"visit calculations" program

[russell kym horsell]

I've been contracted by someone purporting to be associated with
a "department" in the Manchester area, England.

Their project involved writing a simple discrete event simulator (or like)
to calculate how many members of staff would be needed for certain
flows of students on open days.

I have been unable to contact the party for several days now, with
the alleged deadline for the project around 24 hrs off.

If the project (not the scenario ;-) familiar to anyone, please drop me a line.

I've attach the last email I received from the party concerned.

===
$ Reply-To: Matt Nord <[email protected]>
$ Date: Aug 8, 2005 2:50 AM
$ Subject: Re: Project
$ Kym, This is the exact brief I have:
$
$ On visit days to the department (for potential applicants to the course), it
$ would be advantageous to have sufficient staff on duty to ensure that most
$ of the time (say > 90 per cent), that a visitor does not have to wait to
$ talk to a member of staff. This situation can be simulated using random
$ number generators within GSL.
$
$ If you can make it so that we can manually enter the number of visitors
$ expected each hour and then the program would give the number of staff
$ needed to ensure the 90% not waiting figure. I would use 15 mins as the time
$ allowed to each person and assume the visitors arrive uniformly over each
$ hour.
===

 

[osmium]

I've been contracted by someone purporting to be associated with
a "department" in the Manchester area, England.

Their project involved writing a simple discrete event simulator (or like)
to calculate how many members of staff would be needed for certain
flows of students on open days.

I have been unable to contact the party for several days now, with
the alleged deadline for the project around 24 hrs off.

If the project (not the scenario ;-) familiar to anyone, please drop me a
line.

I've attach the last email I received from the party concerned.

===
$ Reply-To: Matt Nord <[email protected]>
$ Date: Aug 8, 2005 2:50 AM
$ Subject: Re: Project
$ Kym, This is the exact brief I have:
$
$ On visit days to the department (for potential applicants to the
course), it
$ would be advantageous to have sufficient staff on duty to ensure that
most
$ of the time (say > 90 per cent), that a visitor does not have to wait to
$ talk to a member of staff. This situation can be simulated using random
$ number generators within GSL.
$
$ If you can make it so that we can manually enter the number of visitors
$ expected each hour and then the program would give the number of staff
$ needed to ensure the 90% not waiting figure. I would use 15 mins as the
time
$ allowed to each person and assume the visitors arrive uniformly over
each
$ hour.
===

The assumption that visitors arrive uniformly means that this is certainly
*not* a simulator, it is a simple calculation.
If you do this, you should warn the recipient that he has specified things
such that the whole problem of queuing is obviated by the wording of the
specification. GIGO, the results will be useless.

 

[russell kym horsell]

The assumption that visitors arrive uniformly means that this is certainly [...]
*not* a simulator, it is a simple calculation.
If you do this, you should warn the recipient that he has specified things
such that the whole problem of queuing is obviated by the wording of the
specification. GIGO, the results will be useless.

The math is simple, but not entirely trivial. In most simple queue systems
everything's (i.e. arrivals and server times) considered Possion -- the most
tractable case.


With uniform distro's you can use the conf intervals of the Binomial
distro, if you ignore regions near the start/end points or other
discontinuities.


Still, the problem is not the problem.