be of
  diminishing return
• So what happens when the synchrotron deletes
  data?
• I would like better volume/service data
• I want you to articulate why it is an issue

8
Case by case

• Neutron Scattering
• From a PfC perspective this seems to be a single
  source of relatively small volumes of data with
  limited demand for real-time external interaction
• Can have storage on-site, replicate into the
  grid, aarnet3 connection
• Could bring grid into ANSTO to support two-way
  interactivity (eg. real-time external simulations
  using HPC, however can be achieved by portals)

9
Case by case

• Microscopy and Microanalysis
• Layer onto APAC Grid resources and use as an
  exemplar for planning for the PfC grid component
  (multi-point data sources, shared allocation
  system, interactivity, real-time processing etc)
• Assumption is that all the issues will arise but
  in data volumes and interactivity levels that can
  be handled by generally existing resources
• Issues catalogues, real-time interaction

10
Case by case

• Imaging
• Need to self-establish the collaboration
  structure of willing participants
• Then my initial thought would be to look support
  through grid developments

11
Big ticket issues

• Who keeps the data
• Note PMs advisory council is examining issues in
  science data
• Expertise, expertise, expertise
• Expensive for all groups to do everything
• But ready access to expertise is essential
• Dynamic interaction
• real-time multi-site multi-resource On Demand
  Research well planned ?
• Is the future, is it the present?
• A shared authorisation system
• Everybody is running their own AAA