Notes on BOINC Workshop 2005 Held at CERN, 11/12 July 2005 Dave wants to have millions of users and hundreds of projects But how does this scale??? - Scalability of platforms (GOOD! Ahead of the Grid!!!) - Scalability of trust (challenging. how to avoid projects that give BOINC bad reputation, without resorting to authoritarian measures? Suggestion of using .peer review. projects point at others they approve of. Suggestion of "ethical guidelines" (e.g. explain who is funding research, who will use data, support open access publication etc.) Support issues - Open source support of community is very good! - Help is needed to bring new scientific groups up to speed. - Package BOINC into user-friendly form for starting new projects. - ("Doing science with BOINC": Possible theme for an externally funded project, or collaboration with Grid projects that do this sort of thing.) Credit issues: - Credit for storage - how should this be measured? - Credit for doing (real) work (e.g. debugging etc.). ?? Registering and signing up should be idiot proof: - BOINC for Dummies! - We seem to be losing lots of users who just give up if things get too hard. - We seem to be reacting well to pro-active users, but ignoring the quiet majority. - User-friendliness may be key to long-term growth to millions of users. - Grid Republic may be part of the answer - would be good to know more about this. (The "join the revolution" slogan seems a bit provocative??) - Informative, up-to-date directory of BOINC projects would be great for users to choose projects quickly, without having to visit each site. Perhaps it could be hosted of at GridCafé website (perhaps in collaboration with Grid Republic???) Discussion of clusters of dormant PCs that could be activated for BOINC usage: what about creating a service mode for lab full of PCs?? Plan for how industry might get involved??? - How does BOINC relate to XGrid, WCG, etc. - Is Microsoft about to short-circuit BOINC with its own distributed computing offering? Areas for improvement in BOINC Server side quality of service: optimised finishing of jobs Use BOINC Simulator SUMMARY SESSION BRUCE ALLEN BOINC is currently a software development project: now let's have stable version of BOINC! Slow-down development, to make versions more reliable Less effort in features and more in making it robust Make a Roadmap for the next year that production projects can rely on Test list of examples that have passed Application and code side have passed this list (There used to be milestones but now more piecemeal) Stable versions come out every six months Need this to get into mainstream releases of Linux Vanilla version - Still possible to add features IGNACIO REGUERO Display graphically the amount of usage etc. Stats sites displayed in terms of credit units/ flops Share technology maps using maximind lookups (costs) Perhaps show it with MonALISA CPDN More detailed sub-project level tailoring required Other wish-list items: - make a BOINC newsletter of some kind - define BOINC guidelines or values (what sort of projects, publication policy etc) - make a BOINC project guide (helps users decide which projects they want to contribute to) - encourage some form of peer review (sites specifying other sites they think well of) - develop BOINC for distributed storage (define credit value, study QoS issues) - look into openlab/EGEE funding opportunities for BOINC development (CERN) - plan for a future pangalactic workshop at Berkeley (you!)