HARDEXP - A STRATEGIC SUPPORT TOOL FOR HARDWARE
EXPANSION
NESTOR MEJIA
INTERCONEXION ELECTRICA S.A.
AA. 8915 3
MEDELLIN, COLOMBIA
ISAAC DYNER
UNIVERSIDAD NACIONAL DE COLOMBIA.
AA 1027
MEDELLIN, COLOMBIA
ABSTRACT
This paper presents a system developed to design strategies for organizational expansion
based on system dynamics and expert system methodologies. The tool was especially
built to plan the expansion of a computing system network.
The prototype developed supports tasks related’ to strategies design, scenarios
generation and system simulation. Examples are exhibited.
1. INTRODUCTION ©
The increase of computational facilities in an organization must be dimensioned and
scheduled in such a way that maximum demand can be satisfied within pre-established
parameters for service quality, maintaining the required reliability indexes.
For an adequate accomplishment of this aim it is required to plan strategically the most
important resources such as: personnel and capital reserves. Large companies know the
value of information resources but under the presumption of low hardware costs they do
not doubt to invest in information technology beyond the present needs. For small
enterprises and for some medium ones it is important to maintain the balance between
the available budget and the actual needs. This, to avoid over-installation expenses.
SYSTEM DYNAMICS '93 349
The purpose.of this work consists in developing.an intelligent system which incorporates
knowledge acquisition and organizational simulations specifically orientated to hardware
expansion planning through a methodology that could be easily adopted to other
problems.
2. PLANNING HARDWARE EXPANSION
The criterion for computer hardware expansion is mainly based on establishing reliable
service levels. The installed computational capacity increments must be dimensioned and
scheduled in such a way that demand is guaranteed within. some quality service standards
and with the desired reliability which must include: devices faults and system
maintenance.
The computational expansion may have two alternatives to satisfy the demand increase:
- Expansion of the existing facilities.
- Incorporation of new technologies.
The previous alternatives oblige to.carry out strategic planning with short horizons since-
technological obsolescence is a critical factor for decisions making in these. matters. In
this study, thirty six months are taken.as a planning horizon.
The. main factors that determine demand satisfaction are. among others maximum time of
system reply, CPU utilization, disks speed and I/O service.
A planning criterion that contemplates high service levels bears high costs of expansion
- and maintenance. It is important to take into account that the adoption of a. very rigid
criterion in terms of high satisfaction levels may be unfeasible from the economic stand
point,
3. CONCEPTUAL MODEL
Nowadays there isa growing interest in. developing decision support systems. There are
developments that. combine models based on knowledge with traditional simulation
models-or system dynamics models.
Levary and Chi (1988) integrates a discréte or continuous simulation model depending.
on the characteristics of the system being simulated and two expert systems: the entrance
expert system verifies the compatibility of the entrance vector and the exit expert system
makes recommendations on design.
Markarian and Koziol (1991) presents a DSS based on system dynamics methods and
expert systems. In this work the system dynamics model was used to simulate the
350 SYSTEM DYNAMICS '93
production processes and to identify possible existing problems. The expert system
diagnoses the possible causes of the problem as well as the possible solutions.
Managerial Support Systems (MSS) are suitable to support planning and decision
processes where uncertainty, sublet experience and different qualitative factors are
considered to be.relevant.
HARDEX? jis a MSS characterized for being a decision making support tool which
incorporates experience to formulate strategies, to build scenarios and to acquire new
knowledge.
The system possesses three main components: an expert system, a system dynamics.
model and a manager interface between the components and the user (Figure 1).
Figure 1. CONCEPTUAL MODEL
EXPERT SYSTEM SIMULATOR
REPORT SCENARIO EXPERIENCE
3.1 THE EXPERT SYSTEM
The-expert system provides elements of judgment in a systematized, controllable and
reproducible form in order to design strategies on hardware expansion. The system
extracts the necessary information and identifies the relevant factors to formulate the
strategy.
‘SYSTEM DYNAMICS '93, 351
Some defined strategies on the knowledge base depend on: minimum. costs,. system
performance, control of disk space, control of main memory use, interactive and batch
priorities and system reliability indexes.
The user is orientated systematically towards one’ of the strategies or to elaborate a
strategy that does not possess an antecedent in the knowledge base. In the last case, the
internal consistence is verify through simulations and then it could be incorporated to the
knowledge base.
For example a strategy orientated towards a good system performance could be obtained
through the following conditions: 3
- The user will wait at most ten seconds fora a system reply.
- The visits of each interactive and batch task are inside of pre-established ranks.
- The use of memory and disk does-not exceed the given parameters.
In the previous case the. system.will recommend with these conditions a strategies based
on high reliability indexes and with no budget restrictions.
3.2 SYSTEM DYNAMICS MODEL
The system dynamics model (Figure 2) is composed by five modules:
~ Users and projects.
-- Budget restrictions.
~ Present capacity installed.
- System performance.
- Expansion of dispositive.
These modules interact to reproduce the scenario that corresponds to the strategy
propounded by the expert system. The system dynamics model uses the decision space
as an entrance vector parameter.
Simulation results show the evolution of the most important variables such as disk and
memory needs and system load.
The user may then evaluate the proposed strategy through simulation results. At this
stage a more adjusted scenario may be built making transformations on the decision
space. Through this process more knowledge is acquired.
352 SYSTEM DYNAMICS '93
Figure 2. CAUSAL DIAGRAM
—_—_Siiiia 4
<
a |
l —}
ne £ DISK
DISK RESERVE
USERS NEEDS ad — > PRESENT nr’ IL o
ee
aes WAXTHUN WAITING rita
TK
— USERS
re . ET
+
Loree
‘SYSTEM
WORKLOAD
INTER-BATCH RELATION
a
arm
| id Hae DISK
HEPAT | He Tabet
SYSTEN
Te ant —_| ae —
THROUGHPUT
|
SYSTEM DYNAMICS '93. 353
3.3 INTERFACE PROGRAM.
The manager program was developed in C language, and it possesses a friendly interface
with the user. Through only one menu the program manages six options:
EXPERT SYSTEM
DYNAMIC SIMULATOR
UPDATE PARAMETERS
SCENARIOS
DECISION SPACE
EXIT
The first option executes the expert system and the result will be a recommended
strategy. The second option can be used to executed Professional Dynamo.
The third option allows to transfer the parameter decision vector produced by the expert
system to the system dynamics model. The fourth option simulates a scenario associated
to a given strategy. The fifth option permits to manipulate the parameter decision vector
to conform a new scenario,
4, STUDY CASES
Figures 3, 4,5 and-6 present simulation results that correspond to two strategies
proposed by the expert system according to hypothetical situations.
First strategy (Figures 3 and 4). The user expansion needs considers no budget
restrictions, normal service and high reliability indexes. Results show a uniform increase
in system load and in system reply. Resources are distributed in time efficiently.
Second strategy (Figures 5 and 6). Budget restrictions and some requirements
“concerning system reply are introduced. Results show how system load and system
throughput are affected in a significant way, indicating that at this point the system could
not attend the demand in an efficient way. Furthermore, more users will not enter the
system for new developments until disbursements will be possible.
For a more extensive evaluation the user will need to manipulate the decision space and
obtain new
scenarios with different variables arrangements.
354 SYSTEM DYNAMICS '93
Figure 3. SYSTEM PERFORMANCE
Strategy with no budget restrictions
nN
o
1-1-1
10
8
6F boa
ar . pang peeedeeren
2
2 woh .
0 5 10 15 20 25 30 35
DEMAND.
QO WORKLOAD + SYSTEM THROUGH © RESPONSE TIME
Figure 4. RESOURCES REQUIRED
- Strategy with no budget restrictions
oO 5 10 1s 20 25 30 35
MONTHS
a piIscs. + MEMORY
SYSTEM DYNAMICS '93
Figure 5. SYSTEM PERFORMANCE
Strategy with restrictions
15
14407
13 7
1207
lof
10 oF
\)
if
OrFPNWAN DIB O
T
(+) 5 10 15 20 25 30 35
DEMAND
© WORKLOAD + SYSTEM THROUGH ° RESPONSE TIME
Figure 6. RESOURCES REQUIRED
Strategy with restrictions
356 SYSTEM. DYNAMICS '93.
5. CONCLUSION
Based on the methodology and on the developed tool, it is possible to assert that:
The interaction of models based on knowledge and system dynamics methods produce
powerful MSS tools.
It is possible the integration of the superficial and deep knowledges in order to help more
efficiently the decision maker.
The problem of the strategical planning of computational recourses is possible to be
solved through
MSS tools.
BIBLIOGRAPHY
FERGIONNE, GUISSEPPI A. 1991. "Decision Technology Systems", Information
Systems Management, V.8(4) FALL, pages 34-43.
LEVARY, REUVEN R. and°CHI Y LIN. 1988. "Hybrid Expert Simulation System
(HESS), Expert Systems, May . Vol. 5, No 2
MARKARIAN, M. and KOZIOL, R. 1991. "System Dynamic: Taming Systems in the
Business World". System Dynamics '91.
SYSTEM DYNAMICS '93 357
CC BY-NC-SA 4.0