A. Kemper and G. Moerkotte. Access support relations: An indexing method for object bases. Information Systems, 17(2):117--146, 1992.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....and integrity constrains. However, while object oriented systems are extremely efficient in navigational type of data querying, their lack of ability to process efficiently large amounts of data to evaluate complex queries based on attribute values is also widely recognized (see, for example, [12]) More specifically, pure object databases are not well suited to support powerful high level query languages inherent to relational and post relational systems. Actually, to introduce any querying facility it is necessary to relax one of the major principles of the object paradigm, namely ....

....Information Systems. Moscow, June 27 30, 1995. the presence of set valued attributes makes set comparison operations more frequent. Several data structures were proposed to support the valuebased querying in the object oriented databases, including join indices [20] access support relations [12], nested indices [2] and signature files [10] Many of these structures were introduced for completely different environments and their relevance and efficiency for support of object querying is not obvious. Therefore, it is important to evaluate the behavior of different indexing structures in ....

[Article contains additional citation context not shown here]

Alfons Kemper and Guido Moekotte. Access support relations: an indexing method for object bases. Inf. Syst. (Oxford), 17(2):117--145, 1992.

.... query processing has been devoted to their optimization: also in this case, the focus is on transforming pointer chasing operations which are considered rather expensive into joins of pointer sets stored in auxiliary access structures, such as class extents [24] access support relations [25] and join indices [38] 41] Although it may seem that a similar approach may be extended to the Web, we show that the more involved nature of access paths in Web sites and the absence of ad hoc auxiliary structures introduces a number of subtleties. We compare two main approaches to query ....

.... reminiscent of the ones that have been proposed for relational databases to optimize selections on a relation with multiple indices [31] and for object oriented query processors to reduce pointer chasing in evaluating pathexpressions assuming a join index on professors and courses is available [25]. However, the following two examples show that, in the Web context, this is not always the optimal solution: in some cases, pointer chasing is less expensive. This is shown in the following example. Example 9.2 [Pointer chasing] Consider the scheme in Figure 1, and suppose we need to answer the ....

A. Kemper and G. Moerkotte. Access support relations: An indexing method for object bases. Information Systems, 17(2):117--145, 1992.

....the problem of efficiently evaluating path queries in this area [CCM96, FS98, MW97] In order to optimize path query processing, several indexing schemes have been proposed in the past for objectoriented databases. Example of such schemes are Path Indexes [BK89, SB96] and Access Support Relations [KM92], which materialize frequently traversed paths in the database in order to support navigation along reference chains leading from one object to another. However, since these approaches are based on the paths found in the schema, it is not possible to use them for the typical schema less XML ....

A. Kemper and G. Moerkotte. Access support relations: an indexing method for object bases. Information Systems, 17(2): 117-145, 1992.

....predicates on in path nodes. Such queries, however, may require the search of many index pages. Path indexes were described and analyzed in [1] A variation of the pathindexing method which can be used to support Joins in relational databases is called Access support relations and is discussed in [5]. In this paper a uniform scheme combining both forms of indexing is suggested. This scheme, called the U index, provides support for class hierarchy with the advantages of directly accessing sub classes (like H trees) while retaining the good performance of single key retrieval of CH trees. The ....

G. Moerkotte and A. Kemper. Access support relations an indexing method for object bases. Information Systems, 17(2):117--145 (1992).

....no address. Materializing joins is attractive if queries ask for all attributes whereas outer joins are also attractive if queries just ask for subsets of the attributes. Some of the related theory of materializing joins and outer joins has been developed in the context of access support relations [18]. Redundancy: just as in any other database, materialized views can be established in order to speed up the execution of queries. Finding the right views to materialize has been an active research area in the context of data warehouses [17, 33] It is likely to be at least as di cult in the ....

A. Kemper and G. Moerkotte. Access Support Relations: an indexing method for object bases. Information Systems, 17(2):117146, 1992.

....into single mixed approach. We show that the mixed approach is not only efficient but can also provide an optimal (otherwise, a near optimal) vertical partitioning scheme. In contrast to partitioning, indexing is a facility in OODBs to reduce the number of disk IOs in query execution (e.g. see [BK89,KM90,KM92]) by reducing the accesses to irrelevant object instances (as compared with sequential scanning) Indexing reduces disk IOs at the object instance level; it still accesses irrelevant instance variables, as not all the instance variables accessed are relevant to the query. Partitioning is also a ....

A. Kemper and G. Moerkotte, "Access Support Relations: An Indexing Method for Object Bases", in Information Systems, Vol. 17, No. 2, pages 117--146, 1992.

.... Examples of such data models are the object oriented model [Cat97] and the object relational data model [SM96] Several indexes dealing with special problems in these data models have been invented, e.g. nested indexes [BK89] path indexes [BK89] multi indexes [MS86] access support relations [KM92], join index hierarchies [XH94] The predominant problem attacked by these index structures is the efficient evaluation of path expressions. However, the problem of indexing data items with set valued attributes is of no less importance. Consider for example the following queries: retrieve all ....

A. Kemper and G. Moerkotte. Access support relations: An indexing method for object bases. Information Systems, 17(2):117--146, 1992.

.... query processing has been devoted to their optimization: also in this case, the focus is on transforming pointer chasing operations which are considered rather expensive into joins of pointer sets stored in auxiliary access structures, such as class extents [17] access support relations [18] and join indices [29] 31] Although it may seem that a similar approach may be extended to the Web, we show that the more involved nature of access paths in Web sites and the absence of ad hoc auxiliary structures introduces a number of subtleties. We compare two main approaches to query ....

.... reminiscent of the ones that have been proposed for relational databases to optimize selections on a relation with multiple indices [22] and for object oriented query processors to reduce pointer chasing in evaluating pathexpressions assuming a join index on professors and courses is available [18]. However, the following two examples show that, in the Web context, this is not always the optimal solution: in some cases, pointer chasing is less expensive. This is shown in the following example. ProfListPage PiProfList oe Rank= 0 F ull 0 ToProf e 6 ProfPage PiCourseList ToCourse ....

A. Kemper and G. Moerkotte. Access support relations: An indexing method for object bases. Information Systems, 17(2):117--145, 1992.

....of large attributes from all other kinds of data considerably speeds up structure oriented queries. And keeping relations between objects (nodes) separately is an approved technique for accelerating navigational queries (in contrast to path indexes in [47] and access support relations in [46], GRAS even avoids duplication of intrinsic relations within an additional index structure) y Partial match queries of the forms ( edge type, and ( attribute value) are not supported. z Clustering within one storage will be discussed below. 8 Bernhard Westfechtel et al. On the ....

....them as a special form of derived node reference attributes, our attribute evaluation algorithm is able to materialize their values on demand. In this way, GRAS is able to efficiently maintain derived information about graphs in a similar way as e.g. in the object oriented database system GOM [33, 46]. But note that the actual attribute (and relation) dependencies of our sample graph in fig. 7 do not really exist in the form of additional relations, as in the system GOM, but will be deduced during the propagation process by means of a static attribute dependency graph (for further details see ....

[Article contains additional citation context not shown here]

A. Kemper, G. Moerkotte. Access Support Relations: An Indexing Method for Object Bases. Information Systems , vol. 17-2, Pergamon Press, pp. 117-145 (1992).

....about query processing has been devoted to their optimization. In this research, the focus is on transforming pointer chasing operations which are considered rather expensive into joins of pointer sets stored in auxiliary access structures, such as class extents [7] access support relations [8] and join indices [19] 20] Although it may seem that a similar approach may be extended to the Web, we show that the more involved nature of access paths in Web sites and the absence of ad hoc auxiliary structures introduces a number of subtleties. We compare two main approaches to query ....

.... of the ones that have been proposed for relational databases to optimize selections on a relation with multiple indices [12] and for object oriented query processors to reduce pointer chasing in evaluating path expressions assuming a join index on professors and courses is available [8]. However, the following two examples show that, in the Web context, this is not always the optimal solution: in some cases, pointer chasing is less expensive. This is shown in the following example. Example 2. Pointer chasing] Consider the scheme in Figure 1, and suppose we need to answer the ....

A. Kemper and G. Moerkotte. Access support relations: An indexing method for object bases. Information Systems, 17(2):117--145, 1992.

....dynamically generated directly from a database without regard to any fixed schema or class hierarchy. For query optimization, we show how the DataGuide can be used as a path index. Substantial research on object oriented query optimization has focused on the design and use of path indexes, e.g. [BK89, CCY94, KM92]. In general, previous work has required explicit specification of the paths to index. The issues of how to create, maintain, and use a path index in a semistructured data model like OEM, where the set of paths in a database may often change over time, have not to the best of our knowledge been ....

....that uses information maintained by a strong DataGuide to significantly speed up query processing for a broad class of Lorel queries. Essentially, a strong DataGuide can also serve as a path index. While path indexes have been studied for traditional object oriented database systems, e.g. [BK89, CCY94, KM92], they are typically created for user specified path expressions; in a semistructured environment, the set of path expressions may be in flux, and isolating useful paths to index may be difficult. Conveniently, we can build and incrementally maintain a comprehensive path index for all possible ....

A. Kemper and G. Moerkotte. Access Support Relations: An Indexing Method for Object Bases. Information Systems, pp. 117-145, 17(2), 1992.

....execution of updates. Whenever an update is performed, some of the precomputed joins become invalid and must, thus, be recomputed. Aggregation indexing techniques can therefore be classified based on whether they maintain some additional information to efficiently perform updates [4, 18] or not [6, 9]. In the former case, no object accesses, except to the modified object, are needed to recompute the join and update the index. In the latter case, objects other than the modified object need to be accessed to recompute the join and update the index. The third group of techniques provides an ....

....for object database systems, thus providing foundations for other researches in this area. It is important to note that, because of its implementation simplicity, the path index is quite attractive and extensions or variations to it have been proposed (see for example the notion of access relation [9]) It is therefore useful to extend the path index to increase its application scope. Moreover, an access structure similar to the path index has been implemented as part of the ObjectStore OODBMS (see the notion of index path in [16] To the best of our knowledge, no other indexing techniques ....

A. Kemper and G. Moerkotte. Access Support Relations: An Indexing Method for Object Bases. Information Systems, 17(2):117--145, 1992.

....path index. A preliminary version of Lore s query language, Lorel, was introduced in [QRS 95] Details of the current version of Lorel appear in [AQM 97] Needless to say, there has been a significant amount of work in indexing for object oriented databases, e.g. KKD89, SS94, RK95, KM92, CCY94, BG92, XH94] All of this work depends on the database having a fixed schema based on a known, strongly typed class hierarchy. In our environment we must take a different approach to indexing, since we do not have a fixed schema, and comparable objects may take on different types. Other ....

A. Kemper and G. Moerkotte. Access support relations: An indexing method for object bases. Information Systems, 17(2):117--145, 1992.

....of the Cartesian product of two relations the join result will cover. In contrast, we argue for a strategy based on the so called fan out of specific object or literal instances. Like some early work on logic program optimization [9] and some recent proposals for object oriented databases [5], we see joins as directed rather than undirected operations. In a sense, the fan out approach requires a left deep strategy in join order optimization which is known to not always find the optimal solutions [12] However, the exploitation of materialized views or pre computed subqueries ....

A. Kemper and G. Moerkotte. Access support relations: An indexing method for object bases. Information Systems, 17(2):117--145, 1992.

....support indexes for translating attribute values into tuple ids (e.g. B trees or hash tables) In object oriented databases, path queries replace the simpler associative queries. Several data structures have been proposed for answering path queries efficiently: e.g. access support relations [14] and path indexes [4] In the case of semistructured data, queries are even more complex, because they may contain regular path expressions [1, 7, 8, 16] The additional flexibility is needed in order to traverse data whose structure is irregular, or partially unknown to the user. For example the ....

....some indexes to speed up the evaluation. Index structures developed for traditional data models are tied to a pre defined schema: e.g. relational databases index on a specific attribute of a specific relation, while objectoriented databases index on a specific path in the object oriented schema [4, 14], e.g. document:section:title. Hence, these index structures are not applicable to semistructured data, because here the schema is unavailable. Full text indexing systems take an opposite approach: given no knowledge on the structure of information, they index all the data. But this is of ....

[Article contains additional citation context not shown here]

Alfons Kemper and Guido Moerkkotte. Access support relations: an indexing method for object bases. Information Systems, 17(2):117--145, 1992.

.... the pointer based data structures to devise proper indexing schemes and retrieve objects efficiently has been identified as a very important issue to further improve the system performance [12] 13] 15] Several indexing schemes have been proposed for nested attribute queries [1] 2] 3] 8] [10] [16] Three index organizations for use in the evaluation of a query in an OODB are introduced in [3] As an extension to [3] performance of path indexes for queries containing several predicates is evaluated in [2] In [10] query processing in an OODB system is improved by maintaining separate ....

.... schemes have been proposed for nested attribute queries [1] 2] 3] 8] 10] 16] Three index organizations for use in the evaluation of a query in an OODB are introduced in [3] As an extension to [3] performance of path indexes for queries containing several predicates is evaluated in [2] In [10] query processing in an OODB system is improved by maintaining separate structures to redundantly store objects which are frequently traversed by database queries. A hybrid indexing technique, called a generalized index, is proposed in [8] to support class hierarchy with complex and primitive ....

A. Kemper and G. Moerkotte. Access Support Relations: An Indexing Method for Object Bases. Information Systems, 17(2):117--145, 1992.

....support indexes for translating attribute values into tuple ids (e.g. B trees or hash tables) In object oriented databases, path queries replace the simpler associative queries. Several data structures have been proposed for answering path queries efficiently: e.g. access support relations [14] and path indexes [4] In the case of semistructured data, queries are even more complex, because they may contain generalized path expressions [1, 7, 8, 16] The additional flexibility is needed in order to traverse data whose structure is irregular, or partially unknown to the user. For example ....

....to use some indexes to speed up the evaluation of such queries. Index structures developed for traditional data models rely on some pre defined database schema: e.g. relational databases index on a specific attribute of a specific relation, while object oriented databases index on a specific path [4, 14] in the object oriented schema (e.g. document:section:title) Hence, these index structures are not applicable to semistructured data, because the schema is missing, unavailable, or only partially known. At the other extreme, full text indexing systems take an opposite approach. Given no ....

[Article contains additional citation context not shown here]

Alfons Kemper and Guido Moerkkotte. Access support relations: an indexing method for object bases. Information Systems, 17(2):117--145, 1992.

....dynamically generated directly from a database without regard to any fixed schema or class hierarchy. For query optimization, we show how the DataGuide can be used as a path index. Substantial research on object oriented query optimization has focused on the design and use of path indexes, e.g. [BK89, CCY94, KM92]. In general, previous work has been based on the class hierarchies of the object oriented model. The 3 issues of how to create, maintain, and use a path index in a semistructured data model such as OEM have not to the best of our knowledge been addressed. 1.2 Paper Outline Section 2 first ....

....how the information maintained by a strong DataGuide can be used to significantly speed up query processing for a broad class of Lorel queries. Essentially, a strong DataGuide can also serve as a path index. While path indexes have been studied for traditional object oriented systems, e.g. [BK89, CCY94, KM92], their use in a semistructured environment has not been addressed. In particular, creating and maintaining a path index without a fixed schema may be quite difficult, yet we can conveniently use strong DataGuides to address the problem. As shown in Section 4 for incremental maintenance, each ....

A. Kemper and G. Moerkotte. Access Support Relations: An Indexing Method for Object Bases. Information Systems, pp. 117-145, 17(2), 1992.

.... how to utilize the pointer based data structures to devise proper indexing schemes and retrieve objects efficiently has been identified as a very important issue to further improve the system performance [11, 12] Several indexing schemes have been proposed for nested attribute queries [1, 2, 3, 8, 9, 13]. Three index organizations for use in the evaluation of a query in an OODB are introduced in [3] As an extension to [3] performance of path indexes for queries containing several predicates is evaluated in [2] In [9] query processing in an OODB system is improved by maintaining separate ....

.... schemes have been proposed for nested attribute queries [1, 2, 3, 8, 9, 13] Three index organizations for use in the evaluation of a query in an OODB are introduced in [3] As an extension to [3] performance of path indexes for queries containing several predicates is evaluated in [2] In [9] query processing in an OODB system is improved by maintaining separate structures to redundantly store objects which are frequently traversed by database queries. A hybrid indexing technique, called a generalized index, is proposed in [8] to support class hierarchy with complex Names ....

A. Kemper and G. Moerkotte. Access Support Relations: An Indexing Method for Object Bases. Information Systems, 17(2):117--145, 1992.

....on evaluation of queries with set valued predicates is few and far between. Several indexes dealing with special problems in the object oriented [7] and the objectrelational data models [23] have been invented, e.g. nested indexes [4] path indexes [4] multi indexes [21] access support relations [18], join index hierarchies [27] The predominant problem attacked by these index structures is the efficient evaluation of path expressions. With the exception of signature files [16] and Russian Doll Trees [14] the problem of indexing data items with set valued attributes has been neglected by the ....

A. Kemper and G. Moerkotte. Access support relations: An indexing method for object bases. Information Systems, 17(2):117--146, 1992.

....while forward traversal of object references are usually very well supported by specialized access mechanisms and would not be treated as ordinary join operations, this is not true for backward traversal. This would require appropriate index structures such as Access Support Relations [KM92] processing of which, in turn, involves handling of potentially very complex join expressions for both initial materialization and maintenance. Hence, there is a demand for optimization techniques that can cope with such complex queries in a cost effective manner. In this paper, we shall examine ....

A. Kemper and G. Moerkotte. Access Support Relations: an indexing method for object bases. Information Systems, 17(2):117--146, 1992.

....while forward traversal of object references are usually very well supported by specialized access mechanisms and would not be treated as ordinary join operations, this is not true for backward traversal. This would require appropriate index structures such as Access Support Relations [KM92] processing of which, in turn, involves handling of potentially very complex join expressions for both initial materialization and maintenance. Hence, there is a demand for optimization techniques that can cope with such complex queries in a cost effective manner. In this paper, we shall examine ....

A. Kemper and G. Moerkotte. Access Support Relations: an indexing method for object bases. Information Systems, 17(2):117--146, 1992.

....QUEL like language GOMql, or the procedural, C like language GOMpl. 2.1 Access Support Relations An Access Support Relation (ASR) is an index structure that supports the evaluation of so called path expressions. We provide a short review of the main principles; details can be derived from [2, 3]. Access Support Relations constitute a generalization of binary join indices originally proposed for the relational model [4] and later extended for object models [5, 6] Binary join indices are an extension of links, developed by Harder [7] The binary links were also applied to object models in ....

A. Kemper and G. Moerkotte, "Access Support Relations: an indexing method for object bases," Information Systems, vol. 17, no. 2, 1992.

....amount of formal work carried out in the context of the relational model over more than two decades. 4. Many queries naturally return relations and not objects (see also [27] on this) 5. Allowing formal reasoning with new (relation based) optimization techniques, e.g. access support relations [16, 17] and generalized materialization relations [15] For us, these arguments seemed cogent enough to incorporate the relational model into our formal model of object oriented databases. 2.1 Sorts Within this subsection we start specifying the value part of our model. We assume that the set of ....

....the query is (entirely) based on extensions and returns a relation, the knowledge of relational query processing and optimization can be utilized. But still, some object oriented specialities must be considered. First, access structures especially developed for object oriented databases (e.g. [15, 17]) must be taken into account. Second, the cost model for estimating the overhead of different evaluation alternatives must be adapted to the particularities of the object oriented model. Furthermore, since object oriented data models have more expressive power than the relationals the optimization ....

A. Kemper and G. Moerkotte. Access Support Relations: an indexing method for object bases. Information Systems, 17(2):117--146, 1992.

....This descriptor, however, need not be consulted if a resident page is re accessed. 200,280) 40,80) 40 41 42 79 . 200 279 Directory Buffer Pages Figure 5: The Directory of a List 3.4. 2 Swizzling and Indexes A reference can be a key of an index, e.g. of Access Support Relations [KM92] or of an index of a large set. It does not make sense to swizzle these references since they are never dereferenced. Furthermore, swizzling these references could induce a complete reorganization of the index, e.g. a B tree [Com79] or a hash table [FNPS79] As a consequence, an index cannot be ....

....swizzling strategy is the most profitable was described. In this section, how, in practice, the variables of the cost model can be derived is outlined, and as a result, the most profitable swizzling strategy can be determined. Some of the ideas were borrowed from clustering in object bases [GKKM92b] there, the same problem of combining characteristics of the object base and the application s profile was addressed. 7.1 Monitoring the Variables of the Cost Model The application s profile can be determined by monitoring. The application is executed in the training mode using no swizzling. ....

[Article contains additional citation context not shown here]

A. Kemper and G. Moerkotte. Access Support Relations: an indexing method for object bases. Information Systems, 17(2):117--146, 1992.

No context found.

A. Kemper and G. Moerkotte. Access support relations: An indexing method for object bases. Information Systems, 17(2):117--146, 1992.

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