I am trying to understand how actually grid arrives with a definite number
between 0 and 1 for normalized priorities. I believe I understand how the
grid calculates the overall priority (),
To give the background, I am calculating the priority from application
stand point by considering various factors (as per the business needs) with
an impression that the higher priority that I sent will be given
importance.
But what is happening is when I submit (I am using DRMAA v2 – java API)
jobs with different priority ranges, many times time lower priority jobs
are winning the slots. (for example sake high priority is 999.978 and lower
priority 654.99). the jobs that are submitted are allocated to the same
project, have the same no of shares in shared tree and weightage factors
are exactly same. The only difference I can think of or block box area is
how grid engine is calculating “Normalized Priorities/Urgencies”.
Well I still agree that normalized urgency might vary based on submitted
time.
I surfed a lot in net to figure out how this function works, I will
appreciate if any hints on how normalized(ppri) or normalized(urg) or
normalized(tckts) works.
*npprior* = normalized(*ppri*)
*nurg* = normalized(*urg*)
*ntckts* = normalized(*tckts*)
*prio* = *weight**_**priority* * *npprio* +
*weight**_**urgency* * *nurg* +
*weight**_**ticket* * *ntckts*
_______________________________________________
users mailing list
users@gridengine.org
https://gridengine.org/mailman/listinfo/users
