Qsim is a event-driven jobs scheduling simulator being part of the Cobalt resource mananger.
It contains several files specific for simulation while using most of the code directly from Cobalt.
The simulation specific parts including event-mananger, queue-simulators, and system simulator, etc.

Usage: qsim.py [options]

Options:
  -h, --help            show this help message and exit
  -j BGJOB, --job=BGJOB
                        file name of the job trace (when scheduling for bg
                        system only)
  -c CJOB, --cjob=CJOB  file name of the job trace from the cluster system
  -p CONFIG_FILE, --partition=CONFIG_FILE
                        file name of the partition configuration of the Blue
                        Gene system
  -o OUTPUTLOG, --output=OUTPUTLOG
                        featuring string for output log
  -i SLEEP_INTERVAL, --interval=SLEEP_INTERVAL
                        seconds to wait at each event when printing screens
  -F BG_FRACTION, --bg_frac=BG_FRACTION
                        parameter to adjust bg workload. All the interval
                        between job arrivals will be multiplied with the
                        parameter
  -f CLUSTER_FRACTION, --cluster_frac=CLUSTER_FRACTION
                        parameter to adjust cluster workload. All the interval
                        between job arrivals will be multiplied with the
                        parameter
  -S BG_TRACE_START, --Start=BG_TRACE_START
                        bg job submission times (in job trace) should be after
                        12.01am on this date.        By default it equals to
                        the first job submission time in job trace 'bgjob'
  -E BG_TRACE_END, --End=BG_TRACE_END
                        bg job submission time (in job trace) should be prior
                        to 12.01am on this date.  By default it equals
                        to the last job submission time in job trace 'bgjob'                                           
  -s C_TRACE_START, --start=C_TRACE_START
                        cluster job submission times (in job trace) should be
                        after 12.01am on this date.         By default it
                        equals to the first job submission time in job trace
                        'cjob'
  -e C_TRACE_END, --end=C_TRACE_END
                        cluster job submission time (in job trace) should be
                        prior to 12.01am on this date         By default it
                        equals to the last job submission time in job trace
                        'cjob'
  -A ANCHOR, --Anchor=ANCHOR
                        the virtual start date of simulation for bqsim. If not
                        specified, it is same as bg_trace_start
  -a ANCHOR, --anchor=ANCHOR
                        the virtual start date of simulation for bqsim. If not
                        specified, it is same as c_trace_start
  -P PREDICT, --prediction=PREDICT
                        [xyz] x,y,z=0|1. x,y,z==1 means to use walltime
                        prediction for (x:queuing / y:backfilling / z:running)
                        jobs
  -W WASS, --walltimeaware=WASS
                        [cons | aggr | both] specify the walltime aware
                        spatial scheduling scheme: cons=conservative scheme,
                        aggr=aggressive scheme, both=cons+aggr
  -C COSCHEDULING, --coscheduling=COSCHEDULING
                        [x y] (x,y=hold | yield). specify the coscheduling
                        scheme: 'hold' or 'yield' resource if mate job can not
                        run. x for bqsim, y for cqsim.
  -v VICINITY, --vicinity=VICINITY
                        Threshold to determine mate jobs in coscheduling.
                        Two jobs can be considered mated only if their
                        submission time difference is smaller than 'vicinity'
  -r MATE_RATIO, --ratio=MATE_RATIO
                        Specifies the ratio of number mate jobs to number
                        total jobs. Used in the case two job traces have the
                        same number of total jobs.
  -b, --batch           enable batch execution model, do not print screen