- Library Home /
- Search Collections /
- Open Collections /
- Browse Collections /
- UBC Theses and Dissertations /
- A state-dependent policy for computer system load regulation
Open Collections
UBC Theses and Dissertations
UBC Theses and Dissertations
A state-dependent policy for computer system load regulation Wong, William Chun Mun
Abstract
One of the principal ideas behind multiprogramming is to make more effective use of the system resources. However, in order to avoid excessive interactions among the competing jobs, which will result in general degradation of system performance, the number and composition of jobs in the multiprogramming set should be carefully controlled. This is the function of the load-control policy. A good load-control policy should be adaptive to work load variation. A load-control policy based on a mathematical model has many advantages including: 1) The effect on performance due to system and/or work load changes can be readily analyzed. 2) It is easy to adopt different optimization criteria by choosing suitable objective functions. 3) The load-control policy is adaptive if the objective function is chosen to be a function of the state of the system. In [11], Chanson and Lo described an adaptive load-control policy based on stochastic control theory. The computer system is modelled as a queueing network with two classes of jobs -batch and interactive. The policy determines the number of jobs of each type to be activated for execution at each system state which is defined to be the total number of jobs of each class in the system(those executing plus those waiting to be activated). The objective is to minimize a weighted sum of the number of jobs in the system. It was shown that the mean response time and throughput rate of the system improve significantly over those in the case of no load control. This thesis extends the results in [11]. The model is generalized to "S" classes of jobs instead of two. The system state is also refined by taking into consideration of the jobs waiting to be activated as well as those that are executing. This eliminates the possibility that the computed number of jobs to be activated in a class is less than the number of executing jobs in that class. It is also shown that this policy performs better than the one described in [11] at the expense of higher computational overhead.
Item Metadata
Title |
A state-dependent policy for computer system load regulation
|
Creator | |
Publisher |
University of British Columbia
|
Date Issued |
1984
|
Description |
One of the principal ideas behind multiprogramming is to make more effective use of the system resources. However, in order to avoid excessive interactions among the competing jobs, which will result in general degradation of system performance, the number and composition of jobs in the multiprogramming set should be carefully controlled. This is the function of the load-control policy. A good load-control policy should be adaptive to work load variation.
A load-control policy based on a mathematical model has many advantages including:
1) The effect on performance due to system and/or work load changes can be readily analyzed.
2) It is easy to adopt different optimization criteria by choosing suitable objective functions.
3) The load-control policy is adaptive if the objective function is chosen to be a function of the state of the system.
In [11], Chanson and Lo described an adaptive load-control policy based on stochastic control theory. The computer system is modelled as a queueing network with two classes of jobs -batch and interactive. The policy determines the number of jobs of each type to be activated for execution at each system state which is defined to be the total number of jobs of each class in the system(those executing plus those waiting to be activated). The objective is to minimize a weighted sum of the number of jobs in the system. It was shown that the mean response time and throughput rate of the system improve significantly over those in the case of no load control.
This thesis extends the results in [11]. The model is generalized to "S" classes of jobs instead of two. The system state is also refined by taking into consideration of the jobs waiting to be activated as well as those that are executing. This eliminates the possibility that the computed number of jobs to be activated in a class is less than the number of executing jobs in that class. It is also shown that this policy performs better than the one described in [11] at the expense of higher computational overhead.
|
Genre | |
Type | |
Language |
eng
|
Date Available |
2010-05-17
|
Provider |
Vancouver : University of British Columbia Library
|
Rights |
For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.
|
DOI |
10.14288/1.0051895
|
URI | |
Degree | |
Program | |
Affiliation | |
Degree Grantor |
University of British Columbia
|
Campus | |
Scholarly Level |
Graduate
|
Aggregated Source Repository |
DSpace
|
Item Media
Item Citations and Data
Rights
For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.