0. Meta question: now that longjobs has graduated to Delivery, should we use Owls as the body for initial consultation and feedback on policy issues that the longjobs team raises (i.e., for general policy decisions, capacity planning, funding/staff-effort recommendations)? (Owls agreed at the 11/19 meeting that it would be the right place, minutes in fuzz[1910].) 1. Should we work on adding Linux slaves now, or wait until there's an explicit demand? If the latter, we would like a commitment that guarantees sufficient lead time (e.g., advance notice of one term, with management working out any necessary compromise with faculty). In general, should demand be considered a factor when determining whether support for a particular platform is "feasible"? (In earlier reports/proposals, we said "If feasible, all platforms in the current Athena release should be represented in the slave pool.") 2. Clarify issues surrounding eligibility, both for sane user communication and staff workflow; identify decision path for eligibility and priority questions. - Our policies and procedures should be in synch with Cluster Patrol and stopit practices; in particular, if an applicant does not meet eligibility requirements or is turned away due to capacity constraints, there should be a clear communication back to the user (and other IS groups as needed for future reference) about whether they may use bldg 37 on weekends, or whether they should be using non-Athena resources; particularly in the Sponsored Research case, someone should follow through with the faculty advisor (stopit, ACST management?) - FLs will handle courses; UROP and thesis applications will go to Accounts, but they should not be burdened with eligibility determination if users doesn't fit into any of those standard categories. Perhaps such inquiries should go instead to a list including longjobs team for technical/capacity questions, ACST mgmt for eligibility questions, or bump to Owls if policy discussion is needed? - We need to establish a clear policy which guides us in accepting registrations for the service. For example, should registrations be granted simply on a first-come/first-serve basis, should we reserve a certain percentage of time for, say, courses (thus possibly delaying or denying registration to individuals), etc. We propose having the application form: A. make eligibility requirements clear up front; if we need an explicit checkoff from users on the Sponsored Research restriction, perhaps add something just above the submit button such as "by submitting this form, you confirm that you will not use longjobs for Sponsored Research" B. include choices for standard academic uses only (i.e., course, thesis, UROP); if use would be "other", the user should be directed to an email address or separate form to be fielded by appropriate decision makers (as decided above) 3. Clarify policy on supporting software not licensed to run on a generic Athena workstation. A previous incarnation of the longjobs project notebook listed among "What longjobs IS NOT": A way to run software that is not supported on Athena platforms. which could be made more explicit, assuming we retain the same notion of scope: A way to run software that does not run in the Athena clusters (whether due to software incompatibility or license restrictions). Background: experience this term raised the possibility that faculty may want to run software that the vendor requires to be node-locked (e.g. that is either currently runnable only on private Athena machines in their dept). Because we have dedicated execution slaves, it would be technically possible to run node-locked software on them, but this would change the generic nature of the machines, imposing an extra burden on Athena Server Operations (e.g., if a machine needs replacement, they would either have to preserve pieces of hardware tied to the license key's node ID, or have ACST get new keys from the vendor) and ACST staff (license manager configuration/troubleshooting and vendor liaison side).