P. Varman and R. M. Verma, "An efficient multiversion access structure," IEEE Transactions on Knowledge and Data Engineering, vol. 9.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....due to version updates. Therefore, retrieving a small segment could require reading a lot of unnecessary data. While UBCC is very effective at supporting full version retrieval queries, complex queries on content and history combined call for indexing techniques such are the Multiversion B Tree [17, 3, 27] and the Multiversion R Tree [13] We investigate the following three approaches: Scheme 1: single Multiversion B Tree, Scheme 2: UBCC with a Multiversion B Tree, and Scheme 3: UBCC with a Multiversion R Tree. The last two approaches still use the UBCC mechanism as the main storage scheme for ....

....versions, new elements are added, updated or deleted. These changes can update the above B tree using the element DNNs. In order to answer queries over a multiversion document we need to maintain the multiple versions of this B tree. Various multiversion B tree structures have been proposed [17, 3, 27]; here we consider the Multiversion B tree (MVBT) 3] which has optimal asymptotic behavior for partial version retrievals and its code is readily available. The MVBT has the form of a directed graph with multiple roots. Associated with each root is a consecutive version interval. A root provides ....

P. Varman and R. Verma, "An Efficient Multiversion Access Structure", IEEE TKDE 9(3), pp. 391-409, 1997.

....algorithms proposed in [27] either do not apply or need to be modified so as to solve the GT and GE join problems. In the rest of the paper we use the Multiversion B tree (MVBT) 3] as a temporal index. However, the proposed algorithms can be applied to other efficient temporal access methods [17, 25]. The main contributions of this paper are summarized below: We propose synchronized, temporal index based 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 ....

P. Varman and R. Verma, "An Efficient Multiversion Access Structure", TKDE 9(3), 1997.

....be modified. Partial persistence fits nicely with the degenerate evolution case since in that case an update corresponds to either an object addition or a deletion. Methods to make a disk based structure (in particular a B tree) partially persistent have appeared in the area of temporal databases [20, 5, 27, 49, 24, 36]. 24] presents the Bitemporal R Tree which is a partially persistent R Tree used to index bitemporal objects. This partiallypersistent R Tree can be easily extended to index the degenerate case of animated objects. The general evolution case where objects change continuously is different. One ....

.... is incorporated in the structure it may have to be merged with another node (this happens if y has less than D e alive records and is called a strong version underflow) or, key split into two nodes (if y has more than B e alive nodes, i.e. a strong version overflow) For details we refer to [24, 49, 5]. An example of a PPR Tree is shown in Figure 5 using the evolution presented in Figure 3 and Figure 4. In particular, Figure 3 shows the MBRs of 20 objects (numbered from 1 to 20) that appeared in a small animated video while Figure 4 depicts the lifetimes of these objects. Here B = 5, D = 2 and ....

P.J. Varman, R.M. Verma. An Efficient Multiversion Access Structure. In IEEE TKDE, 9(3): 391409, 1997.

....creates an exciting window of research opportunities. While there has been very little previous work on versioning web based semistructured documents [5, 21] much relevant work was done for other applications, such as software configuration management [11] CAD systems[10] and temporal databases [1, 3, 9, 12, 16, 18, 19]. Some of the problems occurring in multiversion documents are similar to those of transaction time databases, where object histories are maintained (new objects are added without discarding the old ones) 14] Using timestamping, various efficient indexing and clustering techniques have been ....

.... in multiversion documents are similar to those of transaction time databases, where object histories are maintained (new objects are added without discarding the old ones) 14] Using timestamping, various efficient indexing and clustering techniques have been proposed for temporal relations [1, 3, 9, 12, 16, 18, 19]. Version management schemes have also been proposed in OODBs; however, they are not designed to support documents and to optimize the retrieval of complex documents [2, 4, 8, 10] Change management for semistructured data has been proposed in Chorel [5] the focus though is on modeling version ....

[Article contains additional citation context not shown here]

P.J. Varman and R.M. Verma, An Efficient Multiversion Access Structure, IEEE Trans. on Knowledge and Data Engineering, Vol.9, No. 3, pp: 391-409, 1997.

....events based on the current trajectories of the points. If the trajectory of a point changes, the future versions of the structure might also change. Instead of storing each K i explicitly, we store the differences between K i Gamma1 and K i , using the ideas of persistent data structures [10, 16, 44]. Note that there are two main differences between our structures and standard persistent data structures. First, instead of storing all the past versions, we maintain only a few past versions and we delete a past version when it becomes too old. Second, we also maintain several future versions of ....

....also maintain several future versions of the data structure, which we must update when the trajectory of a point changes. Multiversion kinetic B trees. In the case of one dimensional moving points, we can directly apply the ideas of Driscoll et al. 16] Becker et al. 10] and Varman and Verma [44] to obtain a Indexing Moving Points 23 persistent (or multiversion) B tree. Roughly speaking, each data element is augmented with a life span consisting of the time at which the element was inserted and (possible) the time at which it was deleted. Similarly, each node in the B tree is also ....

[Article contains additional citation context not shown here]

P.J. Varman and R.M. Verma, An efficient multiversion access structure, IEEE Trans. on Knowledge and Data Engineering, 9:391--410, 1997.

....temporal index based join algorithms. Moreover, we introduce 3 two optimization techniques that further improve join performance. Although we concentrate on the Multiversion B tree (MVBT) BGO 96] our techniques and optimizations apply to other efficient tree based temporal indices, as well ([LS89, VV97]) 3. We present an extensive performance study that compares the proposed algorithms with (a) the unsynchronized MVBT join algorithm, b) the synchronized B tree algorithms, and, c) the synchronized R tree algorithms. Our performance results show that the MVBT approaches are consistently ....

P. Varman and R. Verma, "An Efficient Multiversion Access Structure", TKDE 9(3), pp. 391-409, 1997.

....scheme. For an object insertion the hashing scheme finds the integer associated with the position of this insertion and the B tree is updated. Deletions work similarly. Note that deleted integers can be reused. There have been various approaches to making a B tree partially persistent [1] 7] [13]. In our experiments we used the MVBT since its code was readily available to us. The MVBT [1] is a directed acyclic graph that embeds many B trees. It has a number of root nodes, where each root provides access to subsequent versions of the ephemeral B tree s evolution. Like all temporal ....

....Persistent List Various notions of partially persistent lists have appeared in the temporal database literature. Our discussion follows the approach outlined in [6] on how to make an ordered list partially persistent. In [5] a scheme to support non ordered partially persistent lists is presented. [13] presents the C list, which is a list structure made up of a collection of pages that contain versions of data records clustered by oid. However, a Clist solves a different query: given an oid and a version interval, find all versions of this oid during this interval . Let L be an ephemeral list ....

P.J. Varman and R.M. Verma, An Efficient Multiversion Access Structure, IEEE Trans. on Knowledge and Data Engineering, Vol.9, No. 3, pp: 391-409, 1997. 15

....m r b) node accesses, where m is the number of live intervals at the queried timestamp, and r is the number of output intervals. Both the space requirements and query performance are asymptotically optimal. A variation of MVB trees, which reduces the tree size by a constant factor can be found in [VV97] Several algorithms for processing interval queries and temporal joins with MVB trees, are proposed in [BS96] and [ZTS02] respectively. The multi version framework has also been applied to R trees to obtain various bitemporal and spatio temporal access methods [KTF98, KGT 01, TP01] General ....

Varman, P.J., Verma, R.M. An Efficient Multiversion Access Structure. IEEE TKDE, Vol. 9, No. 3, pp. 391-409, 1997.

....we introduce various optimization techniques that further improve join performance. Although we concentrate on the MVBT ( BGO 96] our techniques and optimizations apply to other efficient tree based temporal indices, as well (like the Time Split B tree [LS89] or the Multiversion Access Structure [VV97]) 3. We present an extensive performance study that compares the proposed algorithms with (a) the 3 unsynchronized MVBT join algorithm, b) the synchronized B tree algorithms, and, c) the synchronized R tree algorithms. Our performance results show that the MVBT approaches are consistently ....

P. Varman and R. Verma, "An Efficient Multiversion Access Structure", TKDE 9(3), pp. 391-409, 1997. 24

....B tree using the element DNNs. In order to answer SPaR range queries over a multiversion document (for example: find the elements in version V i with DNNs in range (x; y) we need to maintain the multiple versions of this B tree. Various multiversion B tree structures have been proposed [11, 1, 18]. Here we use the Multiversion B tree (MVBT) 1] which has optimal asymptotic behavior and its code was readily available. The MVBT has the form of a directed graph with multiple roots. Associated with each root is a consecutive version interval. A root provides access to the portion of the ....

P.J. Varman and R.M. Verma, An Efficient Multiversion Access Structure, IEEE Trans. on Knowledge and Data Engineering, Vol.9, No. 3, pp: 391-409, 1997.

....creates an exciting window of research opportunities. While there has been very little previous work on versioning web based semistructured documents [19, 21, 3] much relevant work was done in other applications like software configuration management [10] CAD systems[8] and temporal databases [1, 11, 14, 16, 17]. Some of the problems occurring in multiversion documents are similar to those of transaction time databases, where object histories are maintained (new objects are added without discarding the old ones) 7, 12] Timestamping provides an effective technique to deal with evolving objects in ....

....retrieved from in SCCS by scanning through the file and retrieving valid segments based on their timestamps. Both RCS and SCCS lack sophistication in their secondary storage management (compared to transaction time temporal databases that have developed efficient indexing and clustering techniques [9, 11, 14, 16, 17]) Moreover, neither approach support complex queries, or queries on the evolution of the structure of the document. In fact, most of today s software configuration tools still treat a document as a sequence of lines of text, thus ignoring the rich structure of the document. Another requirement ....

[Article contains additional citation context not shown here]

P.J. Varman and R.M. Verma, An Efficient Multiversion Access Structure, IEEE Trans. on Knowledge and Data Engineering, Vol.9, No. 3, pp: 391409, 1997.

....to frame f i . This will provide an efficient way to select objects from frame f i (in any case, we could not expect to do better than having an individual 2D R tree on that frame ) Methods to make an external memory data structure (in particular a B tree) partially persistent are given in [1, 7, 13]. In [6] a partially persistent R tree is presented to support bitemporal queries. We propose here to extent this methodology for indexing animated objects. Note that another approach to store the evolution of the 2D R tree is based on the notion of overlapping R trees [8] In the overlapping ....

....records alive in page x at f . Page x is considered dead after frame f . We can assume that the deletion frame field of all x s alive records is changed to f even though this is not needed in practice) Then the resulting new page has to be incorporated in the structure (for details we refer to [6, 13, 1]) Answering a spatiotemporal query about region S and frame f has two parts. First, the root alive at f is found. This part is conceptually equivalent to accessing the ephemeral R tree which indexes frame f . Second, the answer is found by searching this tree in a top down fashion as in a regular ....

P.J. Varman, R.M. Verma. An Efficient Multiversion Access Structure. In IEEE TKDE, 9(3): 391-409, 1997.

....current trajectories of the points. If the trajectory of a point changes, the future versions of the structure might also change. Now, instead of storing each K i explicitly, we store the differences between K i Gamma1 and K i , using the ideas of persistent data structures (see, for example, [12, 9, 34]) We call our data structure an interimpersistent data structure. Recall that K has three levels. The primary and secondary structures are variants of Btrees, and we can store the different version of them by adapting the standard techniques [12, 9, 34] Because of lack of space we will not ....

....data structures (see, for example, 12, 9, 34] We call our data structure an interimpersistent data structure. Recall that K has three levels. The primary and secondary structures are variants of Btrees, and we can store the different version of them by adapting the standard techniques [12, 9, 34]. Because of lack of space we will not discuss the modifications needed to tailor these techniques for our purposes. We will instead focus on the auxiliary structure A v stored at each node v of the secondary B tree, which answers a 3 sided range query on a set A of O(B 2 ) points. Recall that ....

[Article contains additional citation context not shown here]

P.J. Varman and R.M. Verma, An efficient multiversion access structure, IEEE Trans. on Knowledge and Data Engineering, 9:391--410, 1997.

....fits nicely with the degenerate case of the problem we address. This is because in the degenerate case an update simply corresponds to object additions deletions. Methods to make a disk based structure (in particular a B tree) partially persistent have appeared in the area of temporal databases [20, 5, 26, 46, 24]. 24] presents the Bitemporal R Tree which is a partially persistent R Tree used to index bitemporal objects. This partiallypersistent R Tree can be easily extended to index the degenerate case of animated objects. However the general case where objects change continuously is different. One ....

....records alive in node x at f . Node x is considered dead after frame f . We can assume that the deletion frame field of all x s alive records is changed to f even though this is not needed in practice) Then the resulting new node has to be incorporated in the structure (for details we refer to [24, 46, 5]) Answering a range query about region S and frame f has two parts. First, the root alive at f is found. This part is conceptually equivalent to accessing the root of ephemeral R Tree which indexes frame f . Second, the objects intersecting S are found by searching this tree in a top down ....

P.J. Varman, R.M. Verma. An Efficient Multiversion Access Structure. In IEEE TKDE, 9(3): 391-409, 1997.

....on the current trajectories of the points. If the trajectory of a point changes, the future versions of the structure might also change. Now, instead of storing each K i explicitly, we store the differences between K i Gamma1 and K i , using the ideas of persistent data structures (see e.g. [11, 8, 32]) We call our data structure an interim persistent data structure. Recall that K has three levels. The primary and secondary structures are variants of B trees, and we can store the different version of them by adapting the standard techniques [11, 8, 32] Because of lack of space we will not ....

....of persistent data structures (see e.g. 11, 8, 32] We call our data structure an interim persistent data structure. Recall that K has three levels. The primary and secondary structures are variants of B trees, and we can store the different version of them by adapting the standard techniques [11, 8, 32]. Because of lack of space we will not discuss the modifications needed to tailor these techniques for our purposes. We will instead focus on the auxiliary structure A v stored at each node v of the secondary B tree, which answers a 3 sided range query on a set A of O(B 2 ) points. Recall that ....

[Article contains additional citation context not shown here]

P.J. Varman and R.M. Verma, An Efficient Multiversion Access Structure, IEEE Trans. on Knowledge and Data Engineering, 9 (3), 1997.

....addresses the hashing problem in a temporal environment. Among existing temporal indexes, four are of special interest for this paper, namely: the Snapshot Index [TK95] the Time Split B tree (TSB) LS89] the Multiversion B Tree (MVBT) BGO 96] and the Multiversion Access Structure (MVAS) [VV97]. A simple model of temporal evolution follows. Assume that time is discrete described by the succession of non negative integers. Consider for simplicity an initially empty set S. As time proceeds, objects can be added to or deleted from this set. When an object is added to S and until (if ever) ....

P.J. Varman, R.M. Verma, "An Efficient Multiversion Access Structure", IEEE Trans. on Knowledge and Data Engineering, Vol 9, No 3, pp 391-409, 1997.

....updated. Multiversioning has been used extensively to provide concurrency control and recovery in (distributed) database systems ( Ree83, SR81, CFL 82, BS83, BG83, AS93, SA93, MWYC96, JMR97, LST97] Specialized multiversion access structures have also been proposed ( LS89, LS90, dBS96, BC97, VV97, MOPW98] In the context of OLTP systems, long read only queries can be allowed to access an older, fully consistent version of the database, while the update transactions are operating on a different version of the database ( BC92b, BC92a] This approach is known as transient versioning ....

Peter J. Varman and Rakesh M. Verma. "An Efficient Multiversion Access Structure". IEEE Transactions on Knowledge and Data Engineering, 9(3):391--409, May 1997.

No context found.

P. Varman and R. M. Verma, "An efficient multiversion access structure," IEEE Transactions on Knowledge and Data Engineering, vol. 9.

No context found.

P.J. Varman and R.M. Verma, An Efficient Multiversion Access Structure, IEEE Trans. on Knowledge and Data Engineering, Vol.9, No. 3, pp: 391-409, 1997.

No context found.

P.J. Varman, R.M. Verma. An Efficient Multiversion Access Structure. In IEEE TKDE, 9(3): 391-409, 1997.

No context found.

P. Varman and R. Verma, "An Efficient Multiversion Access Structure", IEEE TKDE 9(3), 1997.

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