J. M. Hellerstein. Optimization techniques for queries with expensive methods. TODS, 23(2):113--157, 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....2 5 10 20 50 time PullUp PushDown S updated techniques in [RSSB00] to materialized views selection and maintenance. Our work is also related to materialized view maintenance [BLT86] GMS93] Han87] SR88] In our method, intermediate files are updated incrementally. Hellerstein [HS93] Hel94][Hel98] proposed a method called Predicate Migration to produce an optimal plan for queries with expensive methods. Our work is related to this work in that we also try to find an optimal plan for queries by considering pulling up selections above joins. However, we focus on optimizing multiple queries ....

J. M. Hellerstein, "Optimization Techniques for Queries with Expensive Methods," TODS 23(2): 113-157, 1998.

....Optimization Related Work Within the intermediate query representations, there are a lot of opportunities for syntactic calculus or algebraic transformations. A lot of experience has been gained applying these optimizations in the relational or object oriented context [CS96, Cha98, GLR97, Hel98, Ioa96, KPH98, PS96, PS97, RS93, SdBB96, SO95a, SO95b, WM99] This thesis focuses on semantic query optimization, i.e. query transformations based on semantic knowledge rather than syntactic equivalence. In particular, semantic knowledge is represented by association rules of the form: ....

J.M. Hellerstein. Optimization techniques for queries with expensive methods. ACM Transactions on Database Systems, 23(2):113--157, 1998.

....is to determine the set of filter conditions that are to be evaluated using GradeSearch. The rest of the conditions will be evaluated by using Probe. In order to efficiently execute the latter step, we will exploit the known techniques in optimizing the processing of expensive filter conditions [25, 22, 23, 26, 11]. In this section, we first define a space of search minimal executions, which access as few attributes as possible using GradeSearch, and sketch the cost model and the optimization criteria. Next, we describe an optimization algorithm and explain the conditions under which it is optimal. ....

....objects in the repository, IOal Sel(a) o. Optimizing Evaluation of Residues: Given a residue (a, f) the task of determining an optimal eval uation for (a, f) maps to the well studied problem of optimizing the execution of selection conditions containing expensive predicates [25] See also [23, 26, 22, 11]. If (a, f) is a conjunction of atomic conditions a A. A a, there is an efficient algorithm w that finds the optimum probing strategy. Specifically, it can be shown [23, 26] that the order in which the atomic conditions for each object should be probed is given by the rank of each condition ....

[Article contains additional citation context not shown here]

J. M. Hellerstein. Optimization techniques for queries with expensive methods. ACM Transactions on Database Systems, 23(2): 113-157, Sept. 1998.

....for implementing user defined table operators which are able to process tables or tuple streams as input parameters. Furthermore, optimizing queries containing this kind of operators is an important but still open issue. Here, techniques considering foreign functions [5] or expensive predicates [12] during optimization have to be extended. 7. ....

J. Hellerstein. Optimization Techniques for Queries with Expensive Methods. TODS, 23(2):113--157, 1998.

....crucial issues are the study of pattern selection commutativity for useful classes of patterns. The formal study of selection criteria for pattern classes that are more complex than frequent sets is to be done. A framework for object oriented query optimization when using expensive methods [7] can also serve as a basis for optimization strategies. 3 Inductive Databases and KDD Processes Already in the case of a unique class of patterns, real life mining processes are complex. This is due to the dynamic nature of knowledge acquisition, where gathered knowledge often affects the search ....

J. M. Hellerstein. Optimization techniques for queries with expensive methods. ACM Transaction on Database Systems, 1998. Available at http://www.cs.berkeley.edu/jmh/miscpapers/todsxfunc.ps.

....(psflt) Matched Filter (mflt) Channel Model (cm) Figure 11. Relations between channel modeling components in S 4 W. is this form of specification concise, it also enables us to use well known query optimization techniques to push costly operations deeper into the computational pipeline [47]. In particular, query based representations of model instances lose the distinction between conducting a simulation to collect data and looking up (already simulated) data from a database. Coupled with grid information services, such a representation can help determine if specified simulations ....

J.M. Hellerstein. Optimization Techniques for Queries with Expensive Methods. ACM Transactions on Database Systems, Vol. 23(2):pages 113--157, September 1998.

....a sequence of model instances, which can be associated with corresponding simulations and scheduled for execution. Not only is this form of specification concise, it also enables us to use well known query optimization techniques to push costly operations deeper into the computational pipeline [47]. In particular, query based representations of model instances lose the distinction between conducting a simulation to collect data and looking up (already simulated) data from a database. Coupled with grid information services, such a representation can help determine if specified simulations ....

J.M. Hellerstein. Optimization Techniques for Queries with Expensive Methods. ACM Transactions on Database Systems, Vol. 23(2):pages 113--157, September 1998.

.... Thus, only 1 4 of the tuples of S 2 will join (on the date attribute) with tuples of S 1 , and each tuple of S 2 will generate 8 tuples in the result of the join (S 1 has 8 images per day) Hence, the join between IRSat and OzoneSat is not selective, and most existing optimization techniques like [4, 13] will generate a QEP in which the predicate 1 The image collections are very closely inspired from real life data sources, available at http: www.ssec.wisc.edu data comp ir and ftp:daac.gsfc.nasa.gov. RR n 4239 4 I. Manolescu, L. Bouganim, F.Fabret, E.Simon LowOzone (imgHDF) boolean s ....

....for a sorting step. They propose a hybrid hash scheme to ensure the cache ts in memory. We envision using cache as soon as the cost of a program invocation is more expensive than a cache lookup; also, the proposed hybrid hash can be adapted to our algorithm to e ciently manage the cache bu er. In [4, 13], query optimization in the presence of costly predicates is addressed, but unfortunately, the results developed there do not apply to our context. The reason is that predicate ranking [13] and the methods described in [4] are based on the function costs being constant per tuple, which no longer ....

[Article contains additional citation context not shown here]

Joseph M. Hellerstein. Optimization techniques for queries with expensive methods. ACM Transactions on Database Systems (TODS), 23(2):113157, 1998.

....algebra at the physical layer. Most systems do not use a cost function at this level but just (simple) heuristics that usually produce better expressions than the initial input. Commonly known heuristics are: push select project down or the more sophisticated predicate move around [HS93, LMS94, Hel98] most restrictive join rst. Logical algebra to physical algebra translator This stage aims at translating the logical algebra expression into the most ecient program at the physical level. It is usually at this stage that a cost model of the physical layer is used to nd the cheapest ....

Joseph M. Hellerstein, Optimization Techniques for Queries with Expensive Methods, ACM Transactions on Database Systems 23 (1998), no. 2, 113-157.

....through a single remote access to the relevant data sources, instead of one remote access for each CQ. 4. 5 Optimization of Other Complex Trigger Expressions A potential topic for future work is to optimize the processing of selection predicates containing OR s, joins, or very expensive functions [10]. The main idea is to explore optimization opportunities inherent in the particular properties of each type of operators. Consider the trigger conditions that contain OR s. If a single one of the OR ed clauses is true, then the entire predicate is true. By properly ordering the processing of OR ed ....

J. Hellerstein. Optimization Techniques for Queries with Expensive Methods. ACM Transactions on Database Systems, (To appear), 1988. Available at www.cs.berkeley.edu/ jmh.

....a single remote access to the relevant data sources, instead of a single remote access for each CQ. 16 4. 5 Optimization of Other Complex Trigger Expressions A potential topic for future work is to optimize the processing of selection predicates containing ORs, joins, or very expensive functions [10]. The main idea is to explore optimization opportunities inherent in the particular properties of each type of operators. For example, when a conjunct of the given trigger condition containing ORs, in general none, some or all of the ORed clauses it may be true. One extreme case is that if one of ....

J. Hellerstein. Optimization Techniques for Queries with Expensive Methods. ACM Transactions on Database Systems, (To appear), 1988. Available at www.cs.berkeley.edu/ jmh.

....unnecessary calls to external functions, because they are often expensive. For example, if an external function is deterministic, then two calls to that function having the same arguments will yield the same result. Therefore, we cache the arguments and results of each external function call as in [11]. When preparing to call a function, we rst check whether the answer has already been computed. 6 Parallel computation: The parallelization technique used for hash joins can also be used to parallelize the execution of the matching operator, if the neighborhood hash join execution optimization ....

J. M. Hellerstein. Optimization techniques for queries with expensive methods. ACM Trans. on Database Systems, 23(2):113-157, 1998.

....unnecessary calls to external functions, because they are often expensive. For example, if an external function is deterministic, then two calls to that function having the same arguments will yield the same result. Therefore, we cache the arguments and results of each external function call as in [10]. When preparing to call a function, we could rst check whether we had already computed the answer. 6 Parallel computation: The parallelization technique used for hash joins can also be used to parallelize the execution of the matching operator, if the neighborhood hash join execution ....

J. M. Hellerstein. Optimization techniques for queries with expensive methods. ACM Trans. on Database Systems, 23(2):113-157, 1998.

....the quality of resulting plan, and the scalability of the method, and to understand the tradeoffs between the choices. We also plan to study and incorporate the state of art research in efficient rule processing systems [1, 4, 5, 14] and methods for efficiently processing of expensive functions [6] in the proposed project. 3.3 Implementation Strategy Trigger Pattern Extractor Trigger Pattern Recognizer Filter Test Driver Polling Trigger Executor CQ Trigger Index Manager Trigger Condition Tester Distributed Trigger Evaluator Figure 2: Distributed trigger manager architecture ....

J. Hellerstein. Optimization Techniques for Queries with Expensive Methods. ACM Transactions on Database Systems, 1998. Available at www.cs.berkeley.edu/~ jmh.

....by Gupta et al. Gupt89] They cite several types of parallelism that can be exploited, including node, intranode, action, and data parallelism. These overlap with the types of concurrency we outlined in section 6. Work by Hellerstein on performing selections after joins in query processing [Hell98] is related to the issue of performing expensive selections after joins in Gator networks and A TREAT networks [Kand98] Proper placement of selection predicates in Gator networks can improve trigger system performance, and thus scalability. The developers of POSTGRES proposed a markingbased ....

Hellerstein, J., "Optimization Techniques for Queries with Expensive Methods," to appear, ACM Transactions on Database Systems (TODS). Available at www.cs.berkeley.edu/~jmh.

No context found.

J. M. Hellerstein. Optimization techniques for queries with expensive methods. TODS, 23(2):113--157, 1998.

No context found.

J. M. Hellerstein. Optimization techniques for queries with expensive methods. TODS, 23(2):113--157, 1998.

No context found.

J. M. Hellerstein. Optimization techniques for queries with expensive methods. TODS, 23(2):113--157, 1998.

....would go in steps, from (R . 1 S) 2 T to (R . 2 T ) 1 S and then to (T . 2 R) 1 S. This approach treats an operator and its right hand input as a unit (e.g. the unit [ 2 T ] and swaps units; the idea has been used previously in static query optimization schemes [IK84, KBZ86, Hel98] Viewing the situation in this manner, we can naturally consider reordering multiple joins and their inputs, even if the join algorithms are different. In our query (R . 1 S) 2 T , we need [ 1 S] and [ 2 T ] to be mutually commutative, but do not require them to be the same join algorithm. ....

....join between R and S, and an index join between S and T . Since our data is uniformly distributed, Table 1 indicates that the selectivity of the RS join is 1:8 10 4 ; its selectivity with respect to S is 180 i.e. each S tuple entering the join finds 1. 8 matching R tuples on average [Hel98] We artificially set the selectivity of the index join w.r.t. S to be 10 (overall selectivity 1 10 5 ) Figure 7 shows the relative performance of our two eddy schemes and the two static join orderings. The results echo our results for selections, showing the lottery based eddy performing ....

J. M. Hellerstein. Optimization Techniques for Queries with Expensive Methods. ACM Transactions on Database Systems, 23(2):113--157, 1998.

No context found.

J. M. Hellerstein. Optimization techniques for queries with expensive methods. ACM Transactions on Database Systems, 22(2):113--157, June 1998.

No context found.

Hellerstein JM. Optimization techniques for queries with expensive methods. ACM Transactions on Database Systems 1998; 23(2):113--157.

No context found.

Hellerstein, J., "Optimization Techniques for Queries with Expensive Methods, " ACM Transactions on Database Systems, no. To appear, 1998. Available at www.cs.berkeley.edu/ jmh.

No context found.

Joseph M. Hellerstein. Optimization Techniques for Queries with Expensive Methods. ACM Transactions on Database Systems, 23(2):113--157, June 1998.

No context found.

J. Hellerstein. Optimization techniques for queries with expensive methods. ACM TODS, 23(2):113--157, 1998.

No context found.

Joseph M. Hellerstein. Optimization techniques for queries with expensive methods. ACM Transactions on Database Systems (TODS), 23(2):113--157, 1998.

First 50 documents

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