Job scheduler, job schedule control method, and storage medium
A scheduler includes circuitry configured to, based on similarity between execution time and power consumption information of jobs executed in a system, classifies jobs into groups, construct respective time series prediction models for the groups using a power waveform included in each of the groups as teacher data, predict a power waveform at an interval including a first time from each of the constructed time series prediction models, compare a power waveform at an interval including a first time of a job in execution for which power is to be predicted with the predicted power waveform of each of the groups to identify a similar time series prediction model, based on the identified time series prediction model, predict power consumption at a predetermined interval including a second time for the job for which power is to be predicted, and control job execution based on the predicted power consumption.
1. A job scheduler comprising:
a memory; and
circuitry coupled to the memory and configured to:
based on similarity between execution time and power consumption information of a plurality of jobs executed in a system, classifies the plurality of jobs into a plurality of groups,
construct respective time series prediction models for the plurality of groups using a power waveform included in each of the plurality of groups as teacher data,
predict a power waveform at a predetermined interval including a first time from each of the constructed time series prediction models,
compare a power waveform at a predetermined interval including a first time of a job in execution for which power is to be predicted with the predicted power waveform of each of the plurality of groups to identify a similar time series prediction model from among the respective time series prediction models of the plurality of groups,
based on the identified time series prediction model, predict power consumption at a predetermined interval including a second time for the job for which power is to be predicted, and
control job execution based on the predicted power consumption,
the similarity is a sum of distances from first data points of a first waveform to second data points of a second waveform, each of which is closest to each of the first data points.
2. The job scheduler according to claim 1 , wherein the identifying further includes identifying a first time series prediction model that is most similar to a power waveform over an entire section of a first job, and when a second job is performed by the same user as a user performing the first job, identifying the first time series prediction model as a time series prediction model at a first predetermined section of the second job.
3. The job scheduler according to claim 1 , wherein the circuitry is configured to further
determine a weight of each piece of information of submission time information so that the submission time information set at a time of submission of a job included in the classified group is classified into each group,
classify the submission time information into a group according to a result obtained by assigning a weight to each piece of information determined as submission time information of a power prediction target, and
predict power consumption from a time series prediction model of the classified group.
4. A method of controlling a job schedule, the method comprising:
based on similarity between execution time and power consumption information of a plurality of jobs executed in a system, classifying the plurality of jobs into a plurality of groups,
constructing respective time series prediction models for the plurality of groups using a power waveform included in each of the plurality of groups as teacher data,
predicting a power waveform at a predetermined interval including a first time from each of the constructed time series prediction models,
comparing a power waveform at a predetermined interval including a first time of a job in execution for which power is to be predicted with the predicted power waveform of each of the plurality of groups to identify a similar time series prediction model from among the respective time series prediction models of the plurality of groups,
based on the identified time series prediction model, predicting power consumption at a predetermined interval including a second time for the job for which power is to be predicted, and
controlling job execution based on the predicted power consumption,
the similarity is a sum of distances from first data points of a first waveform to second data points of a second waveform, each of which is closest to each of the first data points.
5. The method of controlling the job schedule according to claim 4 , wherein the identifying further includes identifying a first time series prediction model that is most similar to a power waveform over an entire section of a first job, and when a second job is performed by the same user as a user performing the first job, identifying the first time series prediction model as a time series prediction model at a first predetermined section of the second job.
6. The method of controlling the job schedule according to claim 4 , wherein the circuitry is configured to further
determine a weight of each piece of information of submission time information so that the submission time information set at a time of submission of a job included in the classified group is classified into each group,
classify the submission time information into a group according to a result obtained by assigning a weight to each piece of information determined as submission time information of a power prediction target, and
predict power consumption from a time series prediction model of the classified group.
7. A non-transitory computer-readable storage medium storing a program that causes a processor included in a computer to execute a process, the process comprising:
based on similarity between execution time and power consumption information of a plurality of jobs executed in a system, classifying the plurality of jobs into a plurality of groups,
constructing respective time series prediction models for the plurality of groups using a power waveform included in each of the plurality of groups as teacher data,
predicting a power waveform at a predetermined interval including a first time from each of the constructed time series prediction models,
comparing a power waveform at a predetermined interval including a first time of a job in execution for which power is to be predicted with the predicted power waveform of each of the plurality of groups to identify a similar time series prediction model from among the respective time series prediction models of the plurality of groups, and
based on the identified time series prediction model, predicting power consumption at a predetermined interval including a second time for the job for which power is to be predicted,
the similarity is a sum of distances from first data points of a first waveform to second data points of a second waveform, each of which is closest to each of the first data points.