ORBIT-USER: [Fwd: Reservation created for 02/01/2007] - FAIRNESS?

[Ivan Seskar]

>thank you for your explanation. All the rules seem very nice on paper
but when it comes 
>to enforce them you guys do not follow them yourselves.
>According to rule #5 as you have written below, last week the one
conflict I had should
>have been resolved in my favor (since I only had one hour of grid time
that week). 
>Instead you write me back saying that a first-come first-served policy
is used 
>with no fairness at all.
>So, my question to you is, are you administrators going to follow these
rules? 
>(Rhetoric question) I am sorry if I sound upset but the reason is that
I am upset.
>Lately slots have been approved following no logic at all (or following
different 
>rules on different days) and when you are on a deadline a refused slot
might 
>mean missing an important deadline.

Unfortunately it is not easy to do it manually and as a side job (as you
observed); the only real 
solution is to put this algorithm in the Orbit service so that there are
no policy deviations.
Unfortunately we are working on more pressing issues (but scheduling
policy implementation is
definitely on our TO-DO list).

>A few other comments on two other points:
>2) Sandboxes have only two nodes, so it is very hard to be able to test
complex 
>experiments on sandboxes. Perhaps having a few sandboxes with 4 nodes
would 
>help a lot and would also free grid time.

There is no cheap solution to that problem - more than two interfaces
would require quite complex
RF distribution system that we don't have resources for. The 4 virtual
grids are supposed to address
that somewhat by enabling 4 simultaneous experiments on the main grid
but we are still at least few weeks 
away from deploying it (when I say 4 simultaneous experiments that
implies 4 simultaneous interfering 
experiments and is mainly meant for development).

Ivan.