Search

Signature file is one of the efficient access methods for retrieval of text database. In a large database server, parallel device is utilized to achieve concurrent access. Efficient allocation of signature file on parallel device minimizes the query response time and is important in the design of large databases. In this paper, we investigate the design of parallel signature file. We propose a new dynamic allocation technique to distribute the signature file on parallel device. It is an improvement of previous approach of Fragmented Signature File. While Fragmented Signature File uses Quick Filter to distribute the partitioned frame signature file, the proposed Parallel Signature File uses a declustering technique. The proposed Parallel Signature File has some advantages. First, the qualified frame signature blocks are distributed more uniformly than Fragmented Signature Files. Second, the proposed scheme can also be used in dynamic environment. Performance analysis shows that performance of the proposed approach outperforms that of Fragmented Signature File and is not far from theoretical optimal response time.

Retrieval, insertion and cycling performance in dynamic hash file organizations is influenced by a number of factors. These include the growth rate experienced by the file, the distribution of hashkeys over the address space, and the method used for handling overflow records. In this paper we propose a dynamic hashing scheme which uses a generalized exponential hash function and a directory-based overflow handling scheme. The hash function, which is designed to accommodate a variety of file growth rates, ensures that retrieval, insertion and file expansion performance is non-cyclic. We extend the concept of dynamic overflow sharing by distributing overflow space in the file. This is accomplished by combining contiguous sets of pages in the file into groups and assigning each group an additional page, referred to as a preallocated overflow page, for storing some or all of the records that overflow the pages in the group. Overflow records that cannot be stored in the preallocated page are directed to a separate overflow file. The resulting system is analyzed mathematically and results obtained have been confirmed using simulation. Based on these results we conclude that the proposed method, while providing good support for retrieval-intensive databases, is ideal for databases which experience a large proportion of insertion transactions. In addition, the characteristics of the scheme enable efficient implementation in a disk cache environment.

The motivation for this paper is to show that the conversion of a B + tree file into a linear hash file and the conversion of a linear hash file into a B + tree file can be done concurrently with user-transaction processing. The conversion process, in general, is motivated by a change in database processing requirements. For the first case, efficient sequential access and direct access are originally needed but now only efficient direct access is needed. For the second case, the opposite is true. This is quite reasonable for a database system which accommodates new and changing applications. Several existing database systems allow an off-line reorganization to support these changes. We devise an algorithm which performs a concurrent conversion for both cases and present an analytic model of the conversion process for each. We also employ a typical database simulation model to evaluate the reorganization scheme. The results from the analytic model for Case 1 are within 3% (on average) of the observed simulation results and for Case 2 are within 4% (on average) of the observed simulation results.

The file allocation problem considers a file and a fully connected network having n nodes. The problem assumes that the overall file usage over a unit time period is known and it asks for the optimal set of network sites at which to locate copies of the file. This paper considers the same problem but it assumes that the behavior of the user access patterns changes over v planning periods in a manner, known in advance. A model is presented which shows that there are (2n − 1)v possible file allocations. To assist the searching of this large solution space four theorems are presented which are subsequently utilized to analyze the problem and to solve an example case.

Severance and Lohman[1] proposed that all database amendments (i.e. additions, deletions and updates) are localized into a relatively small storage area, called a differential file. When the differential file grows sufficiently large, a reorganization incorporates all changes into a new generation of the database. This paper proposes a modification of the differential file approach in that the differential file only stores record additions. It is also concerned with the selection of the optimum reorganization points for both the original and the modified differential file approaches.