DonNTU   Masters' portal

Abstract

Содержание

Introduction

The idea of cloud computing is very popular last few years in IT. Many different organizations start to realize those technologies aiming to reduce costs and time for administration [1].

Cloud computing is a model of easy network access for general computing resources. It is given operatively with minimal efforts for management [2].

1. Goal and tasks of the research

The goal of this work is development and analysis of method placement on cloud the software for solving dynamic resource allocation problem. To reach this goal is necessary to solve some tasks:

According to the task the measures of choosing the best method are: error, complexity, time, storage.

Every method must be accompanied with instruction and theoretical reference.

2. A review of researches

The basic methods of solving resource allocation problem are described in works of many scientists (V. Mikhalevich, V. Skurba, V. Tanayev, A. Kuksa, W. Maxwell, J. Thompson).

For fast solving of resource allocation tasks can be used combinatorial optimization methods or graph-based methods. The source [3] recommends method of branches and borders, the source [4] - method of elimination of unknowns, in the source [5] is described method, combining [3] and [4].

A big part of resource allocation tasks is solved by dynamic programming methods. Its main idea is very easy [6]. For solving the main task it is necessary to solve separate subtasks and unite all decisions in one.

3. Theme urgency

Resource allocation tasks are the main part of applied tasks. They are used for bankroll allocation, inventories allocation, network resources allocation, time and storage allocation.

Modern scientific research associated with conducting complex calculations using high performance computing resources. The effectiveness of research depends on the availability and accessibility of computer applications for solving a particular problem. There is a large luggage of such applications, including libraries of numerical methods, applied computing packages, computational models, etc. Ready software allows researchers avoid time-consuming implementation code and focus on the task in most cases.

To solve these problems apply high-level environment which remote an access to computing applications through task-oriented interfaces. Such software is implemented as web-portals, and work with applications takes place through a web–browser [10].

This problem has found the solution by moving applications to the cloud last few years. This is due to speed of cloud–based applications, their scalability and profitability [9].

4. Development the software for solving dynamic resource allocation problem

It was decided to develop the software for solving dynamic resource allocation problem in online–mode with placing it on a cloud service that allows a user, by passing the search for the necessary software and installation process to obtain a solution of the task with a description of the decision.

The software will have the following structure:

A typical resource allocation task of the source [11] is presented below.

There is an initial amount of funds ζ0, that must be allocated for the n years between the s enterprises. Funds uki(k=1, ..., n; i=1, ..., s) allocated to the yeark of the enterprisei earn fki(uki) and back to the end of the year in the amount of φki(uki). Income can participate (fully or partially) or not participate in the following distribution.

It is needed to find a way of resource allocation to total income from s enterprises in n years strives to maximum.

Efficiency indicator is the total income from s enterprises.

Формула 1

The amount of resources in the yeark is a value ζk–1.

The equation of state of the process is as follows:

Формула 2

The general problem is formulated as follows [5]:

Формула 3

where J – quality measure optimization problem, lij(t), uij(t),gi(t)) – restrictions, yi0 – initial conditions, Т – total time required for resource allocation, m – number of arcs, n – number of vertices.

5. Development of method placement on cloud the software

To place developing application is selected cloud platform Windows Azure is the one of cloud platforms, provides a service of accommodation executable user application.

Windows Azure is an open and flexible cloud platform that allows creating, deploying and managing applications in a global network. Applications can be developed using any language, tool or platform.

Current version of the Windows Azure supports a variety of computer services, namely, virtual machines, websites, cloud services and mobile services.

Windows Azure cloud services allow creating and deploying applications in almost any programming language and with minimal administration costs.

According to the report of the analytical company Nasuni, the Windows Azure platform is a leader in performance tests when writing and reading data from the cloud, data availability and the minimum number of errors (0 %) [9].

Windows Azure is of commercial interest for companies that do not have specific technological preferences. For example, the company Guppers uses Windows Azure service not only for storage but also for solving computational problems, web–service queries and manipulating data in a database SQL Azure [7].

Conclusion

During the work methods and tools for building applications for the cloud were analysed; the list of techniques that may be applied in online analysis process processes was formulated; evaluation measures were selected. An online–mode was provided using the methods of dynamic programming. Application architecture has been designed according to the proposed method.

The practical value of the development is that users are given the opportunity to quickly solve common tasks for resource allocation by different methods on their own input data with the giving a theoretical description and reference solutions. At the same time, it is not necessary to install software.

References

  1. Шмойлов Д.В. Облачные вычисления: актуальность и проблемы / Д.В. Шмойлов  // Электронное научное периодическое издание «Электроника и информационные технологии» № 1 (10). – МГУ им. Н.П. Огарева, г. Саранск, 2011 г., 7 c.
  2. Бакст Л.А., Бурляева О.К., Кузнецова В.В., Малышева Е.В. Реализация облачных вычислений – актуальная задача развития информационно–вычислительных сетей / Л.А. Бакст, О.К. Бурляева, В.В. Кузнецова, Е.В. Малышева // «Профессиональные инновации» № 7. – г. Москва, 2013 г., С. 26–36.
  3. Сергиенко И.В., Шило В.П. Задачи дискретной оптимизации. – К.: Наукова думка, 2003 г. – 301 с.
  4. Остапенко В.В., Финин Г.С. Метод исключения неизвестных для систем линейных неравенств со структурой графа «Кибернетика и системный анализ», № 5, 1999 г., С. 66–75.
  5. Маслова Н.А. Методы теории вычислений в решении задач управления технологическими процессами / Н.А. Маслова // Штучний інтелект. – 2009 г. – № 3. – С. 165–171.
  6. Михалевич В.С., Кукса А.И. Методы последовательной оптимизации в дискретных сетевых задач оптимального распределения ресурсов. – М.: Наука, 1983 г. – 208 с.
  7. Облачные сервисы. Взгляд из России. Под ред. Е. Гребнева. – М.: Cnews, 2011 г. – 282 с.
  8. Мовчан О.В. Маслова Н.А. Разработка и анализ метода размещения на облаке программного обеспечения для решения динамических задач распределения ресурсов / О.В. Мовчан Н.А. Маслова // Материалы V Всеукраинской научно-технической конференции Информационные управляющие системы и компьютерный мониторинг. – Донецк: ДонНТУ, 2014. С. 450-455.
  9. Таллоч Митч и команда Windows Azure Знакомство с Windows Azure. Для ИТ–специалистов / Таллоч М.; пер. с англ. – М.: ЭКОМ Паблишерз, 2014 г. — 154 с.: ил.
  10. Сухорослов О.В. Облачная платформа для создания вычислительных веб–сервисов на базе инструментария MathCloud / О.В. Сухорослов // Материалы IV Международной конференции «Облачные вычисления. Образование. Исследования. Разработка». – г. Москва, 2013 г., 4 с.
  11. Каллихман И.Л., Войтенко М.А. Динамическое программирование в примерах и задачах. М.: Высшая школа, 1979 г. 124 с.

Note

This abstract is not a full version of the master's work. Writing this abstract of master's work is not yet complete. It is planned to complete the work on the November 2014. Advanced work and a complete list of materials can be obtained from the author or his scientific adviser after that date.