A 56 Kbps communication link is used to transmit 1500-byte long packets. What is the service demand of a packet on that link? Does the service demand of the packet change with the traffic on the link?
A 4-CPU machine is used to run a CPU-intensive benchmark that consists of several programs running concurrently. The programs in the benchmark are equivalent with respect to their use of CPU. What kind of QN model is most appropriate to model this situation? Specify the number of classes, their type (open or closed), the queues, and their types. Justify your answer.
A computer system supports a transaction workload submitting requests at a rate of 10 tps and a workload of requests submitted by 50 client workstations. What kind of QN should be used to model this situation: open, closed, or mixed?
A database server has two identical disks. The service demands of database transactions on these disks are 100 msec and 150 msec, respectively. Show how these service demands would change under the following scenarios:
Disk 1 is replaced by a disk that is 40% faster.
Enough main memory is installed so that the hit rate on the database server's cache is 30%.
The log option of the database management system is enabled. A log record is generated on disk 2 for each update transaction. Updates account for 30% of the transactions and recording a log record takes 15 msec.
The workload of a database server is decomposed into four types of transactions: trivial (TT), medium (MT), complex (CT), and update transactions (UT). Table 2.6 shows the arrival rates and service demands for each class.
Table 2.6. Parameters for a Four-Class Performance Model
Class
TT
MT
CT
UT
Arrival rate (tps)
0.20
0.30
0.20
0.10
Service demand (sec)
CPU
0.15
0.30
0.45
0.70
Disk 1
0.20
0.35
0.55
0.30
Disk 2
0.10
0.30
0.60
0.20
Capacity planners are interested in answering the following questions:
What is the effect on response time of TT transactions if their arrival rate increases by 50%?
What is the effect on response time of UT transactions if their arrival rate is increased by 25%?
What is the effect on the response time of TT transactions if UT transactions are run on a different machine?
From this list of questions, it is clear that the model does not need to consider classes MT and CT separately. How would you aggregate these two classes into a single class? In other words, what is the arrival rate and what are the service demands of the new aggregated class?
A delay resource in a QN can be thought as a special case of a load-dependent resource. Give an expression for the service rate m (n) as a function of the number, n, of customers in the resource and of the average service time, S, per customer.