I. S. Mumick, S. J. Finkelstein, H. Pirahesh, and R. Ramakrishnan. Magic is Relevant. In Proceedings of ACM SIGMOD, Atlantic City, NJ, pp. 247-258 (1990).  Home/Search   Document Details and Download   Summary   Related Articles   Check

....queries for database systems with a powerful view concept. This is in particular the case for systems which will implement the new SQL:1999 standard, and hence will allow the definition of recursive views. The attractiveness of this method lies in its generality and e#ciency. Additionally, in [8, 12] it has been shown that the Magic Set method can improve the performance of nonrecursive queries as well. Thus, it seems worthwhile to implement a Magic Set transformation and a fixpoint based evaluation mechanism on top of a (non)recursive relational database system. In this paper we propose such ....

Mumick, I. S., Finkelstein, S. J., Pirahesh, H., Ramakrishnan, R.: Magic is Relevant. SIGMOD Conference 1990: 247-258.

....hence cannot lead to different proof processes. The general problem, however, of losing strafication after the magic set transformation using an allowed SIP strategy has been applied remains and can be solved, e.g. by using weak stratification [11] or by using the alternating fixpoint [21] In [8, 13] it has been shown that the magic set method can improve the performance of nonrecursive queries as well and hence it seems worthwhile to apply the dynamic approach to nonrecursive systems, too. This is especially true if evaluation costs of body literals differ considerably and the evaluation of ....

Mumick, I. S., Finkelstein, S. J., Pirahesh, H., Ramakrishnan, R.: Magic is Relevant. SIGMOD Conference 1990: 247-258.

....wildcard selections of the form e= A, can be evaluated with a transitive closure operator that walks through the XML tree e to nd all branches with tag A. Transitive closures are very hard to optimize and very few general techniques have been proposed in the literature, such as magic sets [25]. To address this problem, we are planning to incorporate structural recursion over tree types to our XML algebra. Our starting point will be our previous work on structural recursion over tree like data structures [13] which satis es very e ective optimization rules, reminiscent to loop fusion ....

I. Mumick, S. Finkelstein, H. Pirahesh, and R. Ramakrishnan. Magic is Relevant. In Proceedings of the

....P[i] semantics = INS P[i] out.i expr : P[i] out.i expr UNION Reduce(P[i] expr) P[i] semantics = DEL P[i] out.d expr : P[i] out.d expr UNION Reduce(P[i] expr) End case P[i] out.expr : P[i] out.i expr MINUS P[i] out.d expr End for End Where Reduce(expr) 1. Use the technique of [41] to represent SQL queries; if self references exist (e.g. in the case of DEL statements) discriminate between multiple occurrences of the same table. 2. Use the reduction techniques of [48, 33 , 36] wherever applicable to reduce the query definition to a compact form. Fig. 18. Algorithm ....

....set of types belonging to the set SourceSchema, denoting all the types found in the data sources, are treated as source nodes of a graph. For the rest of the types, we can derive an SQL expression by using existing view reduction algorithms, such as [33] corrected with the results of [19, 44, 45] [14, 11, 41, 48, 36]. Our algorithm is applicable to graphs of activities that do not involve updates. In most cases, an update operation can be considered as the combination of insertions and deletions or as the application of the appropriate function to the relevant attributes. The results of the application of ....

I. Mumick, S. Finkelstein, H. Pirahesh, R. Ramakrishnan. Magic is Relevant. In Proc. ACM SIGMOD Intl. Conf. on the Management of Data, pp. 247-258, Atlantic City, NJ (1990).

....no joins in a query, then the only improvement introduced by query unnesting would come from moving predicates between the inner and outer query, which may result in more selective unnest operations. Since predicate move around has already been shown to be e ective by others [Levy et al. 1994; Mumick et al. 1990], we decided to do benchmarks over queries with joins, by translating outer joins into outerblock nested loops. This modi cation was accomplished mainly by mapping path expressions, such as e.dept.instructors in Query 2, into pointer joins between class extents. This optimization, which is known ....

....There is a number of proposals lately that, instead of rewriting correlated queries into more ecient queries or algebraic forms, propose new evaluation algorithms that perform better than the default nested loop evaluation of embedded queries. Examples of such methods include magic decorrelation [Mumick et al. 1990], which promotes predicates in the inner loop of the nested loop evaluation of correlated queries, and predicate move around [Levy et al. 1994] which moves predicates around the query graph. Our decorrelation algorithm not only moves predicates to the right place, it also intermixes operators ....

Mumick, I., Finkelstein, S., Pirahesh, H., and Ramakrishnan, R. 1990. Magic is Relevant. In Proceedings of the 1990 ACM SIGMOD International Conference on Management of Data, Atlantic City, NJ (May 1990), pp. 247-258.

....is a number of proposals lately that, instead of rewriting correlated queries into more ecient queries or algebraic forms, they propose new evaluation algorithms that perform better than the default nested loop evaluation of embedded queries. Examples of such methods include magic decorrelation [MFPR90] which promotes predicates in the inner loop of the nested 41 loop evaluation of correlated queries, and predicate move around [LMS94] which moves predicates around the query graph. Our decorrelation algorithm not only moves predicates to the right place, it also intermixes operators between ....

I. Mumick, S. Finkelstein, H. Pirahesh, and R. Ramakrishnan. Magic is Relevant. In Proceedings of the 1990 ACM SIGMOD International Conference on Management of Data, Atlantic City, NJ, pages 247-258, May 1990.

....algebra and transformation rules incorporate the handling of duplicates and order. We consider operations that eliminate or preserve duplicates. The S equivalence type corresponds to permitting duplicates, e.g. it allows replacing a query expression with a set equivalent one. Mumick et al. [MPR90, Mum90] study the extension of the Magic Sets technique for programs containing multisets and aggregates. They note that the implementation of multisets is efficient, since duplicate checks are not needed. They provide a formal basis for reasoning about optimization techniques when multisets are ....

I. S. Mumick, S. J. Finkelstein, H. Pirahesh, and R. Ramakrishnan. Magic is Relevant. In Proceedings of ACM SIGMOD, Atlantic City, NJ, pp. 247-258 (1990).

....magic rewriting, program decomposition 1 1 Introduction The technique of magic rewriting was expressed for linear rules by F. Bancilhon, D. Maier, Y. Sagiv, and J.D. Ullman [2] and independently, by Rohmer, Lescoeur, and Kerisit [15] Since then, several generalizations have been reported [3, 4, 7, 10, 17]. Generally speaking, the magic rewriting is a transformation that, given a query and a program, produces a new program whose search in computation is guided by the binding information in the query. This new program is called a magic program and consists of two parts: the first part searches for ....

....queries. A straightforward method for computing the answer to query Q is to apply the above rewriting for single binding queries for each binding in B. If the same sips is used in all such rewritings, then only the magic seed is different; everything else is exactly the same. It was observed in [6, 7, 8, 9], for left linear, right linear, and mixed linear rules, that such a fragmented tuple at a time computation suffers from a large amount of redundant work, and an approach based on a single rewriting and single run was proposed for those rules. The basic idea is to partition instantiation of rules ....

I.S. Mumick, S.J. Finkelstein, H. Pirahesh, R. Ramakrishnan, "Magic is relevant," in proceedings of The ACM SIGMOD International Conference on Management of Data, Atlantic City, May 23-25, 1990, ACM, New York, pp. 247-258

....affecting them. For the rest of the types, we can derive an SQL expression by using existing view reduction algorithms. Several complementary proposals exist such as [Kim82 corrected with the results of GaWo87, Mura89, Mura92 (which we will mention as Kim82 in the sequel) Daya87] ChSh96] [MFPR90], PiHH92] LeMS94] The proposed algorithm is applicable to graphs of activities that do not involve updates. In most cases, an update operation can be considered as the combination of insertions and deletions or as the application of the appropriate function to the relevant attributes. ....

....P[i] semantics = INS P[i] out.i expr : P[i] out.i expr UNION Reduce(P[i] expr) P[i] semantics = DEL P[i] out.d expr : P[i] out.d expr UNION Reduce(P[i] expr) End case P[i] out.expr : P[i] out.i expr MINUS P[i] out.d expr End for End Where Reduce(expr) 1. Use the technique of [MFPR90] to represent SQL queries; if self references exist (e.g. in the case of DEL statements) discriminate between multiple occurrences of the same table. 2. Use the reduction techniques of [PiHH92] Kim82 ] LeMS94] wherever applicable to reduce the query definition to a compact form. Fig. 11. ....

I. Mumick, S. Finkelstein, H. Pirahesh, R. Ramakrishnan. Magic is Relevant. In ACM SIGMOD Conference, 1990.

....early on. One line of research has concentrated on extending the traditional selection propagation techniques to nested queries. In these approaches, traditional optimizers are enhanced with additional execution plans, where selection and join predicates are applied as early as possible [MFPR90a, MFPR90b, MPR90, LMS94]. Another line of work has proceeded in an orthogonal direction, introducing execution plans which correspond to alternative structures of nesting. In particular, these approaches consider the possibilities of merging query blocks, denesting queries, and commuting aggregation blocks with joins ....

....joins [Day87, GW87, Kim82, Mur92, PHH92, YL94, HG94] In this paper we propose an approach which unifies and generalizes the approaches mentioned above. We apply the selection propagation idea to certain data dependencies that are implicit in aggregation blocks. Propagation of SQL predicates [MFPR90a, MFPR90b, MPR90, LMS94] is a special case of propagation of these dependencies. At the same time, propagating these 1 dependencies can produce execution plans with alternative nesting structure, as in [Day87, GW87, Kim82, Mur92, PHH92, YL94, HG94] In addition to expressing in a common framework previously proposed ....

[Article contains additional citation context not shown here]

I. Mumick, S. Finkelstein, H. Pirahesh, R. Ramakrishnan. Magic is relevant. In SIGMOD 1990.

....semantics [33, 37] This ensures the applicability of bottom up evaluation and optimization methods to both recursive algebraic expressions and logic rules. Examples of research on algebraic optimization may be found in [5, 7, 10, 13, 16, 26, 37] examples of logic optimization may be found in [11, 12, 39, 41, 42, 43, 47, 48, 51, 52, 68]. In this section we recall results about the use of the transitive closure operator for answering recursive queries. Though the material of this Section is self contained, a general knowledge of the Datalog language might be useful, as it can be achieved through the reading of [14, 44, 56] ....

I.S. MUMICK, S.J. FINKELSTEIN, AND H. PIRAHESH, "Magic is relevant," in Proc. ACM SIGMOD Conference, Atlantic City, 1990.

....(right hand side) of a rewrite rule) programmed in C. Because they are programmed with a general purpose programming language, Starburst rules are able to express a wide variety of transformations including view merging and query unnesting (both discussed in [15] and magic sets transformations ([14]) However, Starburst rules are difficult to understand and reason about, requiring a detailed understanding of the underlying graph based query representation (QGM) Expressing Complex Query Rewrites with Extended Rewrite Rules: Exodus [2] and its successors, Volcano [9] and Cascades [8] and ....

....This predicate holds of an employee e and a department d if e is highly paid or if e is a manager for d and e s salary is more than the average for d. Figure 1b shows an equivalent predicate that has been normalized into CNF. 2 A similar schema is used in papers from the Starburst groupsuch as [14] and [18] P AND Q AND R) OR S (a) P OR S) AND (Q OR S) AND (R OR S) b) P = e:Dno = d:Dno Q = e:Sal d:ASal R = e:Job = Mgr S = e:Sal e:Bon) 100K Figure 1: A Predicate (a) and its CNF Equivalent (b) 2.2 KOLA and CNF KOLA is the language of queries and rewrite rules assumed ....

[Article contains additional citation context not shown here]

I. S. Mumick, S. J. Finkelstein, H. Pirahesh, and R. Ramakrishnan. Magic is relevant. In Proc. ACM SIGMOD Int'l Conference on Management of Data, pages 247--258, 1990.

....the relationship to semijoin more intuitive and less magical. SELECT P.eid, P.sal FROM PartialResult P, LimitedDepAvgSal V WHERE P.did = V.did AND P.sal V. avgsal The above technique can be used in a multi block query containing view (including recursive view) definitions or nested subqueries [42,43,56,57]. In each case, the goal is to avoid redundant computation in the views or the nested subqueries. It is also important to recognize the tradeoff between the cost of computing the views (the view PartialResult in the example above) and use of such views to reduce the cost of computation. The formal ....

Mumick, I.S., Finkelstein, S., Pirahesh, H., Ramakrishnan, R. Magic is Relevant. In Proc. of ACM SIGMOD, Atlantic City, 1990.

....However, it works only on hierarchical queries, which are nonrecursive queries without common subexpressions. Another approach is the adaptation of the magic set transformation for an early evaluation of selection and join predicates in nonrecursive SQL queries with common subexpressions [MFPR90a]. The magicset transformation (see [Ull89] for details) pushes predicates according to the order of doing joins and introduces auxiliary magic views. In this paper, we propose a generalization of the predicate pushdown technique, called predicate movearound, that is capable of pushing predicates ....

....the problem of merging query blocks. Doing so eliminates the need for predicate pushdown. However, that can not always be done (e.g. in the presence of aggregation) Our method can be used in conjunction with the techniques of [PHH92] Our method complements magic sets [BMSU86, BR87] and GMST [MFPR90b, MFPR90a]. The key differences from the magic set approach are as follows. First, the magic set transformation depends upon the join order; it can move predicates up from a relation and then down into another relation that appears later in the join order. In contrast, predicate move around does not depend ....

I. Mumick, S. Finkelstein, H. Pirahesh, and R. Ramakrishnan. Magic is relevant. In SIGMOD 1990, pages 247--258.

.... been adapted to deal with SQL queries (which contain features such as grouping, aggregation, arithmetic conditions and multiset relations that are not present in pure logic queries) and found to be superior to techniques used in commercial database systems for nonrecursive nested SQL queries [MFPR90a] Other variations of magic sets include minimagic [SZ87] variants for propagating arithmetic constraints as selections [BKMR89, MFPR90b, SR92] a variant that can mimic the tail recursion optimization of Prolog systems [Ros91] and magic templates [Ram88] in which tuples with variables in them ....

.... to the development of Magic Sets for programs with SQL features like grouping, aggregation, and arithmetic selections [MPR90] A performance study demonstrated that the techniques developed for dealing with recursive queries actually beat the techniques used in current relational database systems [MFPR90a] They have also extended SQL with recursive view definitions, providing greater expressive power to SQL users, and utilizing the I O optimization facilities of SQL implementations in a direct manner for recursive applications. They have also made contributions to the evaluation of left and ....

[Article contains additional citation context not shown here]

I. S. Mumick, S. Finkelstein, H. Pirahesh, and R. Ramakrishnan. Magic is relevant. In Proceedings of the ACM SIGMOD International Conference on Management of Data, Atlantic City, New Jersey, May 1990.

....idea was introduced to deal with recursion, it provides significant improvements for non recursive queries as well. In [57; 58] it is shown that the technique can be extended to deal with SQL programs, including those containing features like group by, aggregation and arithmetic conditions. In [56] a performance evaluation, carried out on a DB2 relational system, is presented, demonstrating that the technique performs comparably to standard database techniques, and is often significantly better. The Magic and Alexander methods are based on program transformations. Other methods use a ....

I. S. Mumick, S. Finkelstein, H. Pirahesh, and R. Ramakrishnan. Magic is relevant. In Proceedings of the ACM SIGMOD International Conference on Management of Data, Atlantic City, New Jersey, May 1990.

....magic sets with other query optimization techniques. 2 We have done performance experiments on large benchmark data on IBM S DB2 database, and seen that queries transformed by our extended magic sets transformation can execute orders of magnitude faster. Some of the experiments were reported in [MFPR90a] Results of those and other experiments are summarized in Table 1. The experiments compare the execution times Elapsed Time Query Original Correlated EMST Exp A 100.00 0.40 0.47 Exp B 100.00 2.12 0.28 Exp C 100.00 513.27 50.24 Exp D 100.00 5136.49 109.00 Exp E 100.00 52.56 7.62 Exp F 100.00 0.54 ....

....and without using EMST. Then, after you determine the optimal join orders, perhaps you can use the join orders to improve the plan you just selected. We have experimented with our cost based heuristic, and the preliminary results are extremely encouraging. The performance experiments of Table 1 [MFPR90a] used the cost based heuristic (applied manually by iterating through the DB2 optimizer) to achieve impressive gains. The Starburst implementation provides a testbed for the above and other possible heuristics for investigating the interaction between cost based optimization and magic sets ....

[Article contains additional citation context not shown here]

I. Mumick, S. Finkelstein, H. Pirahesh, and R. Ramakrishnan. Magic is relevant. In SIGMOD 1990.

No context found.

I. S. Mumick, S. J. Finkelstein, H. Pirahesh, and R. Ramakrishnan. Magic is relevant. In Proceedings of ACM SIGMOD 1990 International Conference on Management of Data, pages 247--258, Atlantic City, NJ, May 23-25 1990.

....database systems process complex SQL queries involving views, table expressions and subqueries with aggregate functions. Such queries are becoming increasingly important in decision support applications (see, e.g. the TPC D benchmark [TPCD94] The magic sets rewriting technique (see, e.g. [BMSU86, RLK86, BR91, MFPR90, SS94]) has been proposed as a heuristic query transformation to optimize such queries, and can result in dramatic improvements in query execution This work was performed while the author was at the University of Wisconsin, Madison. y This work was performed while the author was at AT T Bell ....

....query transformation to optimize such queries, and can result in dramatic improvements in query execution This work was performed while the author was at the University of Wisconsin, Madison. y This work was performed while the author was at AT T Bell Laboratories, Murray Hill. performance [MFPR90]. There can be many possible variants of this rewriting even for a single query, based upon the decisions made with respect to binding propagation. Some of these variants can actually degrade performance. Prior to this work, there has been no demonstrated algorithm to efficiently choose a variant ....

[Article contains additional citation context not shown here]

I. S. Mumick, S. Finkelstein, H. Pirahesh, and R. Ramakrishnan. Magic is relevant. In Proceedings of ACM SIGMOD International Conference on Management of Data, 1990.

No context found.

I. S. Mumick, S. J. Finkelstein, H. Pirahesh, and R. Ramakrishnan. Magic is Relevant. In Proceedings of ACM SIGMOD, Atlantic City, NJ, pp. 247-258 (1990).

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