CCIT Report #522, March 2005
Bounds on modified SPT heuristics for single machine flow-time
scheduling
with a single breakdown
Omer Gurewitz, Igal Adiri, Technion
Consider the problem of scheduling n jobs with known processing
time on a single machine. The machine
undergoes a breakdown at a fixed point in time, known in advance,
during the processing of the jobs. The job
preempted due to the breakdown has to be restarted from the beginning.
The objective function is minimizing
the sum of completion times of the jobs. It was found that the problem
is NP-hard and the SPT heuristic has
a tight bound of 9/7. In this paper, we present a scheme of a series of
heuristic algorithms as a function of the
number of jobs scheduled before the breakdown. Every algorithm is
characterized, for the worst case, by its bound
and complexity. Thus, we provide the scheduler with a simple tool for
determining his/her choice in the trade-off
between accuracy of the solution (deviation from the optimum) and the
computational effort.