M. Stonebraker, R. Katz, D. Patterson, J. Ousterhout, The design of XPRS, in: Proceedings of the 14th VLDB Conference, Los Angeles, CA, 1988, pp. 318--330.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....in Sect. 8. Its main difference from our approach is that it does have some optimization overhead at run time, and that it uses dynamic programming instead of randomized algorithms. The XPRS project proposes to select at run time a parallel plan from a set of plans based on buffer allocations [SKPO88]. Two different optimization algorithms have been proposed for this task. In an earlier reference [SKPO88] a binary search approach is advocated, where a query is first optimized for the smallest (m) and the largest (M ) possible buffer size; if the two obtained plans are far from optimal for ....

....time, and that it uses dynamic programming instead of randomized algorithms. The XPRS project proposes to select at run time a parallel plan from a set of plans based on buffer allocations [SKPO88] Two different optimization algorithms have been proposed for this task. In an earlier reference [SKPO88], a binary search approach is advocated, where a query is first optimized for the smallest (m) and the largest (M ) possible buffer size; if the two obtained plans are far from optimal for the buffer size for which they were not chosen, the query is optimized again for the midpoint between m and ....

Stonebraker M, Katz R, Patterson D, Ousterhout J (1988) The design of xprs. In: Proc. of the 14th International VLDB Conference, Long Beach, Calif., pp 318--330,

....and Full Parallel. Based on the experiments clear guidelines are given when to use which strategy. 1 Introduction For years now, research has been done on the design, implementation, and performance of parallel DBMSs. Teradata [CaK92] Bubba [BAC90] HC186 16 [BrG89] GAMMA [DGS90] and XPRS [SKP88] are examples of systems that actually were implemented, and many papers were written on their performance. The performance evaluation of these systems is mainly limited to simple queries that involve no more than one or two join operations. Recent developments in the direction of support of ....

M. Stonebraker, R. Katz, D. Patterson & J. Ousterhout, "The design of XPRS," in Proc 14th VLDB Conf, Los Angeles, CA, August 29-September 1, 1988, 318 330.

....HP s mapping of the Image Database into the operating system 20 virtual address space, and Mach s mapped file [TEVA87] mechanism. This approach simplified the implementation of the upper levels of the Bubba software. 3.6. Other Systems Other parallel database system prototypes include XPRS [STON88], Volcano [GRAE90] Arbre [LORI89] and the PERSIST project under development at IBM Research Labs in Hawthorne and Almaden. While both Volcano and XPRS are implemented on sharedmemory multi processors, XPRS is unique in its exploitation of the availability of massive shared memory in its design. ....

Stonebraker, M., R. Katz, D. Patterson, and J. Ousterhout, "The Design of XPRS", Proceedings of the Fourteenth International Conference on Very Large Data Bases, Los Angeles, CA, August, 1988.

....pipelining, bushy trees, hash joins. 1 Introduction In recent years, parallel database machines have drawn a considerable amount of attention from both the academic and industrial communities due to their high potential for parallel execution of complex database operations [1] 3] 8] 11] 17] [32]. In relational database systems, joins are the most expensive operations to execute, especially with the increases in database size and query complexity [23] 38] As a result, it has become imperative to develop solutions for efficient execution of multi join queries for future database ....

M. Stonebraker, R. Katz, D. Patterson, and J. Ousterhout. The Design of XPRS. Proceedings of the 14th International Conference on Very Large Data Bases, pages 318--330, 1988.

....and other optimization choices. Other global dependencies are the following: the order of tuples in intermediate results when sorting algorithms are used [26] the site of intermediate relations in distributed systems [8] and ressource contentions [11] In shared memory systems such as XPRS [10], the way a relation is hashed, or more generally fragmented, is much less important than in 14 Probe2 Join( R1,R2,R3) R1.A = R2.B) and (R2.B = R3.C) Build2 Select R3 R2 R1 Select Select Temp Build1 Probe1 Figure 5: A Query Execution Plan. distributed memory systems since it does not ....

M. Stonebraker et al. The design of xprs. In Proc. Int. Conf. on Very Large Databases, 1988.

.... [SGM86, PGK88] and declustering [RE78, LKB87, GD90] are two popular techniques employed by both general purpose multi disk I O subsystems (e.g. RAID [PGK88] and parallel database management systems (e.g. Gamma [DGS 90] Non Stop SQL [Gro88] DBC 1012 [Ter85] Bubba [BAC 90] XPRS [SPO88] Staggered striping extends these concepts to multimedia information systems, and is novel because its design enables either a multi disk or a parallel system to guarantee a continuous retrieval of an object at the bandwidth required to support its display. There are several proposals which ....

M. Stonebraker, D. Patterson, and J. Ousterhout. The design of XPRS. In proceedings of the International Conference on Very Large Databases, 1988.

....trees, hash joins. 1 Introduction In recent years, multiprocessor based parallel database machines have drawn a considerable amount of attention from both the academic and industrial communities due to their high potential for parallel execution of complex database operations [2] 7] 22] [36] [38] In relational database systems, joins are the most expensive operations to execute, especially with the increases in database size and query complexity [21] 27] 43] Several applications which involve decision support and complex objects, such as graphics, artificial intelligence and ....

M. Stonebraker, R. Katz, D. Patterson, and J. Ousterhout. The Design of XPRS. Proceedings of the 14th International Conference on Very Large Data Bases, pages 318--330, 1988.

....Low level details of how to store reliably the code of compensating transactions, and record their identity in the log records of the saga s subtransactions are also discussed there. A noteworthy approach, which can be classified as a simple type of compensation, is employed in the XPRS system [SKPO88] There, a notion of failure commutativity is defined for complete transactions. Two transactions failure commute if they commute, and if they can both succeed then a unilateral abort by either transaction cannot cause the other to abort. Transactions that are classified as failure commutative ....

M. R. Stonebraker, R. H. Katz, D. A. Patterson, and J. K. Ousterhout. The design of XPRS. In Proceedings of the Fourteenth International Conference on Very Large Databases, Los Angeles, pages 318--330, 1988.

....between plans with different memory requirements, both for individual operator evaluation and for buffer space between successive producer consumer pairs of pipelined operators. This distinction has been discussed in [11] but no notation to distinguish between them is proposed. In the XPRS system [6] the main focus is on intra operator parallelism, and a comprehensive notation for plans is not discussed. Recent work [13] has evaluated the technique of parallelizing a good sequential plan. 5] depicts the parallelism in an example query plan using a time resource diagram, without however a ....

....query plans are far from validation [3] Search Algorithm: Determining the optimal sequential plan is a hard combinatorial problem. Since for every sequential query plan a number of parallel ones are possible, the search space for parallel query optimization is much larger. The XPRS optimizer [6, 13] reduces the search space by finding a good sequential plan and parallelizing it. The bottom up search approach proposed in [14] is a promising one, whose details of how the cost estimate is used to guide the search have not been reported. 9] proposes the use of dynamic programming as the search ....

M. R. Stonebraker et al. The design of xprs. In Proc. of the 14th Very Large Database Conference, August 1988.

....(subtransaction) T from its compensation log records, where the action of T s compensating transaction is recorded. The can precede, cover, and invert relationships between transactions are based on the semantics of transactions, and they can be captured in a similar way to commutativity[LMWF94, Wei88, Kor83, SKPO88], and recoverability[BK92] In order to capture these relationships, the pro le (or code) and input arguments of each transaction must be available. In the Saga model, several possible solutions to the problem of saving code reliably are proposed[GMS87] therefore, these relationships can be ....

....compensating transaction for T 1 will cancel the order from the supplier as well. Although a variety of types of compensation are possible, all of them require semantic knowledge of the application. The notion of commutativity, either of operations [LMWF94, Wei88, Kor83] or of transactions [SKPO88], has been well exploited to enhance concurrency in semantics driven concurrency control. There are several types of commutativity. In operation level, for example, two operations O 1 and O 2 commute forward [Wei88] if for any state s in which O 1 and O 2 are both de ned, O 2 (O 1 (s) O 1 (O 2 ....

[Article contains additional citation context not shown here]

M. Stonebraker, R. Katz, D. Patterson, and J. Ousterhout. The design of XPRS. In Proceedings of the International Conference on Very Large Databases, pages 318-330, Los Angeles, CA, 1988.

....processing, bushy trees, sort merge joins. 1 Introduction In recent years, parallel database machines have drawn a considerable amount of attention from both the academic and industrial communities due to their high potential for parallel execution of complex database operations [4] 8] 16] [27] [30] In relational database systems, joins are the most expensive operations to execute, especially with the increases in database size and query complexity [15] 21] 35] Database applications which involve decision support and complex objects usually have to specify their desired results in ....

M. Stonebraker, R. Katz, D. Patterson, and J. Ousterhout. The Design of XPRS. Proceedings of the 14th International Conference on Very Large Data Bases, pages 318--330, 1988.

....performance. As we have already noted, the IBM 3990 disk controller [8] is capable of buffering updates in its non volatile memory. Both volatile and non volatile buffers are maintained, and updates appear in both. Dirty data in the safe buffer is staged to disk asynchronously when possible. XPRS [18], a POSTGRES based transaction processing system, buffers frequently updated data blocks in safe RAM to provide fast recovery. The techniques used to manage the safe buffer are not described there. The Phoenix file system [5] maintains an entire file system in safe RAM. In [3] the use of ....

....as REDO logging and checkpointing) to guarantee the update safety, the need for the additional mechanism is eliminated. 6 6 A log may be used for other purposes as well, such as to record the commit decisions for each transaction. However, such a log can be much more compact than a REDO log. [18]. 26 9.00 10.00 11.00 12.00 13.00 14.00 15.00 1 9 3 7 5 5 7 3 LRU Volatile WritePurge LRU Volatile WriteUpdate LRU Global WriteUpdate LRU Global WritePurge Read Response Time (ms. Safe Buffer Size Volatile Buffer Size TRACE: serverA 6.00 8.00 10.00 ....

Stonebraker, M., et. al. "The Design of XPRS," Proceedings of the 14th VLDB Conference, Los Angeles, California, 1988, pp. 318-330. 33

....work related to each of these components and the innovations and extensions we propose. 2.1 Architecture Several parallel database architectures have been proposed and built. These include both SN systems (e.g. BUBBA and GAMMA) BAC 90, DGS 90] and SE systems (e.g. XPRS and DBS3) SKPO88, BCV91] SN systems are typically built by interconnecting nodes of processor disk pairs. Data is split into disjoint sets and assigned to different nodes in the system. Then the system can be viewed as a set of interconnected but independent database systems, each managing its own subset of the ....

M. Stonebraker, R. Katz, D. Patterson, and J. Ousterhout. The design of XPRS. In Proceedings of the 14th VLDB Conference, pages 318--330, 1988.

....and Data Engineering, Vol. 6 No. 3, June 1996. 1 Introduction There has been a growing interest in applying general purpose parallel machines to database applications [7, 8, 12, 19, 34, 43, 50] Several research systems have been developed to explore this trend, including GAMMA [16] XPRS [49], DBS3 [4] GRACE [31] and BUBBA [6] Relational databases have a certain natural affinity to parallelism. Relational operations are set oriented and this provides the query optimizer lots of flexibility in selecting the parallelizable access path. In relational database systems, joins are the ....

M. Stonebraker, R. Katz, D. Patterson, and J. Ousterhout. The Design of XPRS. Proceedings of the 14th International Conference on Very Large Data Bases, pages 318--330, 1988.

....in a query increases. 1 Introduction As parallelism is recognized as a powerful and costeffective means to execute complex database operations, it has become imperative to develop efficient solution procedures to handle multi join queries in parallel database systems [2] 7] 8] 9] 11] 13] [24]. A query plan is usually compiled into a tree of operators, called a join sequence tree, where a leaf node represents an input relation and an internal node represents the resulting relation from joining the two relations associated with its two child nodes. There are three categories of query ....

M. Stonebraker, R. Katz, D. Patterson, and J. Ousterhout. The Design of XPRS. Proceedings of the 14th International Conference on Very Large Data Bases, pages 318--330, 1988.

....queries, and some complex queries for such applications may take hours or even days to complete, thus degrading the system performance. As a result, parallelism has been recognized as the solution for the efficient execution of multi join queries for future database management [2] 6] 7] 13] [22]. Intra operator parallelism, which occurs when several processors work in parallel on a single two way join operation, was the focus of most prior studies on exploiting parallelism for database operations [1] 5] 10] 14] 16] 18] 20] In addition, inter operator parallelism allows that ....

M. Stonebraker, R. Katz, D. Patterson, and J. Ousterhout. The design of XPRS. Proceedings of the 14th Int'l Conf. on Very Large Data Bases, pages 318--330, 1988.

....of queries. We present lower bounds for these combinatorial problems and demonstrate how processors may be optimally assigned by tiling the partitioned data grid with optimal configurations. 1 Introduction In highly parallel database machines (e.g. Gamma [2] Bubba [1] Non Stop SQL [11] XPRS [10] and Volcano [6] relations are partitioned (see Livny et al. [8] and Ries and Epstein [9] for early partitioning strategies) across multiple processors. This allows each processor to execute a portion of a query in parallel with the other processors, resulting in a lower response time for the ....

M. Stonebraker, D. Patterson, and J. Ousterhout, The design of XPRS, in Proceedings of the 1988 Very Large Data Bases Conference, Los Angeles, CA, September 1988.

No context found.

Stonebraker, M., Patterson, D., Katz, R. and Ousterhout, J., `The Design of XPRS', Proc. of 14th VLDB Conf., Long Beach, CA, August 1988.

....by making use of information about available main memory and the load average per processor. Both components are described in this paper. 1. INTRODUCTION This paper explores parallelism in the software data base machine XPRS. An initial paper presenting the outline of the system appeared in [STON88] and one on the reliable disk system in [PATT88] This paper explores the file system that we are constructing and the novel way we will obtain intra query parallelism in the resulting system. Recent systems that exploited intra query parallelism include Gamma [DEWI86, DEWI88] Bubba [COPE88] ....

Stonebraker, M. et. al., "The Design of XPRS," Proc. 1988 VLDB Conference, Los Angeles, Ca., Sept. 1988.

No context found.

M. Stonebraker, R. Katz, D. Patterson, J. Ousterhout, The design of XPRS, in: Proceedings of the 14th VLDB Conference, Los Angeles, CA, 1988, pp. 318--330.

No context found.

Stonebraker, M., R. Katz, D. Patterson, and J. Ousterhout, "The Design of XPRS", Proceedings of the Fourteenth International Conference on Very Large Data Bases, Los Angeles, CA, August, 1988. 32

No context found.

M. Stonebraker, R. Katz, D. Patterson, J. Ousterhout, The design of XPRS, in: Proceedings of the 14th VLDB Conference, Los Angeles, CA, 1988, pp. 318--330.

No context found.

M. Stonebraker, D. Patterson, and J. Ousterhout. The design of XPRS. In Proceedings of the 1988 VLDB Conference, Los Angeles, CA, September 1988.

No context found.

M. Stonebraker, R. Katz, D. Patterson, and J. Ousterhout. The Design of XPRS. Proceedings of the 14th International Conference on Very Large Data Bases, pages 318--330, 1988. 44

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