OGF 20, Manchester, UK 7th – 11th May 2007 Introduction During This meeting I attended the following Sessions: Monday 2nd International Workshop on Campus and Community Grids, continuing interoperability, (All Day) Tuesday Keynote, The Social Grid, Tony Hey Keynote: The Grid as a Complex System: Faster, Bigger, Better Science on Global Grid Infrastructure, Peter Coveney Software Licensing for Grids BoF. Presentation and Notes Wednesday Grids Mean Business: Software Licensing - A Way Forward? Presentation 1, Presentation 2 Usage Records-WG, Presentation & Notes The Astronomical Virtual Observatory - Building Operational Services on Pervasive Grids: Standards in Use, Panel Participant Thursday Data Management Area Friday Experience with Application Domains Massimo Lamanna (CERN) Session Reports Monday 2nd International Workshop on Campus Grids - Program Session 1 Introduction Campus grid case studies, connecting local services to a wider context (Part 1) - OxGrid and the UK NGS - GLOW and US Open Science Grid, Wisconsin - White Rose Grid - Crimson Grid, Harvard - Cardiff Campus Grid Session 2 Community grid case studies, connecting local services to a wider context - eMinerals - NEESGrid (or UK NEESGrid) Data Grids: ORION BIRN European Data Grids Session 3 National General Infrastructures, federating smaller grids, specific mechanisms for engaging infrastructures with different software stacks - DGrid - EGEE - Naregi - NGS-UK - OSG - TeraGrid Session 4 Breakout sessions - User management, passing identities and accounting - Support models, both systems and users - Software Licensing within cross organisational systems (This should be used to lead into the discussion in the Grids Mean Business session later in the week) - Data in grids, key topics including, Authenticity and integrity, access controls, technology evolution management. Session 5 Panel discussion and roundup on breakout sessions. This workshop was massively attended with us having at one point to turn people away from the room as we had reach the point of people sitting in the isles. The whole aim though was to illustrate that for each of the topics that were discussed within session 4, a grid of any size, be it organisational or international actually faces the same problems. From this we aim to give a lead to the further topics that should be touched on by GIN as it moves from a technical exercise to a useable long term infrastructure. The output from this workshop is a published OGF informational document. Tuesday Keynote: the Social Grid, Tony Hey Tony spoke about the importance of knowledge management and used many examples of institutional repositories etc. Keynote: The Grid as a Complex System: Faster, Bigger, Better Science on Global Grid Infrastructure, Peter Coveney Peter spoke about Reality Grid and his other wide area projects, introducing the necessity for higher level functionality such as computational steering etc and how we should ensure that standards are developed in thee higher functionality areas as well as the lowest levels. Software Licensing for Grids BoF. This was a summary session for the Roundtable that was held at SC06 on the Uk eScience booth. Production Grid Services Roundtable on Licensing SC06 Participants David Wallom (University of Oxford) Laura McGinnis (Pittsburgh Super Computer Centre) John Gordon (CCLRC) Dave Berry (University of Edinburgh) Pete Clark (University of Edinburgh) Martin Dove (University of Cambridge) Introduction During previous meetings of the Production Grid Services Research Group of the OGF (formerly GGF) there had been several discussions about users having significant issues with the ability to use licensed software within a grid environment. This culminated in the decision to host a roundtable discussion on the UK e-Science stand at SC06. The invitation below was sent to the following companies to encourage them to attend as well as give them a preliminary heads up that the event would be taking place. o Allinea (debuggers, optimizers) o Aurora Vienna (WIEN2K software package) o CEI (ensight.com; simulation software) o Etnus (TotalView debugger) o Numerical Algorithms Group (NAG; mathematics, statistics, data mining s/w) o Platform Computing (LSF) o Wolfram (Mathematica) o Visual Numerics (IMSL numerical libraries) o The Math Works (MatLab) o Cluster Resources (Moab) o Altair (PBS) o The Portland Group (compilers) Discussion So as to kick off discussion a set of slides was shown, which included the OGF IPR agreement. There were two pages of issues, one from the point of view of users, common scenarios for users within a grid environment and the other vendor issues which were completed during the discussion. QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture. o Allinea o Currently has what it considers a grid user friendly license o It is taken as read that companies exist to make money though o NAG o o o o o o o o o o o o o Consider they are currently licensed for grid systems and have invested in license models that users want, for example things such as state wide licenses (e.g. Ohio) John Gordon o A very common current scenario now is that we have user one and user two that are both from separate institutions and want to use resources within a third institution. NAG o A useful approach could be for example cross referencing between sites for example then, with the software tracking back to the users how institution to retrieve a license? Pete Clark o This could also be true within agile companies. Platform o Though a company is often not unwilling to have more open license T&C there may be technical impediments which mean that this is not easily achieved without substantial costly re-engineering. o It is also widely known that site licenses are not perfect o Moving to usage based models would be much more efficient and cost effective for many users though may also introduce more technical problems in other areas, networks etc. Allinea o We have for example produced our own separate license monitoring system. Visual Numerics o Our current license is flexible as to who is actually using the code which would be an advantage within a grid environment. Dave Berry o It should be noted that usage based licensing is not going to solve all the problems either Ohio SCC o It is a case of this type of model of license would never ever be supported by a SCC due to load etc that it could introduce. o Other common problems are; mixed usage where you have dual academic and commercial usage of a system shared usage between commercial companies… Allinea o Why do we all do Academic licenses, All (ish) o Challenging blue sky usage of products to show full capabilities o future business from people already used to certain products Martin Dove o The downside of pay as you go is that within an academic environment you discourage testing and mistakes by researchers who may otherwise have produced new outputs and hence more challenging usage of products Platform o It is important that we don’t confuse the model used with charging which is a different argument. o John Gordon o With grids becoming more ubiquitous there will be a proliferation of models, they must though comfortably live together o TotalView o Who do we license to within this environment? o NAG o The real way forward is to look for very good hybrid models that are flexible as we pass from one environment to another o Adjustments need to be more easily possible o Allinea o That is very possible but of course dependant on the types of software that you are dealing with, i.e. applications would have a very different usage profile to debuggers who are different again to schedulers. o The technology is already there for user tracking within a wide area ‘grid’ which could be used. o Mathematica o The current model requires the IDs of the machines on which you wish to run which would be fine for a tightly coupled grid but not so much for ad-hoc systems o Examples within current environments where this would work would be where licenses were purchased by Virtual Organisations and distributed across their resources. Conclusion and way forward It was decided that the best way for would be for those involved to try to document the use cases of their software has been used within a grid environment. This would mean little extra work for the commercial organisations but hopefully be of great benefit to users. The model we would hope for is where the company within for example its support FAQ has a section on approved license models within grid environments e.g. with short write-ups based on answers already given to current grid users questions of support or sales staff. This could then be referenced from within the Production Grid services website as well as from National Grid projects sites. Steve Little Northern European Sales Executive Platform Computing Limited Units 18/19, Building 2 Intec Business Park, Wade Road Basingstoke, Hampshire RG24 8NE, UK www.platform.com slittle@platform.com +44 (0) 1256 370510 Chris Gottbrath Product Manager Etnus 24 Prime Parkway Natick, MA 01760 508-652-7735 chris.gottbrath@etnus.com www.etnus.com Terasaka Haruo Center for Collaborative Research Institute of Industrial Science The University of Tokyo 4-6- 1, Komaba, Meguro-ku, Tokyo, 153-8505, JAPAN +81-3-5452-6601 terasaka@iis.u-tokyo.ac.jp Susanne M. Balle, Ph.D. Senior Member of Technical Staff High Performance Technical Computing Division Hewlett-Packard MS ZKO2-3/N30 110 Spit Brook Road Nashua, NH 03062-2698 603-884-7732 susanne.balle@hp.com Steve Scalpone The Portland Group Two Centerpointe Drive, Suite 320 Lake Oswego, OR 97035 503-682-2806 steve.scalpone@pgroup.com www.pgroup.com Silvina Grad-Freilich MATLAB Deployment Marketing Manager The Math Works, Inc. 3 Apple Hill Drive Natick, MA 01760-2098 508-647-7189 sgrad@mathworks.com www.mathworks.com Rob Meyer CEO, President Numerical Algorithms Group, Inc. 1431 Opus Place, Suite 220 Downers Grove, IL 60515-1362 630-971-2337 ex. 215 rmeyer@nag.com www.nag.com Andreas Landhauber Solutions & Innovation Scientific and Technical ICT/ HPCeScience T-Systems Solutions for Research GmbH Bunsenstrabe 10, 37073 Gottingen, GERMANY +49 551 7092379 andreas.landhaeusser@t-systems.com Dr. John W. Cobb R&D Staff Member Computer Science and Mathematics Division Oak Ridge National Laboratory One Bethel Valley Road P.O. Box 2008, MS-6016 Oak Ridge, TN 37831-6016 865-576-5439 cobbjw@ornl.gov M. Sean Fitzgerald Vice President of Technology and Consulting Visual Numerics, Inc. 2500 Wilcrest Drive, Suite 200 Houston, TX 77042-2759 713-954-6410 sfitzgerald@vni.com Wednesday Grids Mean Business: Software Licensing - A Way Forward? Chaired by the Grid Computing Now KTNs Ian Osborne, he first gave a presentation on why from the industrial p.o.v licensing is an issue. I gave the second presentation, giving a summary of the PGS Licensing session from the previous day. This was intended to summarise the number of different issues that had been identified. The output from this session is an agreement that a number of those present would agree to construct a set of model use cases within which licensing in a grid environment would be able to be described. This could then be used as a resource by users (both commercial and academic) to negotiate suppliers. It would show how other companies had worked out a suitable licensing model. Usage Records-WG Following on from the release of the version-1 standard into public comment for the second time, this session was used to sound out how we would consider extending it towards version 2. It has been shown that although good for computation there is a lot of extra functionality needed for storage, networking and overall of a service. The Astronomical Virtual Observatory - Building Operational Services on Pervasive Grids: Standards in Use, Panel Participant Attending the final two sessions of this workshop did not really allow me to get too much specifically from the talk. I was though asked to be on the panel as a person involved in production grids and production issues. Thursday Data Management Area Attended the first three sessions of this but found that even though this was supposed to be an EGEE User Forum session that these session were dominated by infrastructure people. The disconnect between any other user than particle physics and the EGEE infrastructure developers was never more clear! Friday Experience with Application Domains This session was much more application focused and useful. If did show the number of user areas that were working on EGEE.