De96 DeWitt, D.: Parallel Object-Relational Database Systems: Challenges & Opportunities, invited lecture at the 4th Int. Conf. on Parallel and Distributed Information Systems, 1996.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... vendors agree that Object Relational DBMSs (ORDBMSs) are the future [24] Though full support of parallel ORDBMS is expected, the techniques used to parallelize relational database systems are not adequate for ORDBMSs due to the complex abstract data types (ADTs) and user defined functions (UDFs) [1]. On the other hand, non traditional database application domains such as time series data analysis, image processing and geoscience information systems often require data objects in order. As a result, a novel mechanism that supports parallelization of user defined operations on both ordered and ....

....satisfying different UDF input requirements. 3.3 Windows Definition 3. 2 Given a data stream ds of type T , a substream of size l on ds is a function ss : integer T such that there exists some integer k 0 and for all 1 i l, ss[i] ds[k i] ss can be represented as ds[u; v] such that ss[1] = ds[u] and ss[l] ds[v] The size of a substream is at least 1, i.e. sizeOf(ss) u Gamma v 1 1. The set of all substreams of data stream ds of type T is denoted as sub(T ) 2 Informally, a window (with respect to the semantics of a UDF) is an input substream of bounded size such that the ....

[Article contains additional citation context not shown here]

D. DeWitt. Parallel object-relational database systems: Challenges and opportunities. In Invited Talk, PDIS, 1996.

....is considered. On the other hand, we considered the dynamically re optimizing QEP configurations in a parallel execution environment. The reason is that users are going to expect full parallel support for all user defined extensions in an extensible database environment, e.g. ORDBMS [5]. Parallelization is important for performance improvement to long running queries. As a consequence, more reconfiguration forms are considered in our research. Using a similar reconfiguration technique as in [16] materializing intermediate results) 25] proposed a two phase (plan rescheduling ....

D. DeWitt. Parallel Object-Relational Database Systems: Challenges and Opportunities. In Invited Talk, PDIS, 1996.

.... that Object Relational DBMSs (ORDBMSs) are the future [23] Though full support of parallel ORDBMS is expected, the techniques used to parallelize relational database systems are not adequate to parallelize the operations in ORDBMSs due to the introduction of complex abstract data types (ADTs) [1]. User defined functions (UDFs) that are associated with ADTs are usually opaque to a query optimizer and allow little optimization opportunities. On the other hand, in non traditional database application domains such as time series data analysis, image processing and geoscience information ....

....UDF input requirements. 3.2 Windows Definition 3. 2 Given a data stream ds of type T , a substream of size l on data stream ds is a function ss : integer T such that there exists some integer k 0 and for all 1 i l, ss[i] ds[k i] ss can also be represented as ds[u; v] such that ss[1] = ds[u] and ss[l] ds[v] The size of a substream is at least 1, i.e. sizeOf(ss) u Gamma v 1 1. The set of all substreams of data stream ds of type T , fds[1; 1] ds[1; 2] Delta Delta Delta ; ds[2; 2] ds[2; 3] Delta Delta Deltag, is denoted as sub(T ) 2 Informally, a window ....

[Article contains additional citation context not shown here]

D. DeWitt. Parallel object-relational database systems: Challenges and opportunities. In Invited Talk, PDIS, 1996.

.... have recently led to significant development efforts for parallel ORDBMS (PORDBMS) of some database vendors ( 7] 13] 12] 28] Although first industrial implementations enter the marketplace and the SQL3 standard [10] is maturing, there are still many topics left for research in this area ([8], 9] 5] 35] One of the current goals for PORDBMS is to move towards a framework for constructing parallel ADTs[35] and more sophisticated query optimization and execution ( 35] 24] ADT functions are completely opaque for the query optimizer and thus allow only very restricted query ....

....join predicates. We are currently working on a framework for userdefined join algorithms to overcome cartesian products and allow efficient parallel execution. How can parallelism be used to efficiently process UDFs on single but very large ADTs [13] and collection types in parallel ([8], 35] We have not addressed these important questions here. Additional future work should be concerned with the impact of user defined aggregate functions and the proposed sorting requirements on query optimization. Another interesting question is how materialized user defined aggregates can be ....

DeWitt, D.: Parallel Object-Relational Database Systems: Challenges & Opportunities, invited talk, PDIS 1996.

....they only reconfigure a QEP at points where no operation is partially done, i.e. no pipelined parallelism, or data flow structure, is considered in their work. We considered the dynamic reconfiguration of QEPs in a parallel execution environment, which is expected for ADT extensions in ORDBMSs [3]. Thus more reconfiguration options are considered in our research. Further, not only the built in operators but also UDFs are considered in our research. Lastly, intermediate results are materialized for plan modification in their work. This may not be practical in the case of the large volume ....

D. DeWitt. Parallel object-relational database systems: Challenges and opportunities. In Invited Talk, PDIS, 1996.

....words, no pipelined parallelism, or data flow structure, is considered in their work. On the other hand, we considered the dynamic reconfiguration of QEPs in a parallel execution environment. The reason is that users are going to expect full parallel support for all ADT extensions in an ORDBMS [3]. Parallelization is important for performance improvement to long running queries. As a consequence, more reconfiguration options are considered in our research. Further, we consider reconfiguration involving not only the built in operators but also user defined functions in an ORDBMS. Lastly, to ....

D. DeWitt. Parallel object-relational database systems: Challenges and opportunities. In Invited Talk, PDIS, 1996.

No context found.

De96 DeWitt, D.: Parallel Object-Relational Database Systems: Challenges & Opportunities, invited lecture at the 4th Int. Conf. on Parallel and Distributed Information Systems, 1996.

No context found.

DeWitt, D.: Parallel Object-Relational Database Systems: Challenges & Opportunities, invited talk, PDIS 1996.

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