active-time-scheduling

Description

This package comprises various algorithms for solving the Active Time Problem developed to this date. The following table provides a list of the implemented algorithms available in the subpackage schedulers and the corresponding articles where these algorithms were developed.

Scheduler	Reference
LazyActivationSchedulerNLogN	"A model for minimizing active processor time" (Chang et al., 2012)
LazyActivationSchedulerT	"A model for minimizing active processor time" (Chang et al., 2012)
MatchingScheduler	"A model for minimizing active processor time" (Chang et al., 2012)
DegreeConstrainedSubgraphScheduler	"A model for minimizing active processor time" (Chang et al., 2012)
GreedyScheduler	"Brief announcement: A greedy 2 approximation for the active time problem" (Kumar and Khuller, 2018)
GreedyIntervalsScheduler	--
MinFeasScheduler	"LP rounding and combinatorial algorithms for minimizing active and busy time" (Chang et al., 2017)
GreedyLocalSearchScheduler	"Brief announcement: A greedy 2 approximation for the active time problem" (Kumar and Khuller, 2018)
GreedyLowestDensityFirstScheduler	--
BruteForceScheduler	Used for testing
LinearProgrammingScheduler	"A model for minimizing active processor time" (Chang et al., 2012)
LinearProgrammingRoundedScheduler	"LP rounding and combinatorial algorithms for minimizing active and busy time" (Chang et al., 2017)
BatchScheduler	"Optimal batch schedules for parallel machines" (Koehler and Khuller, 2013)

Usage Examples

To create a job set, the subclasses of AbstractJobPool from the subpackage models are used. Different subclasses can be to represent different properties for the jobs in them, for example FixedLengthJobPool demands a fixed length from its jobs. The following example demonstrates the process of creating a job pool and adding a job to it:

from models import JobPool

job_pool = JobPool()
job_pool.add_job(release_time=5, deadline=8, duration=2)

To process the job pool, the subclasses of AbstractScheduler from the subpackage schedulers are used. To perform the processing, the job pool should be passed into the process function. The result of the function is the computed job schedule, which, if the problem instance is feasible, contains the information regarding the active time slots as well as individual schedules:

from schedulers import FlowMethod, GreedyScheduler

scheduler = GreedyScheduler(FlowMethod.PREFLOW_PUSH)
schedule = scheduler.process(job_pool, max_concurrency=2)