Salzberg, B. "File Structures, An analytic approach" ISBN 0-13-314691-X; Prentice Hall.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....the leaf nodes and uses the non leaf nodes (index nodes) only as a search directory. The B tree on the other hand, uses the nodes as both a data repository and an indexing directory. The typically large fanout of B trees has the consequence that the height of the B tree index rarely exceeds two [Sal88] Given that the root is always stored in main memory, most insertions, deletions, and exact match queries require no more than four disk accesses [SL91] B trees continue to be studied until recently though from different perspectives. An example is the field of concurrency control where special ....

B. Salzberg. File Structures: An Analytic Approach. Prentice-Hall, Englewood Cliffs, NJ, 1988.

....tree are stored consecutively on 7 disk. This makes it possible to access a node in a file on disk by a start address and an offset. This assumption may be too restrictive in a multi user environment where nodes are allocated dynamically. Here, disk space may be allocated in chunks, termed extents [23]. An extent is a number of consecutive disk pages. All extents contain the same fixed number of disk pages. Within an extent, disk pages can be accessed via a start address and an offset. To make it possible to search a PLI tree, without extra I O cost, an array containing start addresses of all ....

....the size of the PLI tree compared with the size of the backlog is given as follows. Worst case: ii P h Gamma1 m=0 d m j 2 j ffi Gamma (d Gamma 1) d h Gamma1 2 Delta Examples of the worst case and the best case are shown in Table III for height, h = 3 and order, d = 100 [23]. As can be seen, the backlog is approximately d times larger than the PLI tree. The size of the index is very small. The difference for the worst case and best case is insignificant for realistic trees. 9 PLI tree Backlog PLI GammaT ree Backlog Worst case 10,103 990,002 1.0205 Best case ....

B. Salzberg. File Structures: an analytic approach. PrenticeHall, Englewood Cliffs, NJ, USA. ISBN 0--1331--4550--6, 1988.

....physically close to each other. The goal of clustering is to limit the number of disk accesses required to process a query by increasing the likelihood that query results have already been cached. Clustering has been well researched in the field of file structures and access methods (e.g. [Sal88, GR97, Sto94]. B Trees, for instance, provide one dimensional clustering. Multidimensional clustering has been discussed in the field of multidimensional access methods (e.g. GG97, Sam90] A great deal of research has been done in the field of access methods and access path selection. Especially for DW ....

....respect to that partial order defined by the hierarchy levels. For queries with large result sets one dimensional clustering reduces disk accesses by a factor of P. Clustering of one dimensional objects and single object hierarchies has been discussed to a large extent (e.g. ZSL98] BK89] [Sal88]) If the order of dimensions during drill down is known in advance, clustering the data in this order will result in a good query performance. In principle, a concatenated clustering index (i.e. B Tree) on the hierarchy levels of all dimensions in one lexicographic order is maintained. ....

B. Salzberg. File Structures: An Analytic Approach. Prentice Hall, 1988.

....tree are stored consecutively on disk. This makes it possible to access a node in a file on disk by a start address and an offset. This assumption may be too restrictive in a multi user environment where nodes are allocated dynamically. Here, disk space may be allocated in chunks, termed extents [23]. An extent is a number of consecutive disk pages. All extents contain the same fixed number of disk pages. Within an extent, disk pages can be accessed via a start address and an offset. To make it possible to search a PLI tree, without extra I O cost, an array containing start addresses of all ....

....the worst case situation, the size of the PLI tree compared with the size of the backlog is given as follows. Wo rs t ca s e : ## # h 1 m=0 d m # 2 # ## (d 1)d h 1 2 # Examples of the worst case and the best case are shown in Table 3 for height, h = 3 and order, d = 100 [23]. As can be seen, the backlog is approximately d times larger than the PLI tree. The size of the index is very small. The difference for the worst case and best case is insignificant for realistic trees. PLI Tree Backlog PLI Tree Backlog ( Worst case 10,103 990,002 1.0205 Best case 1,010,101 ....

B. Salzberg. File Structures: an analytic approach. Prentice-Hall, Englewood Cliffs, NJ, USA. ISBN 0--1331--4550--6, 1988.

....level 6 of the hierarchy (i.e. the root of the subtree is an object at the supervisor level) is stored in a segment. In the experiments, the segment size is calculated as follows: divide the number of bytes in a subtree rooted at level 6 by 0.69. This gives the same fill factor as a B tree[35]. The overflow mechanism was not programmed for the Link method in the experiments. The experiments did not cause segment overflow, since not many records were inserted. This seems to put Link in a good position. Because if segmentation and subtrees of the hierarchy do not match, as happens with ....

....will be reactivated (i.e. unsetting the flag) in the index. When the IB is finished, cleaning up of pseudo deleted keys should be done. Because the NSF algorithm allows transactions to concurrently insert and delete keys directly in the index tree, it can t build the tree in a bottom up fashion [35]. 5.7.2 Side File Approach (SF) The idea of using a side file is to let the IB be able to build the index in a bottom up fashion without any interference caused by the insertions and deletions of user transactions. The IB scans the data and extracts the keys. It also maintains a position ....

[Article contains additional citation context not shown here]

Betty Salzberg. File Structures: An Analytic Approach. Prentice Hall, 1988.

....dimensions are independent and thus such a hierarchy could be split up in two separate hierarchies (see Section 4.2.2) 4. 2 Multidimensional Hierarchical Clustering Clustering of one dimensional objects and single object hierarchies has been discussed to a large extent (e.g. ZSL98] BK89] [Sal88]) However, OLAP queries often impose restrictions with respect to hierarchies over multiple dimensions. The result set satisfying these restrictions is usually quite large; for presentation it is grouped and aggregated or ranked. Clustering data with respect to multiple hierarchies can ....

B. Salzberg. File Structures: An Analytic Approach. Prentice Hall, 1988.

....in main memory. The availability of large main memories and new technologies for disk drives have modified the models for external sorting and have renewed interest in their study [11, 13] Currently, the sorting of files that are too large to be held in main memory is performed on disk drives [10]. External sorting consists of two phases: a run creation phase and a merge phase. During the first phase, the data to be sorted is divided into smaller sorted 1 This work was partially carried out under grants from the Natural Sciences and Engineering Research Council of Canada and the ....

....M exico. 3 Department of Computer Science, University of Western Ontario, London, Ontario N6A 5B7, Canada. 2 Estivill Castro and Wood sets called initial runs or strings [8] During the second phase, one ore more passes of multiway merge ar used to combine the initial runs into a single run [10, 11]. Because replacement selection allows full overlapping of I O with sequential reading and writing of data and produces initial runs that are larger than the available main memory, it is standard for the run creation phase. We confirm mathematically that the lengths of the runs created by ....

B. Salzberg. File Structures: An Analytic Approach. Prentice-Hall, Inc., Englewood Cliffs, NJ, 1988.

....helpful comments of Vinay Deshpande, to whom I dedicate this paper. Appendix Typical disk systems have a seek time ranging between 10 and 20 ms. with an average rotational delay of 8 ms. and with a transference rate varying from 1 to 4 MB per second. Typical record sizes range from 0. 5 to 2 KB [18]. Considering that the time to access one bucket is approximately 20 ms, including seek time, rotational latency, and transfer time, replacing the extremes values in the formula for R we have 10 R 160. ....

Salzberg, B. File Structures: An Analytic Approach, Prentice Hall, 1988.

....the leaf nodes and uses the non leaf nodes (index nodes) only as a search directory. The B tree on the other hand, uses the nodes as both a data repository and an indexing directory. The typically large fanout of B trees has the consequence that the height of the B tree index rarely exceeds two [Sal88] Given that the root is always stored in main memory, most insertions, deletions, and exact match queries require no more than four disk accesses [SL91] B trees continue to be studied until recently though from different perspectives. An example is the field of concurrency control where ....

B. Salzberg. File Structures: An Analytic Approach. Prentice-Hall, Englewood Cliffs, NJ, 1988.

....an important role in the search performance. The bigger the Delta the larger the linear scan on the leaves of the indexing tree. This may be a problem as we are using a B tree and it is very good for finding small ranges . if the ranges are large, it pays to use sequential reading . [Sal88, Chapter 5]. Fortunately we can improve the search performance by applying a simple idea. We simply split the data items, i.e. the indexing ranges, in two sets, one set containing all ranges in the interval [0; b Delta=2c] and the other one containing all ranges in the interval [b Delta=2c 1; Delta] It is ....

B. Salzberg. File Structures: An Analytic Approach. Prentice-Hall, Englewood Cliffs, NJ, 1988.

....introducing the B tree concept. Instead, the notion of maintaining all data in leaf nodes is repeatedly brought up as an interesting variant. As the importance of B trees gained recognition in the database community, a number of textbooks geared towards databases have presented them. In [Sal88] B tree algorithms are presented, though deletion is in fact incomplete and described as quite a complicated algorithm. Both [Liv90] and [FZ92] cover them as well, but omit deletion. FZ92] does present a useful figure depicting the recursive approach to the insertion algorithm for ....

B. J. Salzberg. File Structures: an Analytic Approach. Prentice Hall, Englewood Cliffs NJ, 1988.

....on a system where N is extremely large and the storage medium is relatively slow will almost always be worse than the most convoluted conversion. ADAMS excels in this area, primarily because it employs sophisticated tree search. B. Tree Search B trees are frequently used to index large files [Sal88]. While they provide reasonable access time, there are undesirable characteristics. In particular, search requires in memory comparison of the search key with individual item keys. This is due to the fact that the key values are not necessarily maintained in the tree in a bit wise lexicographic ....

B. Salzberg, File structures: an analytic approach (Prentice-Hall, Inc. Englewood Cliffs, NJ, 1988).

....that allows users to read and write the shadow as soon as the backup begins. 7. Delete the original copy. Mix s description includes the tasks that a database administrator performs. 2. 3 Read Write Access While Reorganizing and Using Replication DataPropagator Relational [3] and Replidata MVS [5] are products that support replication of DB2 databases. They are not facilities for reorganization, but either of them can be used to achieve the effect of reorganization that allows read write access. They require a table to have a unique key. A database administrator can perform the following ....

....Nov. 1977, pp. 528 533. 2] Haderle, D. J. and Jackson, R. D. IBM Database 2 Overview, IBM Syst. J. Vol. 23, No. 2, 1984, pp. 112 125. 3] IBM Corp. An Introduction to DataPropagator Relational Release 1, GC26 3398 01, 1993. 4] IBM Corp. Implementing Concurrent Copy, GG24 3990 00, Dec. 1993. [5] IBM Corp. and Integrated Systems Solutions Corp. Replidata MVS User s Guide, BLD REP UG00, Jan. 1994. 6] Mix, F. DB2 Reorg and Continuous Select, Proc. 6th Ann. N. Amer. Conf. Intl. DB2 Users Group, IDUG, Chicago, May 10, 1994, pp. 545 563. 7] Mohan, C. IBM s Relational DBMS Products: Features ....

[Article contains additional citation context not shown here]

Betty Salzberg. File Structures: An Analytic Approach. Prentice Hall, 1988.

....organization, a B tree is used to store the hierarchy. The B trees are organized by the concatenated keys of the records on the path to a given record. In object oriented databases, this is often referred to as the path name. Primary B trees cluster data dynamically by pages[Sal88]. This solves a major problem for hierarchical databases because it will maximize the likelihood that data will be found in main memory when accessed [Cat91] The following example illustrates this storage method. Suppose there is a company department employee hierarchy. Suppose there are ....

....the number of bytes in a subtree rooted at level 6 by 0.69. This gives the same fill factor Table 1: Test Configuration Parameter Value(s) Page Size 8 Kbytes Buffer Size 1000 pages Record Size 150 bytes Upper level Fan out 3 Small 20 Bottom level Fan out Medium 100 Large 200 as a B tree[Sal88]. The overflow mechanism was not programmed for the Link method in the experiments. The experiments did not cause segment overflow, since not many records were inserted. This seems to put Link in a good position. Because if segmentation and subtrees of the hierarchy do not match, as happens with ....

Betty Salzberg. File Structures: An Analytic Approach. Prentice Hall, 1988.

....pages are already sorted, we start by reading the base pages from left to right, that is, we read the keys in ascending order. We start building a new B tree in a bottom up fashion. Constructing a B tree from sorted records in a bottom up fashion is described in chapter 5 section 5 of [Sal88]. Essentially, the records are copied to newly allocated empty pages as they arrive. When a new page is added, no splitting is necessary. The first page is filled to a pre assigned fill factor, and then the next records go in the next page. Each new page requires a new entry in the level above. At ....

Betty Salzberg. File Structures: An Analytic Approach. Prentice Hall, 1988.

No context found.

Salzberg, B. "File Structures, An analytic approach" ISBN 0-13-314691-X; Prentice Hall.

No context found.

B. Salzberg. File Structures: an analytic approach. Prentice-Hall, Englewood Cli#s, NJ, USA. ISBN 0--1331--4550--6, 1988.

No context found.

B. Salzberg. File Structures: An Analytic Approach. Prentice-Hall, Inc., Englewood Cliffs, New Jersey, 1988.

No context found.

Salzberg B., "File Structures, an Analytic Approach", Prentice-Hall International, Inc, 1988.

No context found.

Salzberg B., "File Structures, an Analytic Approach", Prentice-Hall International, Inc, 1988.

No context found.

Betty Salzberg. File Structures: An Analytic Approach. Prentice Hall, 1981.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC