Allow client to "claim" jobs

[dsschult]

from slack:

e.g. if the client reports that it has just submitted 10 jobs with 1 CPU and 16 GB RAM each, then the server could deduct that from the state it distributes to other clients

that of course assumes that the jobs will actually run at some point. might need to apply a degree-of-belief scaling to prevent one site from over-promising and starving others

[dsschult]

Probably also need some timeout, before a job goes back to being "available"

[briedel]

How often do you get jobs that have a large RAM per core ratio? Aren't we mostly in 2 GB/core regime?

…

On 5 May 2017 at 10:14, David Schultz ***@***.***> wrote: Probably also need some timeout, before a job goes back to being "available" — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#85 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABG4FqVLyAZUoF7fJwHC7Ykx7puUZeVVks5r2zzsgaJpZM4NSDJy> .

-- Benedikt Riedel Scientific Programmer University of Chicago Computation Institute

[dsschult]

You'd be surprised. It happens quite a lot that we're in at least the 4 GB/core regime, and I've seem batches of jobs in the 20 GB/core regime.