Chen, C., Roussopoulos, N. Adaptive Selectivity Estimation Using Query Feedback. SIGMOD, 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....to choose the cheapest of the possible equivalent expressions, it requires a simple cost model that can predict the cardinalities of intermediate results. Such a cardinality estimate draws heavily on a selectivity model. As such, selectivity estimation has been subject to extensive research [9, 66, 42, 30, 39, 11, 32, 20, 45, 10]. 9 In [25, 38] predicate reordering techniques have been proposed that are even more sophisticated. NF databases require a more sophisticated optimizer in general since the NF paradigm allows nesting of relations [52] This also holds for OO databases. As mentioned in Section 2 the modern ....

C.M. Chen and N. Roussopoulos. Adaptive Selectivity Estimation Using Query Feedback. In Proceedings of the 1994.

.... [28] This broad importance of statistics management has led to a plethora of approximation techniques, for which [15] have coined the general term data synopses : advanced forms of histograms [30, 16, 20] spline synopses [22, 23] sampling [6, 17, 14] and parametric curve fitting techniques [34, 9] all the way to highly sophisticated methods based on kernel estimators [2] or Wavelets and other transforms [26, 25, 4] However, most of these techniques take the local viewpoint of optimizing the approximation error for a single data distribution such as one database table with pre selected ....

C. M. Chen and N. Roussoploulos. Adaptive Selectivity Estimation Using Query Feedback. In Proceedings of the ACM SIGMOD Conference, pages 161--172, 1994.

....with respect to the data distribution without considering how the histograms will be used in estimating user query results. Ignoring user query patterns assumes that user queries are randomly distributed throughout the domain, which is rarely the case in practice. To remedy the second problem, CR94] uses query feedbacks from the query processor to adjust the model functions of the data distributions. However, for real datasets it it rare that the underlying data This work was partially supported by NSF grants EIA 9818320, IIS 98 17432, EIA 9986057 and IIS 99 70700. distributions can be ....

....static assumptions to some degrees, and can adapt to query workloads. 2. 3 Related Work on Adaptive Query Estimation Recently, proposals have been made to extend existing query estimation techniques to accommodate the dynamic natures of data distributions and or the query patterns [GMP97, DIR00, CR94, AC99, BCG01, CDN01, GLR00] In this section, we briefly review some of the techniques. For the category of tuple sampling that summarize the relation R, Ganti et al. GLR00] propose Icicle, a new class of tuple sampling technique that can tune itself to dynamic workloads. The intuition of their ....

Chung-Min Chen and Nick Roussopoulos. Adaptive selectivity estimation using query feedback. In Proceedings of the 1994 ACM SIGMOD International Conference on Management of Data, Minneapolis, Minnesota, May 24-27, 1994, pages 161--172, 1994.

....Kossmann [26] 4.4 Adaptive Query Processing. We believe that techniques for adaptive query processing will be very relevant for data management in sensor networks. Chen and Rousopoulos designed an adaptive selectivity estimation schema that adds statistics gathering to regular query processing [9]; we can envision the use of similar techniques in a sensor network setting where small feedback is piggybacked on results to long running queries. Lack of perfect global knowledge is also an inherent problem in distributed and heterogeneous database systems. One approach to adapting to this ....

C.-M. Chen and N. Roussopoulos. Adaptive selectivity estimation using query feedback. In R. T. Snodgrass and M. Winslett, editors, Proceedings of the 1994.

....is crucial for effective query optimization, and has received considerable research attention. Existing approaches can be classified into two categories depending on whether they take into account only the data distribution [HS92, IP95, GM98, APR99, WAA01] or also consider the query patterns [CR94, GLR00, BCG01, WAA02]. Although our framework can be used with any histogram, for the shake of simplicity and generality, we adopt the equi length method (in fact more sophisticated histograms lead to even better performance) Specifically, the data space is divided into nurnbi bins with equal extents, and ....

Chen, C., Roussopoulos, N. Adaptive Selectivity Estimation Using Query Feedback. ACM S1GMOD, 1994.

....to changes in the data distribution as well as changes in the query patterns. This approach successfully combines the re nement process of adapting to query interest regions with the maintenance process of the data stream. Instead of simply considering the range of a single query as in prior work [5, 1, 3, 27], which cannot capture the distribution of a global query workload that evolves over time. We therefore introduce a workload decay model that represents a weighted summary of the global query workload. Based on the workload decay model, a multi threshold function is proposed in order to eciently ....

....the histograms only based on the changes in the data stream. As has been observed in the database community, precisely capturing data distributions may not be enough for accurately answering queries and hence several proposals exploit the query answer from a DBMS to adjust summary statistics. [5] introduced the concept of using query feedback from the query execution engine. ST histograms [1] are re ned by distributing estimation errors over the domain. STHoles [3] allows nested buckets to capture data regions and drills holes (buckets) selectively from the area where the query feedback ....

[Article contains additional citation context not shown here]

C.M. Chen and N. Roussopoulos. Adaptive Selectivity Estimation Using Query Feedback. In Proceedings of the ACM SIGMOD Conference, pages 161-172, 1994.

....takes advantage of mappings that the extender has already created and optimized for the purpose of search. Second, most are based on space partitioning schemes; this results in summary data space requirements that are exponential in D. Model fitting techniques. Methods based on regression (e.g. [CHEN94, GRAE87, SUN93], wavelets (e.g. MATI98] and neural nets (e.g. BOUL97, LAKS98] can be used to summarize attribute frequency distributions. The proposed techniques have some additional disadvantages. First, like the parametric estimators discussed in this paper, they are all point estimators and provide no ....

....LAKS98] can be used to summarize attribute frequency distributions. The proposed techniques have some additional disadvantages. First, like the parametric estimators discussed in this paper, they are all point estimators and provide no interval bounds. Second, with a few exceptions (e.g. [CHEN94]) the ability to perform dynamic updates of the summary data is generally limited. Histograms. Now well established (see [POOS97] for a recent survey) conventional histograms rely on spacepartitioning schemes (which limits their applicability) When they can be applied, they consitute a very ....

C.M. Chen and N. Roussopoulos, "Adaptive Selectivity Estimation Using Query Feedback," Proc.

....selectivity estimation, which is estimating the percentage of tuples in the table that satisfy a given query. The selectivity of a query depends on the data distribution of the underlying data in the database. Several techniques have been proposed in the literature to estimate query result size [6, 1]. The histograms [3] techniques are the most commonly used form of statistics in practice (e.g. they are used in DB2, Informix, Ingres, Oracle, Microsoft SQL Server, Sybase) and are the focus of this paper. Various kinds of histograms have been proposed, e.g. equiwidth histograms, equi depth ....

Chung-Min Chen and Nick Roussopoulos. Adaptive selectivity estimation using query feedback. In Richard T. Snodgrass and Marianne Winslett, editors, SIGMOD'94, pages 161--172. ACM Press, 1994.

....strategies for unidimensional histograms. As we will see in Section 3, partitioning multidimensional spaces is challenging, and there are no obvious generalizations of these techniques for more than one dimension. The idea of using feedback from the query execution engine is introduced in [4]. Their approach is to represent the data distribution as a linear combination of model functions. The weighting coecients of this linear combination are adjusted using feedback information and a least squares technique. The main problem with this approach is that it depends on the choice of the ....

C.-M. Chen and N. Roussopoulos. Adaptive selectivity estimation using query feedback. In ########### ## ### #### ### ###### ############# ########## ## ##########

....in real life use [Loh94] and have started using more extensive statistical information about the underlying data. Several techniques have been proposed in the literature to estimate query result size [MCS88] including histograms [Koo80] sampling [LNS90, HS95] and parametric techniques [CR94, SLRD93] Of these, histograms approximate the data distribution of an attribute by grouping attribute values into buckets (subsets) and approximating true attribute values and their frequencies in the data based on summary statistics maintained in each bucket. The advantage of histograms over ....

Chung-Min Chen and Nick Roussopoulos. Adaptive selectivity estimation using query feedback. In Richard T. Snodgrass and Marianne Winslett, editors, SIGMOD'94, pages 161--172. ACM Press, 1994.

....another. By constrast, using index structures for estimation takes advantage of mappings that the extender has already created and optimized. Second, space partitioning schemes require storage exponential in D. Model fitting techniques. Methods based on regression, wavelets and neural nets [11, 32, 35] have been used to summarize attribute frequency distributions. The proposed techniques have some additional disadvantages. First, like the parametric estimators discussed in this paper, they are all point estimators and provide no interval bounds. Second, with a few exceptions, the ability to ....

C.M. Chen and N. Roussopoulos, "Adaptive Selectivity Estimation Using Query Feedback," Proc. 1994 SIGMOD, Minneapolis, MN, May 1994, 161-172.

.... different from the statistical approach taken by us and by Hellerstein et al. Statistical techniques: The three major classes of techniques used are sampling (e.g. H OT88, LNS90, HNS94, LN95, HNSS95, GGMS96] histograms (e.g. Koo80, PIHS96, Poo97, APR99] and parametric modeling (e.g. CR94] A survey of various statistical techniques is given in the paper by Barbara et al. [BDF 97] Gibbons and Matias present a framework for studying synopsis data structures for massive data sets [GM99b] and introduced two samplingbased synopses, concise samples and counting samples, that can be ....

C. M. Chen and N. Roussopoulos. Adaptive selectivity estimation using query feedback. In Proc. ACM SIGMOD International Conf. on Management of Data, pages 161--172, May 1994.

....[VH99] data set) It appears that the predicted selectivity values match the measured values. 1. 1 Related work In the literature, a large number of e#orts has been reported on the prediction of selectivity factors in di#erent contexts and under di#erent assumptions [Car75, Yao77, IB86, LNS90, CR94, IP95, GGMS96, PIHS96, CMN98, CMN99] Roughly two directions can be distinguished in the prediction of selectivity factors. Research in the first direction has been focussed to the prediction of the number of page or block accesses, to retrieve # tuples from R tuples which are randomly ....

....polynomial (or other function) is approximated using a least squared error estimation method, sampling queries are run against a sample taken from the real data to compute the selectivity for that sample, followed by extrapolating the selectivity on the sample to the whole data set. We refer to [CR94] for a more detailed description of each of these. Our problem definition certainly matches the one focussed on in this second research direction. But, since we are interested in estimating the selectivity for a fragmented database, it is not exactly the same. Furthermore, the model we propose ....

Chungmin Melvin Chen and Nick Roussopoulos, Adaptive Selectivity Estimation Using Query Feedback, Proceedings of the 1994 ACM SIGMOD International Conference on the Management of Data, ACM Press, May 1994, pp. 161--172.

....from the available data, without necessarily conforming to a formal process model. In this sense, the Work performed while with AT T Labs. data are allowed to speak for themselves. Of the nonparametric methods we consider two approaches: histograms [8, 7, 9, 16, 14, 15] and curve fitting [18, 1]. We briefly review some of the histogram methods (e.g. equiwidth, equidepth, end biased, maxdiff) in Sec. 2.1; an excellent taxonomy of histograms can be found in [16] Curve fitting approaches alternative to the proposed approach are considered in Sec. 6. Many data sets have continuous valued ....

....the maxdiff histogram [16, 15] and the compressed histogram [16, 15] the polynomial based method in [18] wavelet based histograms [13] etc. As shown in [16] the prevailing method is maxdiff(V,A) which we used in our experiments. Remotely related to our work is the query feedback approach of [1]; the use of linear regularization to obtain better estimates from histograms [4] and the CF kernel method of [12] to obtain a fast kernel estimation of the density in very large data sets. 7 Conclusions The main contribution of this paper is the recognition of the need to distinguish between ....

Chungmin M. Chen and Nick Roussopoulos. Adaptive selectivity estimation using query feedback. In Proc. of the ACM-SIGMOD, pages 161--172, Minneapolis, MN, May 1994.

....prototype developed at the University of Maryland [RK86, RES93, DR94] and obtains accurate cost estimates with small CPU overhead but no I O. The ACE module works together with another adaptive module of ADMS Sigma which estimates the selectivities from exactly the sizes of the returned results [CR94] 1.1 The Problem and Related Work Consider the distributed query shown in Figure 1 where two global query plans are considered. In the query, P is a selection predicate on R 1 , Theta a join predicate between R 1 and R 2 . Symbols oe and 1 denote selection and join respectively. The query is ....

....Software Client Comm. Soft. ADMS Buffers Net WAN LAN INGRES ADMS ORACLE6 ADMS ADMS ORACLE7 Figure 2: ADMS Sigma System Architecture directly affects the accuracy of the query cost estimation. ADMS uses another adaptive module, called Adaptive Selectivity Estimator (ASE) CR94] for interpolating the value distributions of attributes which are then used to estimate selectivities. ASE produces accurate estimates of record selectivities from real attribute value distributions which are adaptively approximated by a curve fitting polynomial using the query feedback ....

C. Chen and N. Roussopoulos. Adaptive Selectivity Estimation Using Query Feedback. In Proc. of ACM SIGMOD, 1994.

No context found.

Chen, C., Roussopoulos, N. Adaptive Selectivity Estimation Using Query Feedback. SIGMOD, 1994.

No context found.

C.-M. Chen and N. Roussopoulos. Adaptive selectivity estimation using query feedback. In SIGMOD '94.

No context found.

C. M. Chen and N. Roussopoulos. Adaptive selectivity estimation using query feedback. Proc. of ACM SIGMOD Conf, pages 161--172, May 1994.

No context found.

C. M. Chen and N. Roussopoulos. Adaptive selectivity estimation using query feedback. In Proceedings of the 1994.

No context found.

Chen, C., Roussopoulos, N. Adaptive Selectivity Estimation Using Query Feedback. SIGMOD, 1994.

No context found.

Chen C., Roussopoulos N.: Adaptive Selectivity Estimation Using Query Feedback. SIGMOD Conf. (1994) 161-172

No context found.

Chen, C.M., and N. Roussopoulos, Adaptive Selectivity Estimation Using Query Feedback. Proc. ACM SIGMOD, 1994, 161-172.

No context found.

Chung-Min Chen and Nick Roussopoulos. Adaptive selectivity estimation using query feedback. In Proceedings of the 1994 ACM SIGMOD Interna- 171 tional Conference on Management of Data, Minneapolis, Minnesota, May 24-27, 1994, pages 161--172, 1994.

No context found.

C. M. Chen and N. Roussopoulos. Adaptive selectivity estimation using query feedback. In Proceedings of the 1994.

No context found.

C. Chen. N. Roussopoulos. Adaptive Selectivity Estimation Using Query Feedback. ACM SIGMOD 1994.

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