=Paper=
{{Paper
|id=None
|storemode=property
|title=Issues of Model-Based Distributed Data Processing: Higher Education Resources Evaluation Case Study
|pdfUrl=https://ceur-ws.org/Vol-848/ICTERI-2012-CEUR-WS-paper-25-p-147-154.pdf
|volume=Vol-848
|dblpUrl=https://dblp.org/rec/conf/icteri/CherednichenkoYL12
}}
==Issues of Model-Based Distributed Data Processing: Higher Education Resources Evaluation Case Study==
https://ceur-ws.org/Vol-848/ICTERI-2012-CEUR-WS-paper-25-p-147-154.pdf
Issues of Model-Based Distributed Data Processing:
Higher Education Resources Evaluation Case Study
Olga Cherednichenko1, Olga Yangolenko1, and Iryna Liutenko1
1
National Technical University “Kharkiv Polytechnic Institute”, Frunze st. 21,
61002 Kharkiv, Ukraine
{marxx75, olga_ya26, chliv_68}@mail.ru
Abstract. Higher education resources are considered in this paper as a complex
heterogeneous hierarchical system. The formal mathematical models for
resources evaluation are suggested. The analysis of information and
communication technologies applied in higher education establishments is
conducted. The necessity of distributed hardware and software infrastructure for
data storage and processing in the evaluation activities is shown and
substantiated. The distributed data storage and processing architecture is
presented.
Keywords. Distributed data processing, evaluation, information system,
quality, higher education resources.
Key Terms. ProcessPattern, FormalMethod, Model, SoftwareComponent.
1 Introduction
Nowadays education quality is a crucial factor of both competitiveness of higher
education establishments (HEE) and success of HEE graduates in their careers. That’s
why assessment of education quality is an urgent problem for management activities.
The adequate education quality estimates can be the basis for correct management
decisions and for realization of improvements. Moreover the estimates of higher
education quality may be important for employers and for potential applicants as the
main stakeholders of the system of higher education.
Education quality assessment is directly connected with education quality model.
Various quality models are used in HEEs around the world [1]. Some of them are
oriented on possibilities and results of business processes of HEE. For example,
EFQM Excellence Model adopted for higher education introduces two groups of
criteria: enablers and results [2]. Enabling criteria cover what the organization does,
and the results criteria cover what the organization achieves.
As a rule, HEE plans and manages internal resources in order to support its policy
and strategy, and the effective operation of its processes. Resources are the means that
148 O. Cherednichenko, O. Yangolenko, and I. Liutenko
provide HEE’s functioning. During planning and managing of resources HEE
balances its current and future needs.
There are basic and supporting processes in HEE. Educational process is the
process which realizes the main function of HEE. It supports university’s activities
intended for delivering knowledge and skills to students. The resources of educational
process include the staff, material and technical facilities and courseware. For
example, to teach students some subject, first of all, it is necessary to appoint
competent lecturer and a tutor. Secondly, some well-equipped room for lessons
should be found. Handbooks, guidelines and other courseware are necessary for
representation of teaching material. And finally, to support the educational process
such information resources as curricula, announcements and other useful information
should be available in a convenient way, for example via the local network or the
Internet.
The educational process is justified by academic curricula for different
qualifications, such as bachelor and master. The basic unit of academic curriculum is
a discipline (fig. 1). Every discipline is described by its syllabus. Syllabus determines
all necessary resources for discipline teaching. All HEE’s resources are distributed
among different units and are allocated on vast territories (fig. 2). Moreover, some
resources are in common usage of HEE’s units.
Fig. 1. Discipline structure.
Fig. 2. Resources assessment
Issues of Model-Based Distributed Data Processing … 149
So teaching disciplines requires large amount of various resources that are shared,
commonly used and accounted by different departments of HEE. Under these
conditions assessment of resources that support educational process seems to be a
complex problem. The solution of this problem implies collecting data from different
sources, its storing, processing and taking decisions about resources updating.
The rest of this paper is organized in the following way. Section 2 classifies
information systems in HEE and characterizes mathematical methods used for their
realization. Three mathematical patterns that support resources quality assessment are
considered in Section 3. Section 4 represents discussion about distributed and
centralized data storage and processing. Grid-based architecture is suggested. Section
5 presents conclusions and prospect on future work.
2 Analysis of Information and Communication Technologies in
Higher Education
The HEE’s functioning is based on various processes, therefore HEE has different
information systems (IS) that support these processes. For example, there can be
distinguished information systems of administrative and financial management,
educational process management and support, scientific research management,
information resources management [3]. HEEs use either commercial IS or those
which are elaborated by universities for their particular needs [3]. Often commercial
solutions do not take into account all peculiarities of definite HEE’s functioning, so
they can’t satisfy all requirements. To elaborate its own software HEE usually has a
lack of resources and highly qualified staff. However both types of IS are used and
the problem is to integrate them and to provide efficient interaction between solutions
of different providers.
The work of all university’s IS is based on mathematical models used for solving
management problems. Problems of data accounting are managed by means of data
bases (DB). Queries to DBs are formed with the help of relational algebra [4]. Data
stored in DBs is processed basing on statistical methods which include correlation,
variance, discriminant, factor, cluster and other kinds of analyses [5]. Since testing is
an important and specific educational problem, processing of testing results is
supported by Classical Test Theory [6] and Item Response Theory [7]. HEEs rating
and various assessment problems are solved with the help of expert methods [8].
As we can see there are different tasks that should be automated in the university.
Different mathematical models and IS are used to solve these tasks. All existing and
elaborating IS are intended to the common purpose – to improve university’s
functioning. From the management point of view all IS should exist in the common
informational space with distributed data storage and processing. So we can make the
conclusion that it is necessary to develop distributed hardware and software
infrastructure composed of heterogeneous resources owned and shared by multiple
administrative units which are coordinated to provide transparent, dependable and
consistent computing support to a wide range of applications.
Since quality can be considered as one of the main goals of HEE’s management,
the resources evaluation subsystem has to be a component of HEE’s software
150 O. Cherednichenko, O. Yangolenko, and I. Liutenko
infrastructure. Resources quality influences higher education quality directly. Our task
is to extend the functionality of existing IS by implementing resources evaluation.
Quality monitoring and evaluation (M&E) is a part of the management process.
There are two basic types of monitoring: implementation-focused and results-based
[9]. Implementation-focused M&E is oriented on inputs, activities and outputs of
system’s functioning. Results-based M&E focuses on goals and results giving the
evidences and explanations of existing tendencies. Our research is oriented on results-
based M&E. In the given work we consider the evaluation process which provides
management with necessary estimates towards defined management outcomes.
3 Model-Based Resources Evaluation
We consider the system of university’s resources as a complex heterogeneous
hierarchical structure, which has a large number of parameters. HEE resources are
used, controlled and evaluated by different departments and units. We found
distributed structure of data sources providing partial estimates of resource units. We
distinguish three basic tasks connected to the university’s resources M&E. They are
resources comprehensive quality assessment, internal licensing audit and resources
usage performance evaluation. These tasks are interconnected and related to quality
assessment from different points of view.
Higher education licensing, i.e. the process of granting permissions to provide
certain educational services, is an important process of public administration. The
licensing process is carried out on a regular basis (the license validity is limited), it
requires processing of large data volumes, and supposes that the information related
to HEE is available for public access. This information in particular may include
curricula, syllabus, university infrastructure, etc. The internal audit is required to
prove the sufficiency of existing resources and their quality to obtain license. During
the internal licensing audit HEE is estimated according to the license conditions [10].
Resources usage performance can be evaluated through the set of indicators –
different for each type of resources. For example, performance of capital resources is
measured via loading factors, profitability ratio, usage intensity coefficients. The
efficiency of human resources usage is determined by labor performance, annual
number of workers, economy or excess expenditure of salaries.
The main goal of management in university is improvement of education quality.
Resources of the university provide the possibility to reach this goal. Updating
university resources is the most realistic way for improvement of education quality.
Therefore resources comprehensive quality assessment is the basis of continuous
quality improvement. It allows finding the resources that have to be updated, to
evaluate their influence on educational process, and to elaborate resources updating
projects.
Partial and aggregate indices are used for decision making in all mentioned tasks.
Therefore M&E IS must have the model of transition from a set of partial to aggregate
estimates. To solve this problem we suggest applying Resources Network System
(RNS) [11].
Issues of Model-Based Distributed Data Processing … 151
The structure of evaluation system can be presented as an oriented acyclic graph
G k (K , A k ) , where K is a set of nodes, A k (a ij ) determines the directions of
arcs, that connect the nodes. The given graph has two types of nodes: node-entries
~
K 0 k j K : k i K , a ij 0 and node-aggregates K k j K : k i K , a ij 1 .
Node-entries reflect the partial estimates of resources. The aggregation logic is
defined by particular task. Each aggregate is associated with some composite
function, or estimator.
In this paper we pay our attention to the task of resources comprehensive quality
assessment. We consider direct and reverse tasks of comprehensive assessment. The
direct task is the computation of comprehensive assessment value on the assumption
of known values of estimators. The reverse task is the determination of values for
node-entries on the assumption of user-defined comprehensive assessment.
The ratio scale is used for both types of estimates. We suggest using Qualimetry
Theory (QT) for partial estimates computing. The QT provides generalized principles
of quantitative assessment of quality of objects of any nature [12]. That’s why we
choose the method of qualimetry to develop node-entries assessment methodology.
To aggregate partial estimates we use the following convolutions as estimators:
weighted average arithmetical, weighted average geometric, weighted guaranteed
result, and weighted dominating result index. We use different types of estimators
because of heterogeneity of resources, different influence of various resources types
on the comprehensive resources quality. These convolutions require experts’
judgments.
We use the QT-based approach for partial estimates calculation which includes the
following steps.
1. Situation assessment is defined, i. e. description of the conditions and goals of the
assessment, application of those estimates are identified.
2. The properties tree is constructed. It reflects the hierarchically ordered set of
features of the object and allows to fully describe its quality. The procedure of
properties tree construction is based on a number of claims detailed in the QT [12].
3. For each simple and some complex features it is necessary to assign appropriate
indices. Every index is associated with measurement scale, reference and rejection
values. The absolute values of all indices must be converted into relative ones.
4. Weighted coefficients are calculated. We suggest defining weighted coefficients
using pairwise comparison. The weighted coefficients calculation is based on the
method of eigenvector [13].
5. Partial quality assessment is calculated based on one of the weighted average
methods. We use weighted coefficients and relative values of indices obtained on
the previous step.
Thus, all resources, that support educational process in HEE, are associated with an
oriented graph. Every node of this graph reflects individual or group estimates of
resources. We propose pattern-based data processing and distinguish the following
patterns.
1. The QT-based Partial Assessment Technique (PAT) is a pattern for assessment of
separate resources units. Based on this pattern the evaluation procedure is formed,
which results in partial estimate.
152 O. Cherednichenko, O. Yangolenko, and I. Liutenko
2. The Comprehensive Assessment Technique (CAT) defines the grouping of
estimates of resources units from different points of view. We suggest to represent
CAT as a graph which node-aggregates are associated with one of the convolutions
mentioned above.
3. Weighted coefficients Calculation Technique (WCT) provides expert judgment
method based on pairwise comparisons. This pattern includes the procedure of
processing pairwise comparisons and calculation of weighted coefficients vector.
Resources evaluation is considered on three stages. They are pre-processing, data
processing and interpretation of results. These stages are repeated for each evaluation
task. The pattern is chosen depending on the task that has to be solved. Initial data
sources are determined by RNS. Pre-processing stage includes data extraction from
initial sources, collection of required data and its transformation. Data processing
stage is totally defined by the chosen pattern. On the results interpretation stage the
obtained estimates must be explained from the point of view of the considered task.
Based on the above discussion we suggest creating the resources evaluation
software.
4 Architecture of Resources Evaluation IS
Currently the two common ways of data storage are centralized and distributed
approaches. Centralized data storage applies that a DB is located at a single server.
The data access is executed with the help of remote query or transaction. This is the
simplest approach for realization and maintenance. The disadvantage is that the DB
would be unavailable for remote clients if connection errors occurred. Also database
located on one server has a limited volume of memory. Since all queries are sent to a
single server, there are some obvious constraints in time delays and costs of
connection support.
Distributed approach supposes that data is stored on multiple servers. This allows
increasing of DB volume. Since many queries are satisfied by local DBs, the response
time and the costs of query processing are decreased while availability and reliability
are increased. The disadvantage of distributed data storage is that some queries and
transactions may need the access to all servers which increases the response time and
the costs. Another issue is that all clients have to have information about data location
in distributed DB. This approach is compatible with common usage of local and
global networks.
Distributed data storage and processing provides efficient work with rapidly
changing information that is used by various clients. It supports a large number of
cooperating clients which collect, register, store and deliver information. Distributed
data storage prevents servers from overloading by distributing data among different
computers. Also it provides an access to a huge volume of information stored in the
system.
One of the most convenient approaches of distributed data storage and processing
is grid technology. There are three types of grid systems: computational, data and
network [14]. A computational grid has the processing power as the main computing
resource shared among its nodes. A data grid has the data storage capacity as its main
Issues of Model-Based Distributed Data Processing … 153
shared resource. Such grid can be regarded as a massive data storage system built up
from portions of a large number of storage devices. Network grid has as its main
purpose to provide fault-tolerant and high-performance communication services.
In this work we suggest using data grid architecture for M&E software in HEE
(fig. 3). Data stored by various departments and units is shared through the Internet or
local network. Data processing is conducted by application servers available for
different users via their desktop applications. M&E software services are located on
application servers. Special management system is required to control, coordinate and
synchronize distributed data storage and processing.
The PAT, CAT and WCT services are based on the corresponding patterns
described above. The data collection service provides tools for data extraction from
distributed database. These services can be used by quality evaluation software. If we
need to evaluate resources quality, this software must store RNS model and data
sources for partial estimates. Such software must support pre-processing and
interpretation stages. On the data processing stage it calls the necessary services.
We suppose that Open Grid Services Architecture is the suitable solution for
distributed hardware and software infrastructure. This will be a base for university’s
IS integration.
Fig. 3. Distributed data storage and processing architecture.
5 Conclusions and Future Work
We considered a case study of comprehensive resources quality assessment. Higher
education resources have distributed and heterogeneous structure. This leads to the
154 O. Cherednichenko, O. Yangolenko, and I. Liutenko
idea of distributed data storage and processing. The given paper generalizes methods
of quality assessment into several patterns. The suggested approach allows quality
assessment of heterogeneous objects and has comprehensive facilities for construction
of quality measurement tool. The architecture of information system of resources
evaluation is based on distributed data grid architecture. Such approach may be
generalized for evaluation of different objects, not only resources and not only in
higher education establishment.
Our future work includes elaboration of formal models for solving two remaining
tasks – internal licensing audit and resources performance evaluation. The further
researches will be devoted to detailed architecture design and software construction.
References
1. Existing models of education establishments’ quality systems,
http://ru.ict4um.edu.ru/lib/euro/model (in Russian)
2. EFQM Excellence Model. Higher Education Version 2003. Centre for Integral Excellence,
Sheffield Hallam University (2003)
3. Krukov, V.V., Shahgeldian, K.I.: Corporate Information Environment of University:
Methodology, Models, Solutions. Dalnauka, Vladivostok (2007) (in Russian)
4. Date, C.J.: An Introduction to Database Systems. Addison Wesley (2003)
5. Ross, S.M.: Introduction to Probability and Statistics for Engineers and Scientists. Elsevier
Academic Press , London (2004)
6. Steyer, R.: Classical (Psychometric) Test Theory, http://www.metheval.uni-
jena.de/materialien/ publikationen/ctt.pdf
7. Reeve, B. An Introduction to Modern Measurement Theory, http://www.
http://appliedresearch.cancer.gov/areas/cognitive/immt.pdf
8. Meyer, M. A., Booker, J. M.: Eliciting and Analyzing Expert Judgment: A Practical Guide.
SIAM, Philadelphia (2001)
9. Kusek, J.Z., Rist, R.C.: Ten Steps to a Results-Based Monitoring and Evaluation System: a
Handbook for Development Practitioners. The World Bank, Washington, DC (2004)
10.Cherednichenko, O., Kuklenko, D., Zlatkin, S.: Towards Information Management System
for Licensing in Higher Education: An Ontology-Based Approach. In: Mayr, H.C.,
Karagiannis, D. (eds.) Information Systems Technology and ist Applications 6th
International Conference ISTA 2007. LNI, 84, pp. 33--42. GI, Bonn (2007)
11.Cherednichenko, O., Timchenko, K., Liutenko, I.: Technology of Quality Comprehensive
Assessment (for Resources in the University by Example). In: Vestnik of Kherson National
Technical university, vol. 2 (41), pp. 451--455. KNTU, Kherson (2011) (in Russian)
12.Azgaldov, G.G.: Theory and Practice of Goods Quality Assessment (Basics of Qualimetry).
Economics, Moscow (1982) (in Russian)
13.Saaty, T., Vargas, L.: Decision Making with the Analytic Network Process. Economical,
Political, Social and Technological Applications with Benefits, Opportunities, Costs and
Risks. Springer (2006)
14.Hose, K., Schenkel, R.: Distributed Data Systems: Introduction, http://www.mpi-
inf.mpg.de/departments/d5/teaching/ws10_11/dds/slides/DDS-1-print.pdf