## Design and Evaluation of Data Allocation Algorithms for Distributed Multimedia Database Systems (1996)

### Cached

### Download Links

- [ranger.uta.edu]
- [ranger.uta.edu]
- [www.eee.hku.hk]
- [www.eee.hku.hk]
- DBLP

### Other Repositories/Bibliography

Venue: | IEEE Journal on Selected areas in Communication |

Citations: | 26 - 14 self |

### BibTeX

@ARTICLE{Karlapalem96designand,

author = {Kamalakar Karlapalem and Ishfaq Ahmad and Siu-kai So and Yu-kwong Kwok},

title = {Design and Evaluation of Data Allocation Algorithms for Distributed Multimedia Database Systems},

journal = {IEEE Journal on Selected areas in Communication},

year = {1996},

volume = {14},

pages = {1332--1348}

}

### OpenURL

### Abstract

Given a distributed multimedia database system and a set of queries as well as their frequencies from each site, the objective of a data allocation algorithm is to locate the multimedia data objects (MDOs) at different sites so as to minimize the total data transfer cost incurred in executing the queries. The data allocation problem, however, is NP-complete, and thus requires fast heuristics to generate efficient solutions. In this paper we propose three data allocation algorithms which are based on a genetic technique, an evolutionary process, and neural networks. We have implemented and evaluated these algorithms on our distributed multimedia database system test-bed. A comparison of the algorithms reveals trade-offs between their solution quality and time-complexity. 1

### Citations

8422 |
Genetic Algorithms
- Goldberg
- 1989
(Show Context)
Citation Context ... the multimedia data provider. The proposed algorithms are described as follows: 4.1 The Genetic Algorithm Genetic algorithms manipulate a population of potential solutions to an optimization problem =-=[8]-=-, [9], [12], [18]. They operate on encoded representations of the solutions, equivalent to the genetic material of individuals in nature, and not directly on the solutions themselves. As in nature, th... |

620 |
Data Structures and Network Algorithms
- Tarjan
- 1983
(Show Context)
Citation Context ...x ∇ ⎠ k – 1 x = 0 T = D+ U = ∑ ∑ csite Oj j = 0 j′ = 0 i = 0 We first describe an algorithm that generates optimal solutions for query-site query execution strategy. We use Floyd-Warshall�=-=��s algorithm [26] t-=-o find the minimal cost communication path between any two sites. This gives us the cij′ values in the cost matrix C. The running time of this algorithm is distributed database system. Om where m is... |

244 |
A Partitioning Strategy for Nonuniform Problems on Multiprocessors
- Berger, Bokhari
- 1987
(Show Context)
Citation Context ...roblem. In general, search-based methods can produce better solutions at the expense of much longer execution time, while clustering approaches are much faster but may generate less optimal solutions =-=[3], -=-[16], [23], [24]. Clustering algorithms can be broadly divided into two classes [4]: • Agglomerative Algorithms: The graph to be clustered is initially considered to have N unit clusters. A number o... |

199 |
Synchonization and Storage Models for Multimedia Objects
- Little, Ghafoor
- 1990
(Show Context)
Citation Context ...s to specify queries accessing the distributed multimedia database system and presenting the result to the user. The synchronization for the presentation of the multimedia data is handled by the CMUI =-=[15]-=-. Whereas, the Multimedia Data Provider (MDP) identifies the relevant multimedia data for an user query and facilitates shipping of the multimedia data to the CMUI. The CMUI is a client process and th... |

141 |
Distributed Databases: Principles and Systems
- Ceri, Pelagatti
- 1984
(Show Context)
Citation Context ...model of the dependencies among the data fragments accessed by the query. These dependencies arise because of the partitioning of the relations into data fragments (by using methodologies proposed by =-=[7]-=-) and/or access to multiple relations by a query. We use the distributed query decomposer and data localization algorithms [17] to decompose a distributed query into a set of queries. This decomposed ... |

124 |
Heuristic Algorithms for Task Assignment in Distributed Systems,” in
- Lo
- 1988
(Show Context)
Citation Context ...m. In general, search-based methods can produce better solutions at the expense of much longer execution time, while clustering approaches are much faster but may generate less optimal solutions [3], =-=[16], -=-[23], [24]. Clustering algorithms can be broadly divided into two classes [4]: • Agglomerative Algorithms: The graph to be clustered is initially considered to have N unit clusters. A number of clus... |

80 | Parallel recombinative simulated annealing: a genetic algorithm
- Crainic, Mahfoud, et al.
- 1995
(Show Context)
Citation Context ...media data provider. The proposed algorithms are described as follows: 4.1 The Genetic Algorithm Genetic algorithms manipulate a population of potential solutions to an optimization problem [8], [9], =-=[12]-=-, [18]. They operate on encoded representations of the solutions, equivalent to the genetic material of individuals in nature, and not directly on the solutions themselves. As in nature, the selection... |

74 | Data allocation in distributed database systems
- Apers
- 1988
(Show Context)
Citation Context ...oblem of file allocation for typical distributed database applications with a simple model for transaction execution taking into account the dependencies between accesses to multiple fragments. Apers =-=[1]-=- considered the allocation of the distributed database to the sites so as to minimize total data transfer cost. The author came up with a complicated approach to allocate relations by first partitioni... |

68 | Multiple file allocation in a multiple computer system
- Chu
- 1969
(Show Context)
Citation Context ... shown in Figure 1 (unlike the integrated and heterogeneous computing paradigm based distributed multimedia database systems [2]). Special-purpose distributed multimedia database system architectures =-=[8]-=-, [18] have been proposed. The aim of these architectures is to support specific application domains (like, Medical Databases and News-on-Demand), and thus they are tightly integrated to favor these a... |

66 |
A graph matching approach to optimal task assignment i distributed computing systems using a minimax criterion
- Shen, Tsai
- 1985
(Show Context)
Citation Context ...is formulated as a state-space search problem. - 17 - Total Data Transfer CostsA best-first search algorithm, which is based on similar principles as the A* algorithm, is used for solving the problem =-=[25]-=-. In a state-space search problem, each state description is denoted by a node. Operators applicable to nodes are defined for generating successors of nodes, called node expansion. A solution path of ... |

39 |
Nearest-Neighbor Mapping of Finite Element Graphs onto Processor Meshes
- Sadayappan, Ercal
- 1987
(Show Context)
Citation Context ...l, search-based methods can produce better solutions at the expense of much longer execution time, while clustering approaches are much faster but may generate less optimal solutions [3], [16], [23], =-=[24]. -=-Clustering algorithms can be broadly divided into two classes [4]: • Agglomerative Algorithms: The graph to be clustered is initially considered to have N unit clusters. A number of clustering opera... |

37 |
Allocation of Copies of a File in an Information Network
- Casey
- 1972
(Show Context)
Citation Context ...sor system. He presented a global optimization model to minimize overall processing costs under the constraints of response time and storage capacity with a fixed number of copies of each file. Casey =-=[5] d-=-istinguished between updates and queries on files. Eswaran [11] proved that Casey’s formulation is NP complete. He suggested that a heuristic rather than exhaustive search approach is more suitable.... |

32 |
Graph Partitioning Using Annealed Neural Networks
- Bout, Miller
- 1990
(Show Context)
Citation Context ...ated evolution algorithm, like GA, is O(GP( k +km)), where G is the number of generations and P is the population size. 4.3 The Mean Field Annealing Algorithm The mean field annealing (MFA) technique =-=[3]-=-, [4], combines the collective computation property of the famous Hopfield Neural Network (HNN) with the annealing notion of another well-known optimization algorithm known as the simulated annealing ... |

30 | A new mapping heuristic based on mean field annealing
- Bultan, Aykanat
- 1992
(Show Context)
Citation Context ...evolution algorithm, like GA, is O(GP( k +km)), where G is the number of generations and P is the population size. 4.3 The Mean Field Annealing Algorithm The mean field annealing (MFA) technique [3], =-=[4]-=-, combines the collective computation property of the famous Hopfield Neural Network (HNN) with the annealing notion of another well-known optimization algorithm known as the simulated annealing (SA) ... |

27 | Cluster partitioning approaches to mapping parallel programs onto a hypercube
- Sadayappan, Ercal, et al.
- 1990
(Show Context)
Citation Context ...general, search-based methods can produce better solutions at the expense of much longer execution time, while clustering approaches are much faster but may generate less optimal solutions [3], [16], =-=[23], -=-[24]. Clustering algorithms can be broadly divided into two classes [4]: • Agglomerative Algorithms: The graph to be clustered is initially considered to have N unit clusters. A number of clustering... |

26 |
On the Assignment Problem of Arbitrary Process Systems to Heterogeneous Distributed
- Bowen, Nikolaou, et al.
- 1992
(Show Context)
Citation Context ...of much longer execution time, while clustering approaches are much faster but may generate less optimal solutions [3], [16], [23], [24]. Clustering algorithms can be broadly divided into two classes =-=[4]: -=-• Agglomerative Algorithms: The graph to be clustered is initially considered to have N unit clusters. A number of clustering operations are then performed to produce a certain number of larger clus... |

22 | An ObjectOriented Multimedia Database System for a News-On-Demand Application
- Ozsu, Szafron, et al.
- 1995
(Show Context)
Citation Context ...n in Figure 1 (unlike the integrated and heterogeneous computing paradigm based distributed multimedia database systems [2]). Special-purpose distributed multimedia database system architectures [8], =-=[18]-=- have been proposed. The aim of these architectures is to support specific application domains (like, Medical Databases and News-on-Demand), and thus they are tightly integrated to favor these applica... |

22 | Multimedia Database Systems – The Notions and the Issues
- Rakow, Neuhold, et al.
- 1995
(Show Context)
Citation Context ...climbing heuristics, optimal allocation, max-flow min-cut problem, network flow algorithm, clustering, best-first search algorithm. - 1 -s1 Introduction A distributed multimedia database system [13], =-=[19]-=- is a database system loosely coupled with a multimedia data provider as shown in Figure 1 (unlike the integrated and heterogeneous computing paradigm based distributed multimedia database systems [2]... |

14 |
Computer and Database Location in Distributed Computer Systems
- Gavish, Pirkul
- 1986
(Show Context)
Citation Context ...h is the lack of simplicity in both incorporation of the query execution strategy and the solution procedure. There have been many linear programming formulations proposed for data allocation problem =-=[12]-=-, [20]. The main problem with these approaches is the lack of modeling of the query execution strategy. Lin, Orlawska and Zhang [14] also developed a heuristic algorithm for minimum overall data trans... |

12 |
Placement of Records in a File and File Allocation
- Eswaran
- 1974
(Show Context)
Citation Context ...rategy employed by the distributed database system. This is because different query execution strategies have different MDO migration patterns. The data allocation algorithm is NP-complete in general =-=[11]-=-. In this section, we develop solutions for the allocation problem when query-site and move-small query execution strategies are respectively used by the distributed database management system and the... |

10 |
Multimedia Database Management Systems
- Ghafoor
- 1995
(Show Context)
Citation Context ... hill-climbing heuristics, optimal allocation, max-flow min-cut problem, network flow algorithm, clustering, best-first search algorithm. - 1 -s1 Introduction A distributed multimedia database system =-=[13]-=-, [19] is a database system loosely coupled with a multimedia data provider as shown in Figure 1 (unlike the integrated and heterogeneous computing paradigm based distributed multimedia database syste... |

9 | Parallel A* algorithms and their performance on hypercube multiprocessors
- Dutt, Mahapatra
- 1993
(Show Context)
Citation Context ... is Om . However, it should be noted that with the above properly defined heuristic node evaluation function, the BFS/ESP algorithm can take much less time than the worst case time complexity implies =-=[10]-=-. k ( ) 4.4 Agglomerative Clustering In this section, we describe two clustering approaches to tackle the data allocation problem. In general, search-based methods can produce better solutions at the ... |

9 |
Taskgraph mapping using a genetic algorithm: A comparison of fitness functions
- Hurley
- 1993
(Show Context)
Citation Context ...multimedia data provider. The proposed algorithms are described as follows: 4.1 The Genetic Algorithm Genetic algorithms manipulate a population of potential solutions to an optimization problem [8], =-=[9]-=-, [12], [18]. They operate on encoded representations of the solutions, equivalent to the genetic material of individuals in nature, and not directly on the solutions themselves. As in nature, the sel... |

8 |
An analytical framework for the redesign of distributed databases
- Wilson, Navathe
- 1986
(Show Context)
Citation Context ...ing algorithms are very cost-effective if fast execution is desired. Therefore, these algorithms can be applied to generate allocation schemes for limited redesign of distributed multimedia databases =-=[27]-=-. Limited redesign implies change in allocation scheme but not in fragmentation scheme for a distributed database. The algorithms developed in this paper can be integrated with algorithms for affectin... |

8 |
Genetic Algorithms
- Srinivas, Patnaik
- 1994
(Show Context)
Citation Context ...data provider. The proposed algorithms are described as follows: 4.1 The Genetic Algorithm Genetic algorithms manipulate a population of potential solutions to an optimization problem [8], [9], [12], =-=[18]-=-. They operate on encoded representations of the solutions, equivalent to the genetic material of individuals in nature, and not directly on the solutions themselves. As in nature, the selection mecha... |

7 |
A Model for Database Allocation Incorporating a Concurrency Control Mechanism
- Ram, Marsten
- 1991
(Show Context)
Citation Context ...he lack of simplicity in both incorporation of the query execution strategy and the solution procedure. There have been many linear programming formulations proposed for data allocation problem [12], =-=[20]-=-. The main problem with these approaches is the lack of modeling of the query execution strategy. Lin, Orlawska and Zhang [14] also developed a heuristic algorithm for minimum overall data transfer co... |

3 |
Optimal file allocation for a distributed on a network of minicomputers
- Ceri, Martella, et al.
- 1980
(Show Context)
Citation Context ...d the generalized file allocation problem. They also proposed a model for file migration or reallocation that is identical in formulation to the file allocation problem. Ceri, Martella and Pellagatti =-=[6]-=- considered the problem of file allocation for typical distributed database applications with a simple model for transaction execution taking into account the dependencies between accesses to multiple... |

3 |
Scheduling Data Redistribution
- Rivera-Vega, Varadarjan, et al.
- 1990
(Show Context)
Citation Context ...scheme but not in fragmentation scheme for a distributed database. The algorithms developed in this paper can be integrated with algorithms for affecting the limited redesign of distributed databases =-=[22]-=- and thus facilitating on-line tuning of a distributed multimedia database system. If solution quality is the more prominent factor, the hill-climbing approach is a viable choice for an offline alloca... |

2 |
Site Assignment for Relations and
- Cornell, Yu
- 1988
(Show Context)
Citation Context ...ne of the outputs of the allocation algorithm. This curtails the applicability of this methodology when fragmentation schema is already defined and allocation scheme must be generated. Cornell and Yu =-=[9]-=- proposed a strategy to integrate the treatment of relation assignment and query strategy to optimize performance of a distributed database system. Though they took into consideration the query execut... |

2 |
Database Placement in Communication Networks for Minimizing the overall Transmission
- Lin, Orlowska, et al.
- 1994
(Show Context)
Citation Context ...inear programming formulations proposed for data allocation problem [12], [20]. The main problem with these approaches is the lack of modeling of the query execution strategy. Lin, Orlawska and Zhang =-=[14]-=- also developed a heuristic algorithm for minimum overall data transfer cost, by considering replicated allocation of fragments and both read and update transactions. 8 Conclusions The problem of non-... |

2 |
The Annealing Evolution Algorithm as Function
- Yong, Lishan, et al.
- 1995
(Show Context)
Citation Context ..., combines the collective computation property of the famous Hopfield Neural Network (HNN) with the annealing notion of another well-known optimization algorithm known as the simulated annealing (SA) =-=[20]-=-. The MFA algorithm is derived from an analogy to the Ising spin model which is used to estimate the state of a system of particles or spins in thermal equilibrium. In the Ising spin model, the energy... |

1 |
The Placement of Relations on a Distributed Relational
- Ramamoorthy, Wah
- 1979
(Show Context)
Citation Context ...n updates and queries on files. Eswaran [11] proved that Casey’s formulation is NP complete. He suggested that a heuristic rather than exhaustive search approach is more suitable. Ramamoorthy and Wa=-=h [21]-=- analyzed a file allocation problem in the environment of a distributed database and developed a heuristic approximation algorithm for a simple file allocation problem and the generalized file allocat... |