P. Stenstrom, E. Hagersten, D. J. Lilja, M. Martonosi and M Venugopal, "Trends in Shared Memory Multiprocessing, " IEEE Computer, pp. 44--50, December 1997.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... MMX [93] MVI [19] VIS [123] At the high end server market, database and information processing applications such as online transaction processing (OLTP) 121] and decision support systems (DSS) 122] have emerged as the largest and fastest growing market segment for shared memory servers [115]. While database workloads were once exclusively the domain of main frame systems, the cost advantages of using commodity components with shared memory servers have led to a greater use of these general purpose systems for database workloads. Several of these media processing and database ....

Per Stenstrom, Erik Hagersten, David J. Lilja, Margaret Martonosi, and Madan Venugopal. Trends in Shared Memory Multiprocessing. IEEE Computer, 1997.

....214 11.3. Appendix: Bottleneck Analysis for System Performance . 215 viii List of Figures CHAPTER 1. 1 Figure 1 1. Dataquest server market breakdown [94]. 1 Figure 1 2. TPC C price performance over time [103] 11 Figure 1 3. TPC D (100 GB) price performance over time [104] 12 Figure 1 4. TPC D (300 GB) price performance over time ....

.... commercial server applications, such as database service, file service, media and email service, print service, and custom applications, were the dominant applications run on shared memory multiprocessor server machines in 1995 and are projected to be the dominant server applications in 2000 [94]. As shown in Figure 1 1, commercial applications comprised about 85 of the 1995 server market, and are projected to continue this dominance as the server market grows 15 percent annually. Database workloads alone motivate the sale of vast quantities of symmetric multiprocessor (SMP) machines, ....

[Article contains additional citation context not shown here]

P. Stenstrom, E. Hagersten, D. J. Lilja, M. Martonosi, and M. Venugopal. "Trends in shared memory multiprocessing," IEEE Computer, pages 44-50, December 1997.

....a transient periodic task is created, the overload will remain until it terminates. For such situations, we believe the negotiation scheme to be more appropriate than the other schemes. 2. 2 Parallelism issues Shared memory multiprocessors has demonstrated its potential to yield high performance [8, 9]. In high performance computing, one usually tries to solve a single problem as fast as possible. Because of the communication overhead, induced by executing parallelized programs, the running time consumed to solve a problem will not always decrease as the degree of parallelization increases. In ....

P. Stenstrom et al. Trends in shared-memory multiprocessing. IEEE Computer, 30(12):44--50, December 1997.

....also exhibit temporal locality but the size of its working set critically depends on the structure of the query, the method of scanning, and the size and the content of the database. 1. INTRODUCTION Decision support systems (DSS) are one of the key drivers of the high performance server market [17] taking the form of uniprocessor or parallel servers ranging from shared nothing to shared everything systems [5] While I O latency is an important performance obstacle in these systems, the increasing speed gap between processor and memory technology has also made good cache performance ....

P. Stenstrom, E. Hagersten, D. Lilja, M. Martonosi, and M. Venugopal. Trends in Shared Memory Multiprocessing. IEEE Computer, 30(12):44--50, December 1997.

.... such as online transaction processing (OLTP) and decision support (DSS) database service, file service, media and email service, print service, and custom applications, were the dominant applications run on server machines in 1995 and are projected to be the dominant server applications in 2000 [26]. Commercial applications comprised about 85 of the 1995 server market, and are projected continue this dominance as the server market grows 15 percent annually. Database workloads alone motivate the sale of vast quantities of symmetric multiprocessing machines, and hold the dominant fraction of ....

....annually. Database workloads alone motivate the sale of vast quantities of symmetric multiprocessing machines, and hold the dominant fraction of the massively parallel computing market [19] databases motivated 32 of the server volume in 1995, and will motivate 39 of the 2000 server volume [26]. Despite the widespread usage of commercial applications, they are often ignored in preference to technical benchmarks, such as SPEC or LINPACK, in computer architecture performance studies. This bias is due largely to the lack of available representative multi user traces of commercial ....

P. Stenstrom, et al. "Trends in shared memory multiprocessing. " IEEE Computer, pages 44-50, December, 1997.

.... s performance capability. A consequence of this viewpoint is that much of the computer architecture research to date has focused on improving the performance of computer systems for these types of applications, even though they represent only about 15 of the overall market for large systems [1]. An implicit assumption in this previous research has been that all types of computer systems would benefit from the architectural innovations developed to improve the performance of this subset of applications. However, nonnumeric application programs, specifically, those that do not operate on ....

....to improve performance. B. Parallelization Challenges Numerous compilation techniques and parallel architectures have been developed that have been used to successfully parallelize regularly structured scientific and numerical application programs, such as those commonly written in Fortran [1]. However, parallelizing less regular applications written in languages such as C, C , and Java has proven to be much more difficult due to several characteristics that occur frequently in programs written in these languages. In particular, these languages encourage the use of pointers, which are ....

Per Stenstrom, Erik Hagersten, David J. Lilja, Margaret Martonosi, , and Madan Venogupal, "Trends in shared-memory multiprocessing," in IEEE Computer, December 1997, vol. 30, pp. 44--50.

....not only on absolute performance but also on the ease with which the multiprocessor may be programmed. Parallel computer designers have responded by providing support for the more popular programming models. For example, much of the impetus behind the design of distributed shared memory (or NUMA) [11,15] architectures stems from a desire to provide programmers with a shared memory programming model for a distributed memory architecture. Shared memory programming models are generally considered to be simpler than other alternatives such as message passing. Given the importance placed on designing ....

....fall between message passing and HPF and yet are not far removed from the hardware model on which they run. While the use of this programming paradigm has traditionally been restricted to SMPs with a limited number of processors, the advent of commercial distributed shared multiprocessors [15] allows a shared memory style of programming to be used on larger systems. 15 7. Acknowledgments Support for this project was provided in part by Computing Devices International, Inc. National Science Foundation grant no. MIP 9221900, and a University of Minnesota McKnight Land Grant ....

Stenstrom, P., E. Hagerston, D.J. Lilja, M. Martonosi and M. Venogupal, "Trends in Shared-Memory Multiprocessing," IEEE Computer, v30, n12, 1997, pp. 44-50.

No context found.

P. Stenstrom, E. Hagersten, D. J. Lilja, M. Martonosi and M Venugopal, "Trends in Shared Memory Multiprocessing, " IEEE Computer, pp. 44--50, December 1997.

No context found.

P. Stenstrom, et al. "Trends in shared memory multiprocessing. " IEEE Computer, pages 44-50, December, 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