This page is (a bit better than) a dump of ideas for the Phase two work that Mustie and Alun may lead on.

Currently (-- MarkNorman DateTime(2006-05-11T15:11:24Z)) I have two main ideas:

1. Get grid requirements from non-grid-users (users who don't care about the technology)
2. Provide evidence for the categorisation of users that we suggested in our earlier findings

Initial steps:

• we need a short document (2-3 pages) with aims and vague methodologies
• I need to send a proposal to JISC (they should say yes, as we've heavily underspent)
• Plan the time scales (again)

Requirements

After speaking before, I sent the following in an email (to Mustie) to see if it triggered off thoughts. It kind of sums up where I'm at, but obviously needs tightening up.

 Subject: Idea for requirements/competency work
 Date: Wed, 05 Apr 2006 18:44:24 +0100
 From: Mark Norman <mark.norman@computing-services.oxford.ac.uk>
 To: Mustafizur Rahman <mustafizur.rahman@exeter.oxford.ac.uk>
 CC: 'Alun Edwards' <alun.edwards@computing-services.oxford.ac.uk>

 Mustie,

 This is quite hard, but I thought I should try to get something to you 
 before going on holiday.

 This is incredibly rough - but should give some hints.  Please let me 
 know if it is way down on the uncooked end of half-baked.  i.e. if it 
 triggers off possibilities in your mind - excellent: if it sounds mad or 
 v. badly formed - please say so!

 Time-scale
 ==========
 Alun and I could work with you on something from September onwards. 
 (Actually, I may not be able to do much before October).  You said that 
 you would have to start in August at the latest.  (That might be OK if 
 you could work on the design etc. and start without us).

 Resources/FTE
 =============
 Hopefully, you at least 0.5 FTE, Alun 0.5 FTE from Sept onwards - maybe 
 me 0.2 FTE in October-November?

 Justification
 =============
 It would be good to find some proof that there are people out there 
 whose research would benefit from 'the power of the grid'.  However, 
 these people are non-computer-technical and so would not even consider 
 using the grid as it is now. In fact, it may be so far out of their 
 reach that they may never have heard of grid computing (and probably 
 assume that it is something supercomputer geeks do).

 Idea for work
 =============
 Formulate a study (and hopefully carry out a pilot) to observe or 
 capture the work or desired work of a research 'community' or group.  We 
 may then postulate that 'the grid could help you with this' (and 
 therefore outline a grid-based application or service) and see what they 
 think (without seeing any interface).
 [What we *DON'T* want to do is to go straight to part 2 of the above 
 paragraph - that's what most people have done].

 It would also be good to establish:
 - that much of this researchers' work has common parameters
 (e.g. common searches, common types of data to submit for transformation 
 etc.  Therefore, perfect for building a secure, predictable application 
 for).
 - given a simple interface with something that they want, the 
 researchers will come
 - there is no need for researchers in academic disciplines to be 
 knowledgeable about the technology underlying the delivery of these 
 applications/services

 Method?
 =======
 A questionnaire?!? Hmm. :-p
 Watching users.
 Interviewing users.
 [users = researchers, not actual users of grid technology]
 Participants should be chosen orthogonally to their knowledge of or 
 interest in grid computing.  We don't want the study to be skewed by 
 people with an interest.

 Costs/Resources
 ===============
 We'll probably (I'll know soon!) have at least £20K underspend on the 
 ESP-GRID project, so finances shouldn't be a problem.  The main thing 
 availability of people (you, me Alun).


 It all seems so very vague...

 See you on or after the 18th!

Categorisation of users

This should be fairly easy to do, but we might have a problem in that our set of users chosen for #1 (above) may be quite skewed. In the first of my two (hopeful) [attachment:AllHandsPapers2006/AllHands06TypesUsers.pdf All Hands Paper]s (PDF), we suggest the following users:

• SEU - Service End User (like a user of an app. in a portal), divided into
  • SEUD - pretty standard SEU, running queries on databases, text searches etc.
  • SEUX - special case of SEU (slightly higher risk). Actually runs executable code through an application interface (e.g. Matlab code).
• PUA - Power User Agnostic of which grid resource node his/her job/code runs on.
• PUS - Power User requiring Specific grid resource nodes (most common user at present).
• PUDS - Power user Developing a Service. (As PUA/PUS but developing expertise like SP.)
• SP - Service Provider.
• Grid-Sys - Infrastructure sysadmin. (System administration of grid nodes, possibly with infrastructure delivery and security expertise.)

It's easy to predict that in a mature and popular (with researchers) grid, the SEUs will be dominant in terms of numbers, dwarfing the Power Users (who are really the main security threats).

We can't look into the future, but it would be good to do some kind of work (hopefully alongside the requirements stuff) that helped to justify these labels. Any ideas?

Dates for a brainstorm session

1-2 hours at OUCS? (Or we could come to Begbroke?)

I shall really have to submit this proposal in early June, but we wouldn't have to work on it until later.

Resources/FTE/Time scales

• Mustie 0.5 FTE Aug-Sept, then ???
• Alun 0.5 FTE from Sept onwards
• maybe me 0.2 FTE in October-November?