11.9 Exercises


  1. Show that in a G/G/1 queue, the average number of customers at the server is equal to the utilization of the server.

  2. Derive Eqs. (11.3.9) and (11.3.11) using the Generalized Birth-Death theorem.

  3. Derive the average waiting time for M/M/1 from Eq. (11.7.30).

  4. Consider two Web clusters, A and B. Cluster A has n servers and cluster B has m (m > n) servers. Requests arrive at each cluster at a rate of l requests/sec. A load balancer in front of each cluster evenly distributes the requests to each server in the cluster. The average service time of a request in cluster A is x seconds and the average service time of a request in cluster B is k x x where k > 1. The service time of a request in either cluster has an arbitrary distribution. Derive an expression for the value of m so that the average response of a request in cluster A is the same as in cluster B.

  5. A computer system receives requests from a Poisson process at a rate of 10 requests/sec. Assume that 30% of the requests are of type a and the remaining are of type b. The average service times and the coefficients of variation of the service times for these two types of requests are: E[Sa] = 0.1 seconds, graphics/309fig01.gif, E[Sb] = 0.08 seconds, and graphics/309fig03.gif. Compute the average response time for each type of request under each of the following scenarios: 1) requests of type a and b have equal priorities, 2) requests of type a have non-preemptive priority over requests of type b, 3) requests of type b have non-preemptive priority over requests of type a, 4) requests of type a have preemptive priority over requests of type b, and 5) requests of type b have preemptive priority over requests of type a.

  6. Consider the class 3 requests in Example 11.7 when the server uses a preemptive resume scheduling policy (see Section 11.6.2). It is stated the performance (i.e., the waiting time) of the highest priority requests (i.e., class 3 in this case) is not affected by the lower priority requests. Prove this statement by computing the waiting time of class 3 requests using vanilla M/G/1 results (i.e., from Section 11.4). Compare the result to the value computed in Section 11.6.2.



Performance by Design. Computer Capacity Planning by Example
Performance by Design: Computer Capacity Planning By Example
ISBN: 0130906735
EAN: 2147483647
Year: 2003
Pages: 166

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net