|
|
|
|
|
|
|
Figure 9.18
A generalized CPU-I/O model. |
|
|
|
|
|
|
|
|
Figure 9.19
Simplified model. |
|
|
|
|
|
|
|
|
but it is clear that the bottleneck in the system is the server CPU, since it has the least capacity. For initial analysis purposes, we treat the system as a single server (the server CPU) with a delay center of 22.4 ms to represent the disk plus network time (Figures 9.18 and 9.19). Each user has a timing relationship as shown: |
|
|
|
|
| | | | | | | | |
|
|
|
|
960 ms (User + Pause + delay) |
|
|
|
| | | | | | |
|
|
|
|
|
|
We will look at the system performance for various numbers of workstations from n = 10 to n = 50. We will use the asymptotic performance model. Let us assume the following c2: |
|
|
|
|
| Unit | c2 | | Server CPU | 0 | | Disk | 0.5 | | Network | 0 |
|
|
|
