Present: Paul Anderson, Tim Colles, Neil McGillivray , Dave Robertson, Alastair Scobie, Ian Stark, Perdita Stevens

Minutes and matters arising

Power sockets for laptops is being investigated by Dave Hamilton; Alastair will follow up to know the outcome.

Actions had been done, except that those present except PS had not added use cases. We decided that the main missing category was administrative use cases.

Action: Neil to add these, avoiding listing things for which using non-Informatics university systems is obligatory.(DONE)

Action: The person who asked about URLs was Iain Murray. Perdita will mail him about the outcome.(DONE)

Date and time of next meeting

Wednesday March 27th 2013, 10-11am. Ian requested that in future our meetings start in accord with the University's teaching hours (which 10am is). This was agreed.

Action: Perdita to book room and announce (DONE)

Discussion of strategic direction for provision of IT to Informatics.

We have a pressing need to save effort from somewhere that we can spend on unmet demand (e.g. the development projects that we agree should be done but that repeatedly don't get resourced). We suspect we may be resourcing things that are less valuable than some of the things we are failing to resource. But which? We now have, or will have,

- a list of use cases summarising what services need to be provided somehow;

- a list of services we currently provide.

We discussed (a) radical change [e.g., abandoning DICE] vs (b) incremental change [e.g., turning off particular services that we provide but could get, perhaps with less functionality, from IS or elsewhere]. We do not have, or see how to get to, any realistic radical proposal that could possibly meet our needs, so we consider (b). Here we lack crucial information for a cost/benefit analysis, e.g. for specific X, how much would how many users mind/suffer if we saved effort by ceasing to provide X? The perennial difficulty is to distinguish between cases where providing a service ourselves saves enough effort, mitigates enough risk or adds enough value elsewhere to be worthwhile and cases where, even though it might save some effort, mitigate some risk or add some value, this isn't enough to offset the cost (and opportunity cost) of providing it ourselves. As an example, the issue was raised of whether it is sensible that we run our own authentication service. There was a feeling that it may not be, but no certain conclusion could be reached.

Action: Perdita and Alastair to get together to plan how to elicit input from all users about what services they use (and how much; in particular, where we provide a service better in some way than something IS provide, how much is the increment valued?), then to circulate round CSG for comment before carrying out the exericse. (Discussed and variant DONE)

We also discussed quick decisions and agile implementation of changes vs careful planning and consideration of options. Opinions differed as to whether or not we have a general pattern of being too far on the latter end of the spectrum, but we certainly do need to bear in mind that not all changes need to be made the same way.

Computing support for teaching (VLEs, submit system, etc.): what to do next.

Apparently, Learn will by now have been reconfigured so that it is possible for courses to have publicly available pages (with stable URLs?) If so, it will probably better meet our needs, and we should have a few courses trying it out in Semester 1 of 2013/14.

There is some feeling from students that it would be helpful if there were a page per course in Learn pointing out at the real course homepage, even when the course is not actually in Learn.

Ian felt that the world of VLEs, PLEs etc. is best approached opportunistically; we should continue to encourage people to experiment as they see fit and then disseminate experience, rather than attempt to plan strategically.

Action: Ian and Perdita to discuss Learn technicalities and, if appropriate, encourage more people to use it next year and/or insert dummy pages.(Not done; variant recreated)

Future of local cluster computing provision (see attachments below, from Tim)

We appear to have very limited demand for GridEngine , but Hadoop is more used (for teaching and for research). Perhaps we could shut down half our existing machines, retaining the other half for Hadoop.

Action: Tim to talk to the current GridEngine users about their requirements and alternatives, and to Stratis about the Hadoop options.(DONE)


ECDF: we are in year 3 of a 4 year charging model, currently using only about 50% of our base allocation. Liz has projected zero spend on ECDF from 2014 onwards, chiefly because the current charging model seems very much to our disbenefit.

Action: Someone (Perdita by default) to ask Liz how much of the ECDF cost is, in fact, covered by grants.(Carried; not done)

Action: Perdita to ask Jim to come to our next meeting and, if he wishes, provide a paper, concerning the value and use of ECDF compared with alternatives.(Carried; DONE)

-- PerditaStevens - 05 Nov 2012

Topic attachments
I Attachment Action Size Date Who Comment
pdfpdf beowulf_power.pdf manage 21.2 K 08 Jan 2013 - 11:20 PerditaStevens  
pdfpdf cluster_usage.pdf manage 28.6 K 08 Jan 2013 - 11:19 PerditaStevens  
pdfpdf clustering.pdf manage 329.3 K 08 Jan 2013 - 11:19 PerditaStevens  
Topic revision: r8 - 12 Mar 2014 - 21:22:43 - PerditaStevens
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback
This Wiki uses Cookies