C. Baru et al. DB2 Parallel Edition. IBM Systems Journal, 34(2), 1995.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....more than 30 joins [4] over some Terabytes of data. In this context, the fast development of high performance parallel machines provided with efficient multi tasking opens the possibility to exploit massive parallelism for each complex query execution and higher throughput of concurrent execution [5]. In addition to the sequential optimization issues (i.e. operator ordering and implementation method choice) a parallel query optimizer must determine the degree of inter operator parallelism, i.e. the number of operators to be executed concurrently. The number of possible orderings faces a ....

....For every generated ordering a physical scheduling is done. We adapted a one phase optimization for our sharednothing implementation, because the risk for excluding a good parallizable operator ordering in the first phase is very high for the bushy tree space on shared nothing or hybrid systems [5, 22]. However, if we want to run our methods within a two phase optimization, we would apply the resource allocation only to the solution of the first phase. Assumptions In the remainder of the article we concentrate on equi join operators with join predicates of the form R:attr1 = S:attr2 for some ....

[Article contains additional citation context not shown here]

C.K. Baru, G. Fecteau, A. Goyal, H. Hsiao, A. Jhingran, S. Padmanabhan, G.P. Copeland, and W.G. Wilson. DB2 Parallel Edition. IBM Systems Journal, 34(2):292--323, 1995.

.... [1] Different research prototypes e.g. those followed by XPRS [2] GAMMA [3] DBS3 [4] Volcano [5] and commercial products as Teradata [6] Tandems NonStopSQL [7] Informix Online Xps [8] Sysbase SQL Navigation Server [9] Oracles Parallel Server V7 [10] and IBM s DB2 V4 [11] Parallel Edition [12] have been yet proposed. However many problems remain widely open. Among them static query optimization, its run time adaptation [13] and load balancing [14] are the most crucial problems. In this paper, we firstly introduce a static optimization architecture and present then a special hierarchy ....

C. K. Baru G. Fecteau A. Goyal H. Hsiao A. Jhingran S. Padmanabhan and G. P. Copeland. DB2 Parallel Edition. IBM Systems Journal, 34(2), 1995.

.... Enkidu Let us have a rst glance at commercial products. Oracle Parallel Server [1] seems now to reach its maturity. This product, while said to be working on workstations networks, is nevertheless working in a shared disk way, and has a mostly centralized control. IBM DB2 Parallel Edition [2] is, just like Oracle PE, a parallel port of a sequential product. DB2 o ers limited distribution techniques (mostly hashing) and is also supposed to work on big con guration, like SP2 or Sun SMPs. Last, Informix XPS [13] which is frankly shared to be workstations oriented , is in reality ....

C.K. Baru, G. Fecteau, A. Goya, H. Hsiao, Jhingran A., Padmanabhan S., G.P. Copeland, and Willson W.G. DB2 Parallel Edition. IBM Systems Journal, 34(2):292322, 1995.

....explicitly and provide the infrastructure to process them efficiently. Moreover, our approach allows insertions of new documents to run concurrently to retrieval without sacrificing the ACID properties. Research on parallel database systems has investigated hash partitioning for data placement [3, 7, 10, 26]. We take over hash partitioning techniques to distribute document and index data. 3 3 Architecture of the Document Engine To explain our architecture, Subsection 3.1 shows how multi level transactions increase parallelism. The following subsections then describe our search engine in more ....

C. Baru, G. Fecteau, A. Goyal, H. Hsiao, A. Jhingran, S. Padmanabhan, G. Copeland, and W. Wilson. Db2 parallel edition. IBM Systems Journal, 34(2):292--321, 1995.

....parallel # This work was supported in part by the National Science Foundation under contract No. NSF ASC9318183 and the Advanced Research Projects Agency under contract No. DABT63 94 C 0049. The authors assume all responsibility for the contents of the paper. server [13] DB2 parallel edition [2]) In such systems, database relations are generally partitioned horizontally and distributed across multiple processors. The other approach is to employ disk arrays [6, 21] or parallel file systems [9, 27] In both the approaches, the key motivation is to exploit parallelism (especially in I O) ....

C. K. Baru, G. Fecteau, A. Goyal, H. Hsiao, A. Jhingran, S. Padmanabhan, G. P. Copeland, and W. G. Wilson. DB2 parallel edition. IBM Systems Journal, 34(2):292--322, April 1995.

....(e.g. 21,22] These works have not yet consider operator orderings including inter operation parallelism because of its high scheduling complexity and its difficult synchronization. In the last years, several parallel DBMS products, as PARIS [19] Prisma [23] and the DB2 Parallel edition DBMS [24] have integrated inter operator parallelism into the query execution machine and schedule a member of the GBT space. Performance evaluations demonstrate that in the context of sufficient high resource availability (in terms of number of processors and memory size) inter operator parallelism ....

....when considering the structure of the rules right hand expression, however the last one can only be derived after some more complicated considerations. Deterministic search based query optimizer For parallel optimizers based on deterministic search (e.g. the dynamic programming in the IBM DB2 PE [24] and in the NonStopSql product [39] an extension of the searching technique for the best linear tree operator ordering is simple to achieve. Let us concentrate on the bottom up dynamic programming technique, as utilized in the NonStopSql product. It works iteratively over the number of relations ....

C.K. Baru, G. Fecteau, A. Goyal, H. Hsiao, A. Jhingran, S. Padmanabhan, G.P. Copeland, and W.G. Wilson. DB2 Parallel Edition. IBM Systems Journal, 34(2):292--323, 1995.

....effectiveness of parallelization in improving performance has been proved through the research [4, 7, 9, 2, 5, 16] Based (a) Currently at Mitsubishi Electric Corporation. on the results, several vendors are now shipping commercial parallel DBMS products running on commercial parallel platforms[1]. While research interests and efforts have concentrated on the parallelism exhibited in speedup and scaleup characteristics, the efficiency of fundamental I O operations tended to be ignored. Few papers present analyses on I O performance including disk transfer and network communication on real ....

C. Baru, G. Fecteau, et al., DB2 parallel edition, IBM Systems Journal, Vol.34, No.2, 1995.

....introduce communication operators. To our knowledge, only three systems have adopted this latter approach : the par LERA database language [18] with its constructor node (see section 6. 1) the Volcano prototype [19] exchange operator) and the commercial product DB2 Parallel Edition of IBM [20](data exchange operator) However, all these operators carry very poor information which strongly limits their representation power : no communication algorithm can be specified ; no means are provided to specify the access methods used to read the relation partitions to be communicated. In fact, ....

C.K. Baru G. Fecteau A. Goyal H. Hsiao A. Jhingran S. Padmanabhan G.P. Copeland W.G. Wilson. DB2 Parallel Edition. IBM Systems Journal, 34(2), 1995.

....require reorganization of the data to re balance the load of the system. The shared nothing architecture has been adopted by many commercial database systems such as Tandem, Teradata (one of the earliest and most successful commercial database machine) Informix XPS, and BD2 Parallel Edition [5] as well by numerous research prototypes including Gamma [22] and Bubba[8] 4.2.4 Hierarchical Hybrid Architecture The hierarchical or hybrid architecture represents a combination of the shared memory, shared disk and shared nothing architectures [81] The main vehicle of this architecture is an ....

C. Baru, G. Fecteau, A. Goyal, H. Hsiao, A. Jhingran, S. Padmannabhan, G. Copeland, and W. Wilson. DB2 Parallel Edition. IBM Systems Journal, 34(2):292--322, 1995.

....Large Data Sets using DB2 Parallel Edition Sriram Padmanabhan IBM T.J. Watson Research Center srp watson.ibm.com Commercial parallel database systems such as DB2 Parallel Edition (DB2 PE) [1, 2] are delivering the ability to execute complex queries on very large databases. However, the serial application interface to these database systems can become a bottleneck for a growing list of applications such as mailing list generation and data propagation from a warehouse to smaller data ....

Baru, C. K., et al. DB2 Parallel Edition. IBM Systems Journal 34, 2 (1995), 292--322.

....University research groups are divided between these two schools. DBS3 [16] and Gamma belong to the first category, whereas Volcano [17] Midas [18] and XPRS [19] integrate parallelism within an existing DBMS. Concerning industrial vendors, we mainly find the second approach, like IBM products [20], Oracle [21] and Informix [22] This is not astonishing as the software can be more quickly developed when already running codes could be reused. However the amount of work needed to achieve parallel functionnalities remains important. For example, the implementation of Volcano took about five ....

C.K. Baru et al. DB2 Parallel Edition. IBM Systems Journal, 34(2), 1995.

....number of relation pages are moved from heavily loaded nodes or disks to more lightly loaded nodes [2, 19] The algorithms used simple predefined strategies, and they did not factor in the cost of movement into their decision. A few reorganization utilities have been described in the literature [18, 14, 15, 1, 12]. Each utility provides a fixed reorganization and strategy, such as index creation. All of these utilities would be possible candidates for a decision algorithm to consider. Therefore, a method should be created to automatically choose a utility and compare different utilities. To compare ....

C. K. Baru et al. DB2 Parallel Edition. IBM Systems Journal, pages 292-- 322, February 1995.

....research groups are divided between these two schools. DBS3 [VAL91] and Gamma belong to the first category, whereas Volcano [GRA93] Midas [BOZ96] and XPRS [HON92] integrate parallelism within an existing DBMS. Concerning industrial vendors, we mainly find the second approach, like IBM products [BAR95], Oracle [LIN93] and Informix [GER95] This is not astonishing as the software can be more quickly developed when already running codes can be reused. However the amount of work needed to achieve parallel functionalities remains important. For example, the implementation of Volcano took about ....

C. K. Baru, G. Fecteau, A. Goyal, H. Hsiaoand, A. Jhingran, S. Padmanabhan, G. Copeland and W.G. Wilson. DB2 Parallel Edition. IBM Systems Journal, vol.34, 1995.

....effort has focused on developing high performance database management systems. One approach is to build multiprocessor database machines, which have become increasingly popular (for example, Bubba [4] Gamma [12] Teradata [5, 13] Tandem [25] Oracle parallel server [14] DB2 parallel edition [2]) In such systems, database relations are generally partitioned horizontally and distributed across multiple processors. Another approach is to employ disk arrays [6, 22] or parallel file systems [10, 28] In both approaches, the key motivation is to exploit parallelism (especially in I O) by ....

Chaitanya K. Baru et al. DB2 parallel edition. IBM Systems Journal, 34(2):292--322, April 1995.

....parallel This work was supported in part by the National Science Foundation under contract No. NSF ASC9318183 and the Advanced Research Projects Agency under contract No. DABT63 94 C 0049. The authors assume all responsibility for the contents of the paper. server [13] DB2 parallel edition [2]) In such systems, database relations are generally partitioned horizontally and distributed across multiple processors. The other approach is to employ disk arrays [6, 21] or parallel file systems [9, 27] In both the approaches, the key motivation is to exploit parallelism (especially in I O) ....

C. K. Baru, G. Fecteau, A. Goyal, H. Hsiao, A. Jhingran, S. Padmanabhan, G. P. Copeland, and W. G. Wilson. DB2 parallel edition. IBM Systems Journal, 34(2):292--322, April 1995.

....memory [Stonebraker, 1986] This categorisation was used to facilitate discussion about appropriate parallel hardware architectures for DBMS. Many researchers participated in the ensuing discussion; see e.g. Hua et al. 1991] DeWitt and Gray, 1992] Bergsten et al. 1993] Valduriez, 1993b] [Baru et al. 1995]; many others base their arguments on it. We summarise their conclusions and briefly describe the architectural categories, namely shared memory (SM) shareddisk (SD) and shared nothing (SN) The following basic arguments and characteristics are often given for these architectures. These are not ....

....do not scale up well: Bhide and Stonebraker, 1988] showed that beyond a certain number of processors, access to main memory can become a bottleneck that limits the system s processing speed. Thus SM systems are limited to a small number ( 20 30) of processors [Valduriez, 1993a] cf. [Baru et al. 1995]. Shared Disk In a Shared disk system, each processor has its private memory, but access to disks is shared by all processors; see Figure 2. It is argued that the costs for SD system are relatively low as the interconnect could be a bus system based on standard technology. It is also argued that ....

[Article contains additional citation context not shown here]

Baru, C., Fecteau, G., Goyal, A., Hsiao, H., Jhingran, A., Padmanabhan, S., Copeland, G., and Wilson, W. (1995). DB2 Parallel Edition. IBM Systems Journal, 34(2):292--322.

....parallel database and parallel application models. Parallel database model: In order to simplify our discussion, we assume that the underlying database architecture is a Shared Nothing (SN) architecture [18] based on a Massively Parallel Processing (MPP) system. In the SN parallel database model [3, 4, 6], each table is partitioned and declustered across a number of nodes. Figure 2 shows a typical database application. From the DBMS perspective, the application starts by connecting to a database which results in the creation of a database coordinator (dbcoord) The application can issue queries ....

....database system must provide in order to realize the various interfaces. The discussion includes both the usage of existing parallel database technology and proposed extensions and enhancements. These discussions are mostly based on the IBM DB2 Parallel Edition (DB2 PE) parallel database system [3, 8]. Implementing SCI does not require any database system support because the database application interface does not change. However, the application structure must change so that the appcoord retrieves the data from the database and distributes it to the appslaves in a partitioned manner. In the ....

C. K. Baru et al. DB2 Parallel Edition. IBM Systems Journal, 34(2):292--322, 1995.

....with disks attached to the nodes Table 1. 3: Some representative Shared Nothing parallel databases As stated in [49] although all three configurations have been built in the past, a parallel database architecture based on the shared nothing hardware design has emerged as the most favoured approach [12, 27, 48]. It is noteworthy that the underlying technological and economic reasons for this convergence are exactly the same as the ones discussed in section 1.1. The advantages claimed for the SN architecture include: 1) partitioning allows multiple processors to scan large relations in parallel without ....

C. K. Baru, G. Fecteau, A. Goyal, H. Hsiao, A. Jhingran, S. Padmanabhan, G. P. Copeland, and W. G. Wilson. DB2 Parallel Edition. IBM Systems Journal, 34(2):292--322, 1995.

No context found.

C. Baru et al. DB2 Parallel Edition. IBM Systems Journal, 34(2), 1995.

No context found.

Baru, W., C., & al. DB2 Parallel Edition. IBM Syst. Journal, 34, 2, 1995. 292-322.

No context found.

Baru, W., C., & al. DB2 Parallel Edition. IBM Syst. Journal, 34, 2, 1995. 292-322.

No context found.

Baru, W., C., & al. DB2 Parallel Edition. IBM Syst. Journal, 34, 2, 1995. 292-322.

No context found.

Baru, W., C., & al. DB2 Parallel Edition. IBM Syst. Journal, 34, 2, 1995. 292-322.

No context found.

C.K. Baru, G. Fecteau, A. Goyal, H. Hsiao, A. Jhingran, S. Padmanabhan, G.P. Copeland, and W.G. Wilson. DB2 Parallel Edition. IBM Systems Journal, 34(2), 1995.

No context found.

C. K. Baru, G. Fecteau, A. Goyal, H. Hsiao, A. Jhingran, S. Padmanabhan, G. P. Copeland, and W. G. Wilson. DB2 Parallel Edition. IBM Systems Journal, 34(2):292--322., 1995.

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