Deadline based evaluator for GRID schedules
This tool is designed for evaluation of GRID schedules. If you need more information, you can contact the author at toth@fi.muni.cz
User Documentation
Use --help
to print a simple help overview.
The required options are --input filename
to specify the input file, --cpuAllowance
to specify the number of cores all users receive in their share, --ramAllowance
to specify the number of MB of RAM all users receive in their share.
To actually generate a heatmap, also specify --generateHeatmap
. This requires gnuplot
installed.
Input file
The expected format is JOBID ARRIVAL WAITTIME RUNTIME CPUREQ RAMREQ USERID QUEUE
for each line.
ARRIVAL
, WAITTIME
, RUNTIME
are expected to be time offsets in seconds from the schedule start.
CPUREQ
is expected to be the number of cores requested by the job.
RAMREQ
is expected to be the number kB of RAM requested by the job.
JOBID
, USERID
, QUEUE
are expected to be simple string tokens (without spaces).
Output files
The tool will generate several files. These files are named after the input filename.
Consumption file .consumption
This file contains time overview of CPU and memory consumption. File contains one record (line) per-minute in the format TIME CPU MEM
.
Deadlines file .deadlines
This file contains data for heatmap generation. File contains one record for each 2D coordinate (timex
user). Format of each line is USERID TIME DEADLINE_COUNT
.
Jobs file .jobs
File contains processed input information. File contains one record for each job in the input file. Format of each line is JOBID ARRIVAL WAITTIME RUNTIME CPUREQ RAMREQ USERID DEADLINE_SATISFIED DEADLINE_DIFF
.
DEADLINE_SATISFIED
is a boolean representing whether the job had it's deadline violated or satisfied
DEADLINE_DIFF
is the time difference (can be both positive or negative) against the calculated deadline
Users file .users
File contains processed information from the model, grouped by users. File contains one record for each user in the input file. Format of each line is USERID JOBCOUNT VIOLATED_COUNT SUM_DIFF SUM_POSITIVE_DIFF RELATIVE_VIOLATED RELATIVE_SUM RELATIVE_SUM_POSITIVE
VIOLATED_COUNT
is the total number of violated deadlines for this user
SUM_DIFF
is the total sum of all differences between calculated deadlines and actual job end times
SUM_POSITIVE_DIFF
is the total sum of differences between calculated deadlines and actual job end times for jobs that didn't meet their deadline
RELATIVE_VIOLATED
relative count of violated deadlines VIOLATED_COUNT/JOBCOUNT
RELATIVE_SUM
and RELATIVE_SUM_POSITIVE
are relative variants of SUM_DIFF
and SUM_POSITIVE_DIFF
(divided by JOBCOUNT
)