Knowledgebase/Blog
Trip report: AHM 2011
Andy Powell
posted this on September 30, 2011 16:33
I've spent most of this week up in York for the All Hands Meeting, AHM 2011, my first time at the whole thing though I've been several times in the past for odd days here and there.
It was pretty good overall, though someone must have put the conference in a hot wash at some point because it seems to have got a lot smaller than it used to be :-) - there were about 175 delegates this year I think. There was also a slight imbalance (for my liking) between a very technical, hands-on focus on the nuts and bolts of eScience and a more strategic view of the wider issues facing 'research'. I guess that's how AHM delegates like it? So, whilst there ware several (good) papers discussing the minute detail of performance comparisons between HPC and cloud computing there was relatively little discussion of open access, open science or the potential funding changes that are being mooted for JISC and JANET (just to take a few examples). Perhaps I just went to the wrong sessions? It is interesting to note that the conference will change next year, becoming Digital Research 2012 (to be held in Oxford in September), reflecting in part a problem with the eScience 'brand' (some researchers not recognising eScience as being what they do) as well as indicating a desire for a broader perspective.
As to my personal highlights from the conference... there was one Linked Data paper by Arif Sharon of STFC describing an open source system for publishing geospatial data which I though was pretty good. Also a couple of 'cloud performance' papers, the first by Bruce Berriman who looked in detail at the cost and performance of using various cloud platforms to undertake astronomical research, and the second by Adrian Jackson (though it was presented by someone else) looking at benchmarking cloud infrastructure for HPC. The prize-winning paper by Valentin Tablan, et al, on GATECloud, text processing in the cloud, was also very good and a deserving winner IMHO. Finally, my favorite presentation was the closing keynote by Paul Watson of the University of Newcastle who talked about the work of the SiDE Project and how e-Science can underpin a social inclusion agenda. This paper felt like a breath of fresh air to me, moving right way from the detail of how, and instead focusing on why this stuff is important and the ways it can be used to improve the lives of ordinary people.
For my part, I was at the conference for various reasons: to take part in a jointly organised JISC/EPSRC workshop on 'cloud' before the conference proper and about which I'll return once the notes are made available; to speak in a workshop on Federating IaaS Clouds for UK Research organised by David Wallom on the Wednesday; to give a presentation about our UMF Cloud Pilot on the final day, and to take part in a JISC workshop on UMF after the conference ended.
In the 'federating' workshop I made the point that federations are typically "a relatively loose collection of organisations agreeing a set of policies and/or standards that are of mutual benefit in achieving a given objective (and where mutual benefit might relate to: usability, widespread adoption, sustainability, etc.)"... the important words being 'loose collection' and 'mutual benefit'. Also that we might learn from our experiences of the UK Access Management Federation. In particular, the importance of:
- mandating (or encouraging) open standards rather than particular software packages
- encouraging ‘diversity’ of implementation
- making usability (in its broadest sense) a high priority
- acknowledging that we don’t control everything and acting accordingly, i.e. recognising that stuff will happen around the edges.
Listening to the other presentations, and the ensuing discussion, I suspect that I might have been guilty of confusing the verb, to federate, with the noun, a federation, but I hope my contribution was useful nonetheless. My gut feeling is that cloud is relatively immature and that we need to encourage diversity and experimentation in all aspects (technology, business models, policies) and that we need to be careful that federating (or indeed federations) don't discourage that from happening at this stage.
The slides for my UMF Cloud Pilot presentation are available from Slideshare - feel free to take a look. Not a lot to report in terms of feedback though I was asked about whether we had a steering group (which we don't - though it might be something that we should think about) and that there was some discussion about the dislike of credit card billing models. My sense, at this stage, is that we need to think about a 'tariff' model for researchers (as an alternative to pay-as-you-go) and that we should investigate the possibilities for alternatives to credit cards as the payment mechanisms. Watch this space... but if you have thoughts on this in the meantime, please do get in touch.
There was some interesting discussion during the final JISC UMF workshop, particularly on the role of the JANET Brokerage and it's level of mandate around purchasing decisions for cloud within institutions, and around SaaS business models (though I must confess that by that stage my heart was firmly set on getting the train home). It was also noted that the UMF SaaS projects focus on data management of one kind or another but not much on compute (e.g. visualisation). I pointed out that a long term tape tier of storage would therefore be a useful addition to the 'fast' and 'slow' disk storage currently on offer as part of the UMF Cloud Pilot.
Also, there was surprisingly little use of Twitter as a back-channel (or anything else) at the conference which surprised me - I was expecting more.
Comments
Interesting that you pickup on the dislike of the credit card approach to billing, not sure what the alternative would be, monthly billing, also how, would be set up a steering group and what would that entail, what are the benefits and pitfalls of doing so.