Han Shen, Beng Chin Ooi, and Hongjun Lu. The tp-index: A dynamic and efficient indexing mechanism for temporal databases. In Proceedings of the Tenth International Conference on Data Engineering, February 14-18, 1994, Houston, Texas, USA, pages 274--281. IEEE Computer Society, 1994.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works, must be obtained from the IEEE. Contact: Manager, Copyrights and Permissions IEEE Service Center 445 Hoes Lane P.O. Box 1331 Piscataway, NJ 08855 1331, USA. Temporal Databases [SOL94] Quality Classes, Personalization and Fuzzy Logic Matching where intervals can be utilized to describe the problem; Spatial Data where a spatial object can be approximated by a bounding box or set of intervals on a space filling curve. FR89, BKK99, KMPS01] For point data there are only a few ....

....DBMS technology, e.g. the external Segment Tree [BG94] is a nontrivial mapping of the Segment Tree. Further more we want to focus on indexing structures which can be used by exploiting the techniques of commercial RDBMSs, e.g. indexes like the B Tree or UB Tree. Therefore, we do not consider [SOL94, GLOT96, KS91] which require indexing techniques not available in any commercial DBMS. Ore90] utilizes an approach quite similar to ours, also featuring a Z curve based access method and the parameter space transformation. However, he focuses on spatial objects and spatial joins, but not on ....

[Article contains additional citation context not shown here]

Han Shen, Beng Chin Ooi, and Hongjun Lu. The tp-index: A dynamic and efficient indexing mechanism for temporal databases. In Proceedings of the Tenth International Conference on Data Engineering, February 14-18, 1994, Houston, Texas, USA, pages 274--281. IEEE Computer Society, 1994.

.... The proposed approach may be combined with any algorithm for the implementation of the coalesce operator and allows the exploitation of index structures (i.e. queries retrieving tuples having valid times included in a given range amongst others [1] 3] 8] 12] 15] 18] 20] 21] [22]) The remainder of this paper is organised as follows: in section 2 the optimisation scheme is described in detail and it is proved that the optimised execution scheme yields the same result with the coalesce valid time selection operator sequence. In section 3 the performance of the ....

H. Shen, B. Ooi and H. Lu, "The TP--Index: A Dynamic and Efficient Indexing Mechanism for Temporal Databases", Proceedings of the 10 th IEEE International Conference on Data Engineering, pp. 274-281, Houston, Texas, February 1994.

....join algorithms for the GT Join and GE Join. Top down and sideways traversal algorithms are presented. We examine other approaches, including the unsynchronized approach, the B tree and R tree based synchronized approaches, and an approach based on spatial partitioning (using the TP Index [21]) Results from an extensive experimental evaluation are also presented. Our experimental study shows that the sideways, linkbased, synchronized traversal is the most robust among the examined methods. Depending on the join characteristics, other methods can be competitive and should be considered ....

....intervals so that each partition fits in memory. 16] proposed a (plain) T Join algorithm based on spatial partitioning. Here a record s interval i is mapped to a point (i:start, i:end i:start) in a two dimensional space. These points are then indexed by an R tree like method (the TP Index [21]) which partitions the space. However, a partition in one relation may be joined with many partitions in the other relation. When an R tree is used as an index, a temporal join can be considered as a special case of a spatial join. 4] presents a depth first while [10] proposes a breath first ....

H. Shen, B. Ooi and H. Lu, "The TP-Index: A Dynamic and Efficient Indexing Mechanism for Temporal Databases", Proc. of ICDE, 1994.

....TSB tree [LS89, LS90] the MVBT [Bec96] the Two Level Time Index [EWK93] the R tree [Gut84] and the Segment Rtree [Ko193] a variant of the R tree specifically suited for temporal databases. Temporal index structures like the Snapshot Index [TK95] the Time Index [EWK93, EKW91] and the TP Index [SOL94] aim only at supporting specific query types effciently. Comparing them with other index structures is only meaningful based on a specific kind of application. Among the index structures with a general aim, the TSB tree has demonstrated very good query performance [ST99] Therefore, we have chosen ....

H. Shen, B.C. Ooi, H. Lu, The TP-Index: A Dynamic and Efficient Indexing Mechanism for Temporal Databases, Proc. Data Engineering Conf., 1994

....access methods are outperformed by factors of up to 42 for disk accesses and 4.9 for query response time. 1Introduction There is a growing demand for database applications that handle temporal and spatial data. Intervals occur as transaction time and valid time ranges in temporal databases [SOL 94] Ram 97] B 98] as line segments on a space filling curve in spatial applications [FR 89] BKK 99] as inaccurate measurements with tolerances in engineering databases, for hierarchical type systems in object oriented databases [KRVV 93] Ram 97] or for handling interval and finite domain ....

....than the minimum O(log n r) which Edelsbrunner s interval tree guarantees. The concept of time splits is introduced as a successful heuristics to avoid large fruitless scans. Again, the augmentation is an obstacle for the integration into commercial systems. The TP Index of Shen, Ooi and Lu [SOL 94] is based on a transformation of intervals into a triangular 2D space. Duplicates are avoided and the index is well suited for appending intervals since the data space may grow dynamically at the upper bound. The access method is highly specialized to the suggested mapping, and an integration ....

Shen H., Ooi B. C., Lu H.: The TP-Index: A Dynamic and Efficient Indexing Mechanism for Temporal Databases. Proc. IEEE Int. Conf. on Data Engineering, 274-281, 1994.

....addresses 2TBDs. In what follows we shall present, rather briefly, the research developed by Elmasri et al. [EWK90, EKW91, EJK92, EJK93, K 94] Time Index and its derivatives, Lomet and Salzberg [LS89, LS90, LS93] Time Split B tree, Segev and Gunadhi [GS93] AP tree, and by Shen et al. [SOL94] TP index. All these authors attempt to include some type of temporal dimension into a tuple. We focus on the indexing structures used and how the temporal indexing is accomplished. A more introductory survey can be found in [NE95a] while a more analytical survey that encompasses the above ....

....the problem of having a two level indexing, in a very similar fashion as in the 2TI, presenting the NST (Nested Tree) where the higher level structure indexes a non temporal key and points to an AP tree, indexing the history of such a key. The TP index (where TP stands for Time Polygon) SOL94] has a similar approach to the TI but uses a different technique to store the data. It maps temporal data into a two dimensional space, clustering based on time. Then this data is partitioned into polygons which correspond to data pages. In dealing with the time polygons, which are arranged ....

H. Shen, B.C. Ooi, and H. Lu. The TP-Index: A dynamic and efficient indexing mechanism for temporal databases. In Proceedings of the Tenth International Conference on Data Engineering, pages 274--281, Houston, TX, February 1994. IEEE.

....but not both. Hence none really addresses 2TDBs. In what follows we present, rather briefly, the research developed by Elmasri and others [4, 13] Time Index and its derivatives, Lomet and Salzberg [14] Time Split B tree, Segev and Gunadhi [8] Append Only tree, Shen et al. (Time Polygon Index) [18] and by Ang and Tan [1] IBT. All these authors attempt to include some type of temporal dimension in a tuple. We focus on the indexing structures used and how the temporal indexing is accomplished. A more introductory survey, covering also other papers not listed above, can be found in [16, 15] ....

....Segev also addressed the problem of having a two level index, in a very similar fashion as in the 2TI. This is called the NST (Nested Tree) where the higher level structure indexes a non temporal key and points to an AP tree, indexing the history of such a key. The Time Polygon Index (TP index) [18] maps temporal data into a two dimensional space, clustering based on time. This data is then partitioned into polygons which correspond to data pages. In dealing with the time polygons, which are arranged hierarchically, the TP index is similar to the R Tree [9] in some aspects. As in the TI, The ....

H. Shen, B.C. Ooi, and H. Lu. The TP-Index: A dynamic and efficient indexing mechanism for temporal databases. In Proceedings of the Tenth International Conference on Data Engineering, pages 274--281, Houston, TX, February 1994. IEEE.

....table be used, the updates would cost (in the expected amortized sense) O(1) The cost of the queries for the R tree and the TSB tree are expect behavior, as the worst case behavior happens seldomly and is O(N=B) Recently, a few other papers have come to our attention. The Time Polygon Index [SOL94] maps temporal data into a two dimensional space. It remings a R tree [Gut84] as the mapped temporal space is divided in pre determind shapes, however, this leads to one of its drawbacks of the scheme, which suggests that such mapping is biased towards some specific types of queries only. The ....

H. Shen, B.C. Ooi, and H. Lu. The TP-Index: A dynamic and efficient indexing mechanism for temporal databases. In Proceedings of the Tenth International Conference on Data Engineering, pages 274--281, Houston, TX, February 1994. IEEE.

....in our experimental study. An actual performance study of different types of temporal databases would entail using time indexes to better optimize the unpack, join and other temporal operations. Quite a number of studies have been done on the development of time indexes for temporal databases [6, 7, 21, 22]. However, all of them assume an underlying temporal model based on tuple timestamping. It is clear that further research should also be carried out for the addition of a time index to improve processing of certain class of temporal queries in a temporal model using attribute timestamping. Once ....

Shen, H., Ooi, B.C., Lu, H., "The TP-Index: A Dynamic and Efficient Indexing Mechanism for Temporal Databases," in Proc. of the International Conference on Data Engineering, 1994.

....for implementation and comparison by utilizing the generated benchmarks. Some of the indexing structures we will consider are the following (this is not an exaustive list) ffl Those considering valid time ranges only: Time Index [EWK90, EJK92, K 94] MAP21 [NDK95] Time Polygon Index [SOL94] and Interval B tree [AT95] ffl Those considering valid time but also assuming that input data is monotonically growing (append only) Monotonic B tree [EJK92] and Append Only Tree [GS93] ffl Those indexing transaction time only: Time Split B tree [LS93] and Snapshot Index [TK93] ffl ....

H. Shen, B.C. Ooi, and H. Lu. The TP-Index: A dynamic and efficient indexing mechanism for temporal databases. In Proceedings of the Tenth International Conference on Data Engineering, pages 274--281, Houston, TX, February 1994. IEEE.

....few. Some of these structures are also proven to guarantee optimal query time [BGO 93, ST95] We do not elaborate much on these structures due to limited space. There are also index structures that are used for dynamic management of time intervals. Segment R trees [KS91] R trees [Gut84] TP index [SOL94] are some of them to name a few. These structures are mostly multidimensional index structures that can be directly used (such as R trees) to index time intervals, or variations of multidimensional index structures tailored for indexing temporal domains (such as TP index) Being a popular spatial ....

....There is also some work that takes the approach of utilizing B trees for temporal indexing. The use of regular B trees for indexing valid time intervals was suggested in [GLOT96] Here, first, the intervals are mapped to two dimensional points with the same mapping function used for TP index [SOL94]. These two dimensional points are mapped back to one dimensional points (not intervals) by defining a total order among them using either horizontal, vertical, or diagonal sweep lines. B trees are used to index these points after the final transformation. Temporal queries also go through these ....

H. Shen, B. C. Ooi, H. Lu, " The TP-Index: A Dynamic and Efficient Indexing Mechanism for Temporal Databases", IEEE Data Engineering Conference , pages 274-281, 1994.

....[LS89, LS90] the MVBT [Bec96] the Two Level Time Index [EWK93] the R tree [Gut84] and the SegmentR tree[Kol93] a variant of the R tree specifically suited for tempo ral databases. Temporal index structures like the Snapshot Index [TK95] the Time Index [EWK93, EKW91] and the TP Index [SOL94] aim only at supporting specific query types efficiently. Comparing them with other index structures is only meaningful based on a specific kind of application. Among the index structures with a general aim, the TSB tree has demonstrated very good query performance [ST94] Therefore, we have ....

H. Shen, B. C. Ooi, H. Lu, "The TP-Index: A Dynamic and Efficient Indexing Mechanism for Temporal Data- bases", Proc. Data Engineering, 1994

....few. Some of these structures are also proven to guarantee optimal query time [BGO 93, ST95] We do not elaborate much on these structures due to limited space. There are also index structures that are used for dynamic management of time intervals. Segment R trees [KS91] R trees [Gut84] TP index [SOL94] are some of them to name a few. These structures are mostly multidimensional index structures that can be directly used (such as R trees) to index time intervals, or variations of multidimensional index structures tailored for indexing temporal domains (such as TP index) Being a popular spatial ....

H. Shen, B. C. Ooi, H. Lu, " The TP-Index: A Dynamic and Efficient Indexing Mechanism for Temporal Databases", IEEE Data Engineering Conference , pages 274-281, 1994.

.... LLV [K 94] 23 13 Handling Very Long Lived Versions VLLV [K 94] 23 14 Sketch of a two level AP tree : 26 15 The five polygons used in the TP index [SOL94] 28 16 A tentative new polygon shape : 29 17 Bitemporal indexing using a two level approach : 33 18 Bitemporal indexing using shareable trees : ....

....will be able to handle the floating queries. On the other, we propose a data structure which should do it. 3 Indexing Techniques for TDB In what follows we present a summary of several researches on indexing techniques for TDBs [KS89, LS89, LS90, LS93, EKW91, EKW91, EJK92, EJK93, GS93, SOL93, SOL94] We present such a summary in a format that may be more suitable for quick review, yet with details to explain the concepts being introduced. Also the text that follows may lack some flow , as it is intended to be presented as a (sort of) fact sheet. In what follows, for each technique, we ....

[Article contains additional citation context not shown here]

H. Shen, B.C. Ooi, and H. Lu. The TP-Index: A dynamic and efficient indexing mechanism for temporal databases. In Proceedings of the Tenth International Conference on Data Engineering, pages 274--281, Houston, TX, February 1994. IEEE.

....none will be able to handle the floating queries. On the other, we propose a data structure which should do it. 3 Indexing Techniques for TDB In what follows we present a summary of several researches on indexing techniques for TDBs [KS89, LS89, LS90, LS93, EKW91, EKW91, EJK92, EJK93, GS93, SOL93, SOL94] We present such a summary in a format that may be more suitable for quick review, yet with details to explain the concepts being introduced. Also the text that follows may lack some flow , as it is intended to be presented as a (sort of) fact sheet. In what follows, for each ....

....use of a pointer that links the key to the most recent version of the record, which then saves traversing the temporal index, the trade off is only one additional pointer in the upper level (indexing the key) tree. 3. 5 Shen, Ooi and Lu, 1994 The TP index (where TP stands for Time Polygon) SOL93, SOL94] has a similar approach to the TI but uses a different technique to store the data. It maps temporal data into a two dimensional space, applying clustering based on time. Then this data is partitioned into polygons which correspond to data pages. In dealing with the time polygons, which ....

H. Shen, B.C. Ooi, and H. Lu. The TP-Index: A dynamic and efficient indexing mechanism for temporal databases. Technical Report TR6/93, National University of Singapore, June 1993.

....section. problem of having a two level indexing, in a very similar fashion as in the 2TI, presenting the NST (Nested Tree) where the higher level structure indexes a non temporal key and points to an AP tree, indexing the history of such a key. The TP index (where TP stands for Time Polygon) SOL94] has a similar approach to the TI but uses a different technique to store the data. It maps temporal data into a two dimensional space, clustering based on time. Then this data is partitioned into polygons which correspond to data pages. In dealing with the time polygons, which are arranged ....

H. Shen, B.C. Ooi, and H. Lu. The TP-Index: A dynamic and efficient indexing mechanism for temporal databases. In Proceedings of the Tenth International Conference on Data Engineering, pages 274--281, Houston, TX, February 1994. IEEE.

....hashing table be used, the updates would cost (in the expected amortized sense) O(1) The cost of the queries for the R tree and the TSB tree are expected behavior, as the worst case behavior happens seldomly and is O(N=B) Recently, a few other papers have come to our attention. The TP Index [12] maps temporal data into a two dimensional space. The mapped temporal space is divided into pre determined shapes. However, this leads to one of its drawbacks of the scheme, which suggests that such mapping is biased towards some specific types of queries only. The Interval B tree [1] uses several ....

.... we compare the efficiency in using the MAP21 indexing structure to that of the Time Index [3] Some of the reasons for using the Time Index as a comparison tool are the following: i) it is a relatively simple data structure; ii) it has already been used as a comparison benchmark in [1] and [12]; and (iii) it also allows retroactive and proactive updates. 6.1 Time Index Review In this section we briefly review the Time Index (for further details we refer the reader to [3, 4, 8] As this paper focuses on MAP21, we just comment on how the Time Index is used in order to process the ....

H. Shen, B.C. Ooi, and H. Lu. The TP-Index: A dynamic and efficient indexing mechanism for temporal databases. In Proceedings of the Tenth International Conference on Data Engineering, pages 274--281, Houston, TX, 1994. IEEE.

....[LS89, LS90] the MVBT [Bec96] the Two Level Time Index [EWK93] the R tree [Gut84] and the SegmentR tree[Kol93] a variant of the R tree specifically suited for temporal databases. Temporal index structures like the Snapshot Index [TK95] the Time Index [EWK93, EKW91] and the TP Index [SOL94] aim only at 3 supporting specific query types efficiently. Comparing them with other index structures is only meaningful based on a specific kind of application. Among the index structures with a general aim, the TSB tree has dem onstrated very good query performance [ST94] Therefore, ....

H. Shen, B. C. Ooi, H. Lu, "The TP-Index: A Dynamic and Efficient Indexing Mechanism for Temporal Databases", Proc. Data Engineering, 1994

....few have addressed the issue of indexing temporal data. In what follows we briefly review some of the work done in the area. We do not intend for this review to be exhaustive though. For a good survey we refer the reader to [ST94] The Time Index [EWK93] Time Index [K 94] TP index [SOL94] Interval B tree [AT95] and the B tree based TP index 2 [G 96] all index valid time ranges, therefore no ordering in the ranges input is imposed. The drawback of the Time Index is the size of the index itself which is quadratic on the number of indexed ranges. The Time Index is ....

....simulated (we discuss such simulation in Appendix A.1) We chose the Time Index because it is a conceptually simple structure, which despite its inefficient use of storage may yield good average query processing time. It has also been used as a reference structure in other published research [SOL94, K 94, AT95, G 96] Furthermore, the TP index and the Interval B tree have rather unique internal data structures, and we believe that the use of simple data structures, such as the B trees, is a very desirable important feature of temporal indices, if they aim at being of practical ....

[Article contains additional citation context not shown here]

H. Shen, B.C. Ooi, and H. Lu. The TP-Index: A dynamic and efficient indexing mechanism for temporal databases. In Proceedings of the 10th IEEE International Conference on Data Engineering, pages 274--281, Houston, TX, February 1994.

....entity for a given time instant or an interval. To speed up the process of retrieving versions that were valid for a given time, we need to build indices on the temporal dimension, which brings us to the problem of efficient indexing of the lifespan intervals of object versions [10] 13] 24] [26], 28] In spatial databases, the intervals may themselves be objects; so indexing intervals is essential for retrieving objects [23] In our domain of deformation study experiments, it is necessary to index the time intervals of content objects so that temporal relationships among them can be ....

....point of the query interval is reached. The difference between two approaches is demonstrated in Figure 5.2. We have the flexibility to choose one method or the other based on our knowledge of the distribution of the intervals. 10 20 R (22) 17) 41) 10,13] 14,17] 7,8] 6,11] 4,22] 20,32] [26,41] [29,33] C C C 1 2 3 Figure 5.2 An IB tree of order 3 with height 2 Example 5.1: Consider the IB tree in Figure 5.2. Assume that we want to find the intervals that intersect with the query interval [18,25] We can evaluate this query in two ways. In the first one, we find the first intersecting ....

Shen, H., Ooi, B.C., Lu, H., "The TP-Index: A Dynamic and Efficient Indexing Mechanism for Temporal Databases", IEEE Data Engineering Conference, pages 274-281, 1994.

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