Workshop on Progress in Distributed Operating Systems and Distributed Systems Management, German National Research Center for Computer Science (GMD), Berlin, April 19, 1989 Some Unsolved Problems of Distributed System Management, as seen from Project Athena by J. H. Saltzer Talk notes (transcribed from handwritten version) Two differences 1. Title of workshop = European Update but the work I report is from North America - Done in the Cambridge near Boston - I am on sabbatical in the Cambridge near London (perhaps there was some confusion) 2. Two themes Advances in } Dist. Sys. Management of } Lots of speakers are willing to stand up in front of an audience and report they are making advances. Not so many are willing to say they are just managing. It happens that my project does emphasize system management, so my talk is a scarce commodity. ___________________________________________ Project Athena Began 1983 - Not research. Purpose is to provide M.I.T. undergrads with networked eng'g workstations for doing homework. - Overall architecture is a client-server model - based on UNIX everywhere (slightly modified BSD 4.3 with additions) - 10 Mbit network interconnects everything. - Scale (first slide) indicates why management receives a lot of emphasis. Motivation: minimize labor while permitting both - development - day to day operation For performance the workstations have disks. Some system materials are there. Issue is managing the bits stored on the workstation. (easily changed, require a visit to examine in the case of trouble.) (slide 2) Strategy: revise UNIX, make conventions, to eliminate contents of lower left corner. - upper left corner has special server, Name server - upper right can be divided arbitrarily between w/s and file server. - installation of workstation has only one item to hand tailor: the network name. (Slide 3) Eliminates - setup (8 hr average for UNIX) - service call for software For private W/S: We manage the system disk. (Slide 4) See abstract