I am working on a data set which has the waiting times taken of jobs running on a cluster. I need to come up with a method to use this historical data to come up with a prediction for the future. Even probably try simulating the full history (as in I have history of the job submission time and running time,etc). So I can run through the actual history and at every job submission, depending on the status of the cluster, try to predict the waiting time. Can I do this using any of the models used in R? Particularly "forecast"...Can I used ARIMA or ARMA for this? ...The problem is I don't think I'm dealing with time series because the measurements of waiting times (at a particular state i.e. job submission) ain't done at regular intervals. Could anyone please suggest a model for this? Thanking you, RS [[alternative HTML version deleted]]
Sounds more like you would want to explore the use of some sort of Queueing model. A quick search of R help did not yield any packages that could be used to develop such models, but I think that modeling simple queuing systems and estimating wait times is pretty straight forward and could be programmed in R. Hopefully someone out there will have more/better advice. Good luck, Spencer On 6/19/07, Roshan Sumbaly <rsumbaly@gmail.com> wrote:> > I am working on a data set which has the waiting times taken of jobs > running > on a cluster. I need to come up with a method to use this historical data > to > come up with a prediction for the future. Even probably try simulating the > full history (as in I have history of the job submission time and running > time,etc). So I can run through the actual history and at every job > submission, depending on the status of the cluster, try to predict the > waiting time. Can I do this using any of the models used in R? > > Particularly "forecast"...Can I used ARIMA or ARMA for this? ...The > problem > is I don't think I'm dealing with time series because the measurements of > waiting times (at a particular state i.e. job submission) ain't done at > regular intervals. Could anyone please suggest a model for this? > > Thanking you, > RS > > [[alternative HTML version deleted]] > > ______________________________________________ > R-help@stat.math.ethz.ch mailing list > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide > http://www.R-project.org/posting-guide.html > and provide commented, minimal, self-contained, reproducible code. >[[alternative HTML version deleted]]
Hi Roshan
see inline for answer
waiting time. Can I do this using any of the models used in R?
/********* i feel this sounds more or less like a queueing model. Try
searching "M/M/1 queueing model" on internet.
Further, i feel look at poission distribution may also be helpful. It was
Long Long time back, more than a decade, when i used GPSS software as i
cannot see anything susbstantial in
help.search("queue")/help.search("queueingmodel")to model
queues. i would
be of little help in remembering all those, but i would exten help to
whaever extent i can. *****************/
Particularly "forecast"...Can I used ARIMA or ARMA for this? ...The
problem
/***********
i doubt but may regularizing the series by either transformations and then
imputing it may turn irregular series into a time series. but the
possibilities are bleak. ******/
Particularly "forecast"...Can I used ARIMA or ARMA for this? ...The
problem
is I don't think I'm dealing with time series because the measurements
of
waiting times (at a particular state i.e. job submission) ain't done at
regular intervals. Could anyone please suggest a model for this?
Thanking you,
RS
[[alternative HTML version deleted]]
______________________________________________
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
===========================================================================================DISCLAIMER
AND CONFIDENTIALITY CAUTION:\ \ This message and ...{{dropped}}