Available via license: CC BY-ND 4.0
Content may be subject to copyright.
801
MODELS AND SOLUTIONS FOR THE IMPLEMENTATION OF
DISTRIBUTED SYSTEMS
Ghencea Adrian
Titu Maiorescu University, Faculty of Economics
Vătuiu Teodora
Titu Maiorescu University, Faculty of Economics
Ț
ȚȚ
Țarcă Naiana
University of Oradea, Faculty of Economics
Software applications may have different degrees of complexity depending on the problems they
try to solve and can integrate very complex elements that bring together functionality that
sometimes are competing or conflicting. We can take for example a mobile communications
system. Functionalities of such a system are difficult to understand, and they add to the non-
functional requirements such as the use in practice, performance, cost, durability and security.
The transition from local computer networks to cover large networks that allow millions of
machines around the world at speeds exceeding one gigabit per second allowed universal access
to data and design of applications that require simultaneous use of computing power of several
interconnected systems. The result of these technologies has enabled the evolution from
centralized to distributed systems that connect a large number of computers. To enable the
exploitation of the advantages of distributed systems one had developed software and
communications tools that have enabled the implementation of distributed processing of complex
solutions. The objective of this document is to present all the hardware, software and
communication tools, closely related to the possibility of their application in integrated social
and economic level as a result of globalization and the evolution of e-society. These objectives
and national priorities are based on current needs and realities of Romanian society, while being
consistent with the requirements of Romania’s European orientation towards the knowledge
society, strengthening the information society, the target goal representing the accomplishment
of e-Romania, with its strategic e-government component. Achieving this objective repositions
Romania and gives an advantage for sustainable growth, positive international image, rapid
convergence in Europe, inclusion and strengthening areas of high competence, in line with
Europe 2020, launched by the European Council in June 2010.
Keywords: information society, databases, distributed systems, e-society, implementation of
distributed systems
Cod JEL: O33, M15, L86
Introduction
The concept of “information society” is a very generous program that covers all sectors of
government. The main goal is to create a society that includes all citizens’ access to public
services provided in electronic form, by increasing the capacity to use information society
services, reforming the government operational models and increasing operational efficiency
through appropriate use of information and communication technologies and the increase of
competitiveness of the business through advanced use of ICT - Information and communications
technologies.
The complexity of software applications is an essential property that derives from: the
complexity of the problem domain, managing the development process, the degree of flexibility
allowed by the software and problems that characterize the behavior of discrete systems.
802
Since the last decades of the previous century growth rate has been calculated mainly by tricks,
by dividing tasks in the computer system, by introducing devices interrupt requests from the
input / output or direct memory access (Lungu, Vătuiu and Fodor, 2006: 45). Then specialized
systems for digital imaging systems appeared that sought to compensate for insufficient speed
through parallel processing, allocating each pixel in an image line of each unit of computing - a
processor dedicated local operations from the image. This is how the first configurations of
parallel computing and the first parallel algorithms appeared. One aspect of distributed
processing which has been subject to attention in recent years refers to an environment in which
they can engage cycles and spare storage space of the CPU tens, hundreds and thousands of
networked systems to work on a specific problem that involves a very high processing capacity.
Developing these processing models was limited however by the lack of attractive broadband
connections and problems, combined with the real challenges of security, management and
standardization. A distributed system can be defined as a group of independent computers that
the user perceives as a coherent unit.
Fig. 1 Distributed storage and query of arbitrary data
A distributed system must not be built only because it can be built. There are many different
types of distributed processing systems and there are quite a few challenges to overcome to
successfully design such a system. Distributed processing systems aim to connect users and
resources in a transparent, open and scalable way (Vătuiu and Popeangă, 2006: 134). It would be
ideal if this system would show more tolerance for errors and also if it would be more powerful
than any combination of independent computer systems.
1.Parallel processing systems
Parallel processing is the simultaneous execution of the same instructions (divided and especially
adapted) on multiple processors in order to obtain faster results. The idea is based on the fact that
the process of solving a problem can usually be divided into smaller instructions that can be
achieved simultaneously if they are coordinated. A parallel processing system is a computer that
is equipped with more than one processor for parallel processing (Lungu and Ghencea 2011:122).
Also new multicore processors are parallel processing systems. There are several types of parallel
processing computers. They are differentiated by the type of interconnection between processing
(known as "processing elements"), between processors and memory. Besides the classification
made by Flynn who consider the type of processor instructions executed, there is a classification
that is based on how memory is constructed: parallel processing computers that have distributed
memory have multiple processors which access all the available memory like an addressing
global space; computers which processes in parallel and have distributed memory with multi-
803
processors, but each processor can only access local memory - no global space for memory
access between them.
Fig. 2 Parallel processing systems
1.1.Implementation of parallel processing for different computer systems
Parallel processing, defining two or more processors that execute concurrent processes and acts
as a single unit, allows one to run complex applications.
Most management systems relational databases are currently being upgraded to take advantage of
parallel processing in heterogeneous systems, and to allow complex run of mission-critical
applications. The optimal distribution of data is difficult in terms of technology; this process can
be strongly dependent on the requirements for ensuring a good response to requests, and ensuring
data integrity, continuous availability, interoperability, etc.
Management systems of modern databases use a series of abstract concepts and strategies to meet
the requirements associated with current applications. The transaction, for example, can be used
in the distributed data network to ensure the passage of groups of data and associated operations
from a client post to server or from one server to another. Most of the producers of databases
have monitors for the transaction process (TP - Transaction Processing) which represent
advanced tools to manage distributed transactions in heterogeneous networks (Cristea 2007:44).
The standard method of communication between TP monitors and database systems has been
accepted as Protocol X / A as part of a group of standards X / Open. Currently, Sybase System 10
servers and Oracle 7 supports protocol X / A, while indirectly Informix Online adhere to this
protocol through an ancillary product, Informix TP / XA.
In terms of ensuring data integrity in systems client / server distributed management software
vendors have addressed these databases strategies:
- The "two-phase commit" technique: all changes required by a transaction on a database are
either committed (the execution of the transaction is completed) or void, with the return of
the database to the previous state. This strategy is not suitable for complex heterogeneous
networks in which the probability of failure in any node is large and mission-critical
systems. It is used as a way to ensure that all servers hold identical copies of the database
at any time;
804
- The data replication strategy is now a solution adopted by Oracle, Sybase and Informix.
Replication is a process in which multiple servers hold many identical copies of a database.
The replication strategy differs significantly from the database "two phase commit" in that
it guarantees the identity of copies of distributed databases only at certain times or under
certain conditions. The data replication technique used by the Oracle Server 7 is called
"Table Snapshots" by which the central server (master) copies at certain moments of time
only those parts of the database that have changed, then disseminating these changes in the
network. The replication mechanism used by the server is Informix database with the
method "snapshots" used by Oracle. Informix uses the file "log" to "backup" to yield data
from database tables to be replicated.
Replication servers are only the beginning of a whole generation of software that implements the
abstract concepts related to data sharing in heterogeneous environments with advanced
management and parallel processing, optimizing transactions.
1.2 Distributed Parallel Processing in Neural Networks
This type of processing is done using:
- Processing units: by analogy with the human brain, they correspond to neurons, and
collectively to concepts such as character, features of pictures or objects in a PPP
scheme. Individual units do not interpret the detailed design; they can be represented
only by groups of units. We obtain a robust architecture that is not dependent on the
efficiency of individual units and which assigns responsibility;
- Connections and the rule of activation;
- Internal input data (usually spread);
- External input data;
- Results obtained from processing units.
1.3.Open Distributed Processing
Open Distributed Processing (PDD) is the ISO standardization effort in the field of distributed
processing. PDD 10 746 is the set of standards produced by ISO / IEC and ITU-T X.900 products
(Telecommunication International Union). The notion of distributed processing seems very
technical and complicated. However, with the development of large coverage networks,
distributed processing is increasingly used. PDD’s goal is to facilitate the use of distributed
processing in areas as diverse as the wide deployment technology (Tannenbaum and Steen 2006:
241).
Fig. 3 Open distributed processing model
805
PDD is easy to use by programmers and operators because of its nature and means of
distribution. In other words, both programming and use of distributed applications is presented as
if that application would not be distributed. The way to get this perspective is transparent
homogeneous PDD. Transparency provides users and developers with a consistent view of the
network system as in the case when a message can traverse several different networks, without
the user knowing the details of this process; distributed processing may involve different areas
controlled by different authorities and equipment heterogeneous with very different hardware and
software.
In general, the difference between distributed and network solutions is that when working in a
network, the user is aware that the system runs on multiple machines, while in case of distributed
processing the system appears as a single entity.
2. Implementation of Distributed Systems Solutions
2.1 Company Informix, symmetric multiprocessing
Informix Company has made a complete redesign of the system or introduced incentives for
symmetric multiprocessing (SMP). Informix Online, is a dynamically scalable architecture
(DSA). In situations in which the Informix server used version 6.0 of parallel processing,
performance increased in comparison with older architectures. Informix Server version 6.0 of
parallelism is allowed only on certain operations (create indexes, sorting, backup, data recovery)
and runs on some platforms of symmetric multiprocessing (SMP). -Online/DSA Informix server
(Dynamic Scalable Architecture). Version 7.0 has a scalable architecture that can manage scans,
unification (join) the parallel sorting and querying databases.
2.2. Oracle 2.2 - parallel processing technology
Oracle 7 parallel Server creates a good balance between the need for better management and a
multiprocessing system. Both Informix and Sybase disks require special partitioning in parallel
processing, unlike Oracle, which is regarded as superior in this regard. The Oracle 7, Version 7.1
automatically maintains multiple copies of the same data on multiple servers, which eliminates
the need for intensive disk-partitioning labor. Parallel query options in the Oracle Server 7 allow
machinery working in symmetric multiprocessing (SMP), in clustered multiprocessing and
massively parallel processing (MP) to execute a single application on multiple processing units,
thus providing an almost linear scalability with each processor added to the configuration.
2.3 Sybase - Build Monumentum
Sybase company made a major step in designing database systems from the new generation by
launching a product with "multithreading" work features. Sybase Build Monumentum runs in
several ways to control the execution on both platforms -Windows NT and UNIX platforms,
while adding to the existing architectural features for working with object-oriented database. The
multiprocessing on distributed networks is done however only from the "client" point of view.
Build Monumentum has an administrator who manages the "multithreading" processes even if
the host operating system normally does not support this.
2.4 - IBM DB2 Parallel Edition
IBM developed its system of relational database management, DB2/6000, in a parallel multi-
processor architecture with hardware ranging from local networks and systems to single
processor RISC System/6000 IBM Power Parallel Systems SPX. Parallel Server DB2 Parallel
Edition database can efficiently handle very large databases using strategies of sharing data and
applications on their parallel execution.
The DB2/6000 management system extends to support a "shared nothing" architecture selected
for two reasons:
- It is a scalable architecture to the level of hundreds of processors;
806
- Provides high portability because it requires only one communication link
between processors, so it can be ported on any platform.
In the "shared nothing" architecture with the parallel implementation of the management system
DB2 in a network of machines RISC/6000 IBM, database storage is done in a network of
processors that provide buffers, lock structures, files log and separate records for each process.
This prevents competition on cache structure resulting from the fact that all processors share the
same set of resources.
In the architecture with multiple processing nodes, for large databases, data placement becomes a
complex problem and system administration can be difficult. The implementation of parallel
systems using a DB2 requires a definition language and management tools necessary for data
partitioning. DB2/6000 provides two important features for partitioning large database tables: the
use of partitioning keys and groups of nodes for a database that contains several tables; the user
who develops the application can define a partitioning key for each table.
Conclusions
The internal administration of distributed databases is demanding and generally difficult, because
one has to ensure that:
- Distribution is transparent (invisible and unobtrusive) - users must be able to interact with
the system as if it were a non-distributed one (monolithic);
- Transactions must also have a transparent structure (invisible and unobtrusive). The course
of each transaction must maintain database integrity, despite the multiplicity of partitions.
For this they are usually divided into several transactions, each of them working with only
one partition.
These requirements lead to the harmonization of the smart development strategy for Romania, in
line with the priorities of “Europe 2020”. Europeans should have the chance to use services on
networks that are equal to or better than those available in other countries. In this direction have
been made following steps:
- In 2010: adoption of a Recommendation to encourage investment in next generation
access networks (NGA).
- In 2011: Monitoring of the implementation of the NGA Recommendation by national
regulatory authorities and Monitoring of NGA deployment and broadband competition in
the Member States.
- In 2012: guidance on pricing and/or costing methodologies in national regulatory
measures.
Bibliography
1. Cristea Valentin, Algoritmi de prelucrare paralela, Bucuresti, Matrix Rom, 2005
2. Frishberg Leo, Architecture and User Experience (Part 8: A Form of Software Architecture?),
Computer-Human Interaction Forum of Oregon, 2009
3. Lungu Ion, Ghencea Adrian, Distributed databases. Security, consistency and their replication,
16th IBIMA Conference Kuala Lumpur, Malaysia, 2011
4. Lungu I., Vătuiu T, Fodor A. G., Fragmentation solutions used in the projection of the distributed
database systems, Proceedings of the 6th International Conference "ELEKTRO 2006", pp. 44-48,
Edis-Zilina University Publishers, 2006
5. Tannenbaum Andrew S., Steen Van Maarten, Distributed sysems: Principles and paradigms, New
Jersey, Pearson Prentice Hall, 2006
6. Vătuiu Teodora, Popeangă Vasile, Project risk management for migration from client/server
distributed Oracle Database architecture to Web oriented centralized database architecture of The
Ministry of Public Finance’s level, „Universitaria SIMPRO 2006”- Petroşani, 2006, ISSN 1842-4449,
pag. 134.
7. http://ec.europa.eu/information_society/newsroom/cf/fiche-dae.cfm?action_id=203,
accessed April 20, 2011
