Results 1 to 5 of 5

Math Help - A Problem in Queuing Theory

  1. #1
    Junior Member
    Joined
    Nov 2009
    Posts
    57

    A Problem in Queuing Theory

    A company gets an average of five calls an hour from prospective clients. It takes a company representative an average of twenty minutes to handle one call (the distribution of time to handle one call is exponential). A prospective client who cannot immediately talk to a representative never calls again. For each prospective client that talks to a representative the company makes one thousand dollars. How many representatives should the company maintain if each is paid ten dollars an hour?

    Thanks for your help.
    Follow Math Help Forum on Facebook and Google+

  2. #2
    MHF Contributor

    Joined
    Aug 2008
    From
    Paris, France
    Posts
    1,174
    Quote Originally Posted by akbar View Post
    A company gets an average of five calls an hour from prospective clients. It takes a company representative an average of twenty minutes to handle one call (the distribution of time to handle one call is exponential). A prospective client who cannot immediately talk to a representative never calls again. For each prospective client that talks to a representative the company makes one thousand dollars. How many representatives should the company maintain if each is paid ten dollars an hour?

    Thanks for your help.
    You have to write the problem in terms of a Markov process, and look for an invariant distribution.

    Using the memoryless property of exponential random variables, the fact that if X\sim\mathcal{E}(\lambda), Y\sim\mathcal{E}(\mu), then Z=\min(X,Y)\sim\mathcal{E}(\lambda+\mu) and P(Z=X)=\frac{\mu}{\lambda+\mu}, you can get the following (up to possible errors).

    Suppose there are N representatives, that the arrivals of clients are given by a Poisson process of parameter \lambda and that the length of a call is an exponential random variable of parameter \mu. The queuing process is described by a Markov process (X_t)_{t\geq 0} on state space \{0,\ldots,N\}, corresponding to the number of simultaneously busy representatives, with X_0=0, holding times at k=0,\ldots,N-1 being exponential with parameter q(k)=\lambda+k\mu and with parameter q(N)=N\mu at N, and transition probabilities between "neighbours" only, given by p_{k,k+1}=\frac{\lambda}{\lambda+k\mu}=1-p_{k,k-1}.

    If \pi is the invariant distribution, then the average number of people on the phone per amount of time is the expectation of \pi. This leads to the answer.
    Follow Math Help Forum on Facebook and Google+

  3. #3
    Junior Member
    Joined
    Nov 2009
    Posts
    57
    My first question is: Why do you assume that the arrival of customers is described by a Poisson process? Apart from a remote allusion ("five calls an hour from prospective clients"), nothing in the problem outline gives an idea about the distribution of the client calls.

    Is this a common assumption in queuing theory problems? True that once such an assumption is made (clearly the most manageable since the inter-arrival times are exponentially distributed, and we can then use its memoryless property), the problem becomes much more manageable. I guess this relates more to the general way of approaching queuing problems.

    Thank you for your answer.
    Last edited by akbar; January 27th 2010 at 02:34 PM.
    Follow Math Help Forum on Facebook and Google+

  4. #4
    MHF Contributor

    Joined
    Aug 2008
    From
    Paris, France
    Posts
    1,174
    Quote Originally Posted by akbar View Post
    My first question is: Why do you assume that the arrival of customers is described by a Poisson process?
    I don't know one queuing problem where arrival times don't make a Poisson process. Anyway, how did you expect to find an answer without knowing the distribution of arrivals?? If all the phone calls arrive at the same time, you'll obviously need more representatives than if the phone calls are evenly spaced. Thus, without an implicit assumption, the problem wouldn't be manageable at all... this should ring bells
    Follow Math Help Forum on Facebook and Google+

  5. #5
    Junior Member
    Joined
    Nov 2009
    Posts
    57
    I certainly didn't expect to solve the problem without an assumption, but maybe another (just as realistic) distribution would have been possible. Nothing in the problem outline prevented it.

    Quote Originally Posted by Laurent View Post
    I don't know one queuing problem where arrival times don't make a Poisson process.
    Knowing from an expert that one can confidently make such assumption in general problems in Queuing Theory is particularly helpful. I haven't done as many problems involving queues as you certainly did, but this was the first one I came across without any reference to the distribution of arrivals. This is the reason of my doubts.

    In any case, the idea of finding an "equilibrium" distribution on the states of representatives makes perfectly sense.

    Thanks a lot for your wisdom .
    Follow Math Help Forum on Facebook and Google+

Similar Math Help Forum Discussions

  1. Queuing Theory problem, M/M/1/K queue with twist
    Posted in the Advanced Statistics Forum
    Replies: 2
    Last Post: July 12th 2011, 10:05 PM
  2. a problem with queuing theory
    Posted in the Advanced Statistics Forum
    Replies: 0
    Last Post: April 21st 2010, 01:52 AM
  3. Queuing theory problem SOS
    Posted in the Advanced Statistics Forum
    Replies: 0
    Last Post: April 17th 2010, 06:56 PM
  4. Queuing theory
    Posted in the Advanced Statistics Forum
    Replies: 0
    Last Post: October 10th 2009, 01:37 AM
  5. Queuing theory
    Posted in the Statistics Forum
    Replies: 0
    Last Post: September 25th 2009, 05:46 AM

Search Tags


/mathhelpforum @mathhelpforum