A multi-model algorithm for the cost-oriented design of Internet-based systems

The selection of a cost-minimizing combination of hardware and network components of Internet-based systems to satisfy organizational requirements is a complex design problem with multiple degrees of freedom. Decisions must be made on how to distribute the overall computing load onto multiple comput...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Information sciences Ročník 176; číslo 21; s. 3105 - 3131
Hlavní autoři: Ardagna, Danilo, Francalanci, Chiara, Trubian, Marco
Médium: Journal Article
Jazyk:angličtina
Vydáno: Elsevier Inc 03.11.2006
Témata:
Algorithms
Combinatorial optimization
Costs minimization
Design tool
Distributed applications
Web-based services
Design tool
Algorithms
Distributed applications
Web-based services
Combinatorial optimization
Costs minimization
ISSN:0020-0255, 1872-6291
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:The selection of a cost-minimizing combination of hardware and network components of Internet-based systems to satisfy organizational requirements is a complex design problem with multiple degrees of freedom. Decisions must be made on how to distribute the overall computing load onto multiple computers, where to locate computers and how to take advantage of legacy components. The corresponding optimization problem not only embeds the structure of NP-hard problems, but also represents a challenge with a well-structured heuristic approach. A scientific approach has been rarely applied to cost minimization and a rigorous methodological support to cost issues of the design of Internet-based distributed systems is still lacking. The methodological contribution of this paper is the representation of complex design issues as a set of four intertwined cost-minimization sub-problems: two set-partitionings, a set-packing and a min k-cut with a non-linear objective function. Optimization is accomplished by sequentially solving these sub-problems with a heuristic approach and tuning their solution with a local-search approach. Results indicate that decomposition significantly reduces optimization time and solutions have also lower costs than those identified without prior decomposition (20–60%). Cost reductions considerably grow (25–70%) when methodological outputs are compared with practitioners’ solutions.
ISSN:0020-0255
1872-6291
DOI:10.1016/j.ins.2005.12.013