Thakkar, S.S. and M. Sweiger, "Performance of an OLTP Application on Symmetry Multiprocessor System," Proceedings of the 17th Annual International Symposium on Computer Architecture, Seattle, WA., May, 1990.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....first to study the performance of memory consistency models in the context of database workloads. There are a number of studies based on the performance of out of order processors for non database workloads(e.g. 42, 85, 89] Most previous studies of databases are based on in order processors [9, 27, 28, 34, 75, 94, 118, 119], and therefore do not address the benefits of more aggressive processor architectures. A number of the studies are limited to 102 uniprocessor systems [27, 34, 72, 75] As discussed in Section 4.4, data communication misses play a more dominant role in multiprocessor executions and somewhat ....

....75] As discussed in Section 4.4, data communication misses play a more dominant role in multiprocessor executions and somewhat reduce the relative effect of instruction stall times. Another important distinction among the database studies is whether they are based on monitoring existing systems [3, 13, 18, 26, 27, 28, 57, 62, 118] (typically through performance counters) or based on simulations [7, 8, 34, 72, 75, 82, 94, 106, 119, 120] one study uses a combination of both techniques [9] Monitoring studies have the advantage of using larger scale versions of the workloads and allowing for a larger number of experiments to ....

Shreekant S. Thakkar and Mark Sweiger. Performance of an OLTP Application on Symmetry Multiprocessor System. In Proceedings of the 17th Annual International Symposium on Computer Architecture, pages 228--238, June 1990.

....processing, decision support, microbenchmark, and performance evaluation. 1. INTRODUCTION In the last five to ten years, several studies have explored the architectural characteristics of online transaction processing (OLTP) database workloads [3] 7] 8] 9] 16] 17] 18] 19] 22] 23] 24] [26] [27] 1 This work was performed as part of the author s dissertation research. The author s present address is: Storage Systems Program, Hewlett Packard Laboratories, 1501 Page Mill Road, M S 1U 13, Palo Alto, CA 94304 1126. Her current email address is kkeeton hpl.hp.com. 50 [33] and decision ....

....approach for producing a representative random microbenchmark lies in posing multiple read only queries. 5. RELATED WORK Many of the studies that use database workloads to evaluate computer architecture innovations have employed the complex OLTP [3] 7] 8] 9] 16] 17] 18] 19] 22] 23] 24] [26] [27] 33] and DSS [3] 5] 15] 17] 18] 23] 31] workloads defined by the TPC. These studies vary in their usage of full scale data sets versus in memory data sets. One study provides rules of thumb for using an in memory version of the TPC B OLTP benchmark to approximate the processor and ....

S. S. Thakkar and M. Sweiger. "Performance of an OLTP application on Symmetry multiprocessor system." In Proc. of the 17th ISCA, June 1990, pp. 228-238.

....Multiprocessor Studies 3.8.2.1. Thakkar and Sweiger: Sequent Symmetry Thakkar and Sweiger used a hardware monitoring tool to study the bus utilization, cache coherence miss rates, and I O rates for the TP1 benchmark running on a Sequent Symmetry cache coherent symmetric multiprocessor (SMP) [100]. In addition, they explored the effects of larger caches, a faster clock rate, and process migration. For memory resident databases (e.g. fully cached ) they found that performance is limited by bus saturation caused by processes ping ponging between processors. This effect was improved by ....

....effects of larger caches, a faster clock rate, and process migration. For memory resident databases (e.g. fully cached ) they found that performance is limited by bus saturation caused by processes ping ponging between processors. This effect was improved by pro 64 Characteristics Thakkar90 [100] Cvetan96b [29] Barroso98 [13] Lovett96 [61] Ranganath98 [83] System evaluated 24 proc. Sequent Symmetry 4 proc. AlphaServer 4100 SMP 4 proc. AlphaServer 4100 SMP 32 proc. Sequent CCNUMA 4 proc. CC NUMA server Processor Alpha 21164 (in order) Alpha 21164 (in order) ....

S. S. Thakkar and M. Sweiger. "Performance of an OLTP application on Symmetry multiprocessor system," Proc. of the 17th ISCA, pages 228-238, June 1990.

....that can scale to thousands of nodes. To reduce network traffic and to minimize latency, each processor is given a large private cache. Measurements of shared memory multiprocessors running database workloads show that loading and flushing these caches considerably degrades processor performance [THAK90]. As parallelism increases, interference on shared resources limits performance. Multi processor systems often use an affinity scheduling mechanism to reduce this interference; giving each process an affinity to a particular processor. This is a form of data partitioning; it represents an ....

Thakkar, S.S. and M. Sweiger, "Performance of an OLTP Application on Symmetry Multiprocessor System," Proceedings of the 17th Annual International Symposium on Computer Architecture, Seattle, WA., May, 1990.

....has already been referenced in earlier sections. We further discuss some of the previous work pertinent to database workloads and CMP in this section. There have been a large number of recent studies of database applications (both OLTP and DSS) due to the increasing importance of these workloads [4,7,8,12,21,27,28,34,35,36,42,46]. To the best of our knowledge, this is the first paper that provides a detailed evaluation of database workloads in the context of chip multiprocessing. Ranganathan et al. 35] study user level traces of database workloads in the context of wide issue out of order processors, and show that the ....

S. S. Thakkar and M. Sweiger. Performance of an OLTP Application on Symmetry Multiprocessor System. In 17 th Annual International Symposium on Computer Architecture, pages 228--238, May 1990.

....For example, I O latencies can be hidden more effectively by using more server processes (as is typically done in audit runs of TPC benchmarks) Finally, a few studies have used OLTP and DSS workloads to evaluate multiprocessor systems or investigate other performance issues. Thakkar and Sweiger [18] identified disk I O and the system bus as bottlenecks for the TP1 benchmark on a Sequent Symmetry. Lovett and Clapp [11] focused on the L2 cache behavior and scalability of the Sequent STiNG CCNUMA multiprocessor. Verghese, et al. 24] evaluated page migration and replication for CC NUMA ....

S. Thakkar and M. Sweiger. Performance of an OLTP application on Symmetry multiprocessor system. In 17th Ann. Int'l Symp. on Computer Arch., p. 228--238, May 1990.

....cache coherency protocol, the Selective Redo scheme would be the best choice. 8 Related Work Many studies have demonstrated the performance advantages of using SM implementation platforms for database systems. Based on a TP1 benchmark performed on a Sequent Symmetry shared memory multiprocessor, [27] conclude that an SM database system can deliver very high performance. In [28] an analytical and simulation study compares SN (shared nothing) SD, and SIM (shared intermediate memory 7 ) This comparison concludes that the data sharing architectures, especially SIM, are more resilient to ....

S. Thakkar and M. Sweiger. Performance of an OLTP Application on Symmetry Multiprocessor System. Proceedings of the 17th International Symposium on Computer Architecture, pages 228--238, May 1990.

....when there are restrictions on how PEs are allocated, as in the Cray T3E which requires a contiguous set of nodes [346] or subcube allocation in hypercubes. The problem with this approach is that migration requires a potentially large context to be copied, which loads communication resources [571]. Also, this does not solve internal fragmentation. The only way to reduce internal fragmentation is to remove the restrictions on PE allocation [375, 606] A third solution is to reduce the effect of fragmentation by combining partitioning with time slicing. This works because scheduling is ....

S. S. Thakkar and M. Sweiger, "Performance of an OLTP application on Symmetry multiprocessor system". In 17th Ann. Intl. Symp. Computer Architecture Conf. Proc., pp. 228--238, May 1990.

....the processor s cache. This type of processor affinity is particularly interesting, because it may diminish with time, and because it is dependent on fairly complex characteristics of the workload. There are a significant number of references in the literature [19] 24] 26] 41] 43] 53] 54] [57] [60] 63] that deal with this type of processor affinity; this work is the subject of the first two sections of this chapter. In the remaining section, two new affinity scheduling algorithms are proposed. 3.1 Principles and Algorithms of Affinity Scheduling A parallel application executes most ....

....execution for any reason, it should be resumed on the same processor, avoiding migration whenever possible. The analytical model was used to demonstrate that ignoring cache affinity in multiprocessor scheduling can result in significant degradation of program performance. Thakkar and Sweiger [57] studied the performance of an Online Transaction Processing (OLTP) application on the Sequent Multiprocessor system using the TP1 [4] benchmark. Written in SQL, the benchmark simulates a banking system workload. The TP1 testing revealed a hierarchy of performance bottlenecks that show the ....

S. S. Thakkar, M. Sweiger, "Performance of an OLTP Application on Symmetry Multiprocessor System", Proceedings of the 17th Annual International Symposium on Computer Architecture, May 1990, pp. 228-238.

....and data clustering were evaluated [17] and found to improve join and aggregate queries. Each of these studies is targeted to a specific task and concentrate on ways to make it faster. The first hardware evaluation of a relational DBMS running an on line transaction processing (OLTP) workload [22] concentrated on multiprocessor system issues, such as assigning processes to different processors to avoid bandwidth bottlenecks. Contrasting scientific and commercial workloads [14] using TPC A and TPC C on another relational DBMS showed that commercial workloads exhibit large instruction ....

S. S. Thakkar and M. Sweiger. Performance of an OLTP Application on Symmetry Multiprocessor System. In Proceedings of the International Symposium on Computer Architecture, 1990.

....data partitioning, loop fusion, and data clustering were evaluated [17] and found to improve join and aggregate queries. All of the above studies are targeted to a specific task and concentrate on ways to make it faster. The first hardware evaluation of a relational DBMS running an OLTP workload [22] concentrated on multiprocessor system issues, such as assigning processes to different processors to avoid bandwidth bottlenecks. Contrasting scientific and commercial workloads [14] using TPC A and TPC C on another relational DBMS showed that commercial workloads exhibit large instruction ....

S. S. Thakkar and M. Sweiger. Performance of an OLTP Application on Symmetry Multiprocessor System. In Proceedings of the International Symposium on Computer Architecture, 1990.

....emitted. Unfortunately, obtaining the source code of a tuned database is hard because it is usually proprietary. For these reasons, there is little previous work in this area. In addition, most of the work has addressed the problem from a high level point of view. For example, Thakkar and Sweiger [12] looked at the performance of OLTP workloads running on a Sequent multiprocessor and highlighted the importance of process scheduling and the I O capability of the machine. Maynard et al. [5] contrasted the cache performance of technical and commercial workloads and concluded that the latter is ....

S. S. Thakkar and M. Sweiger. Performance of an OLTP Application on Symmetry Multiprocessor System. In Proceedings of the 17th Annual International Symposium on Computer Architecture, pages 228--238, May 1990.

....to name a few see [3, 24, 25] These studies are based on analytical and simulation models, and consider uniprocessor and SD database systems. In contrast, our study has considered implementation and performance issues in SM database systems, and has focused on empirical measurements. In [23], the performance of on line transaction processing applications are examined on a Sequent shared memory multiprocessor. This work examined the performance effects of the interaction between many system components, including the caching subsystem, process migration, and disk I O. However, in this ....

S. Thakkar and M. Sweiger. Performance of an OLTP Application on Symmetry Multiprocessor System. Proceedings of the 17th International Symposium on Computer Architecture, pages 228--238, May 1990.

.... along with academic research in this area, have been heavily influenced by popular scientific and engineering benchmarks such as SPLASH 2 [26] and STREAMS [13] with only a handful of published architectural studies that have in some way tried to address issues specific to commercial workloads [3, 7, 9, 12, 14, 16, 20, 21, 24]. The lack of architectural research on commercial applications is partly due to the fact that I O issues have been historically considered as the primary performance bottleneck for such workloads. However, innovations in disk subsystems (RAID arrays, use of non volatile memory) and software ....

....memory system performance issues and analyze their trends. One such trend is the increasing sensitivity of OLTP applications to dirty miss latency. Some previous studies have identified different bottlenecks for the performance of OLTP applications than the ones we underscore. Thakkar and Sweiger [20] have measured the performance of an early OLTP benchmark (TP1) in the Sequent Symmetry multiprocessor, using both memory resident and out of memory databases. Rosenblum et al. 16] focused on the impact of architectural trends on the operating system performance using three applications, ....

S. S. Thakkar and M. Sweiger. Performance of an OLTP application on Symmetry multiprocessor system. In Proceedings of the 17th Annual International Symposium on Computer Architecture, pages 228--238, June 1990.

....the first to study the performance of memory consistency models in the context of database workloads. There are a number of studies based on the performance of outof order processors for non database workloads(e.g. 8, 16, 18] Most previous studies of databases are based on in order processors [2, 4, 5, 6, 14, 20, 27, 28], and therefore do not address the benefits of more aggressive processor architectures. A number of the studies are limited to uniprocessor systems [4, 6, 13, 14] As discussed in Section 3, data communication misses play a more 3 The increase in the local and remote components of read latency ....

....such optimizations in their code if it leads to significant gains. dominant role in multiprocessor executions and somewhat reduce the relative effect of instruction stall times. Another important distinction among the database studies is whether they are based on monitoring existing systems [4, 5, 11, 27] (typically through performance counters) or based on simulations [6, 13, 14, 20, 21, 28] one study uses a combination of both techniques [2] Monitoring studies have the advantage of using larger scale versions of the workloads and allowing for a larger number of experiments to be run. However, ....

S. S. Thakkar and M. Sweiger. Performance of an OLTP application on Symmetry multiprocessor system. In Proceedings of the 17th Annual International Symposium on Computer Architecture, pages 228-- 238, June 1990.

....of the work has addressed the performance of these workloads from a high level point of view. For example, DeWitt and Gray [3] studied parallel database systems and indicated that a shared nothing architecture seems to be more cost effective than a shared memory architecture. Thakkar and Sweiger [9] looked at the performance of On Line Transaction Processing (OLTP) workloads running on a Sequent cache coherent shared memory multiprocessor and highlighted the importance of process scheduling and the I O capability of the machine. Maynard et al. [6] contrasted the cache performance of technical ....

S. S. Thakkar and M. Sweiger. Performance of an OLTP Application on Symmetry Multiprocessor System. In Proceedings of the 17th Annual International Symposium on Computer Architecture, pages 228--238, May 1990.

....of processors, our experiments used only 32 of the 64 processors magnitude better than CL. There has been considerable debate about whether a SM implementation platform is suitable for database systems. Based on a TP1 benchmark performed on a Sequent Symmetry shared memory multiprocessor, [21] concludes that a SM database system can deliver very high performance. In [22] an analytical and simulation study compares SN (shared nothing) SD, and SIM (shared intermediate memory 10 ) This comparison concludes that the data sharing architectures, especially SIM, are more resilient to ....

S. Thakkar and M. Sweiger. Performance of an OLTP Application on Symmetry Multiprocessor System. Proceedings of the Seventeenth Annual International Symposium on Computer Architecture, pages 228--238, May 1990.

....utilization or program execution. In order to reduce network traffic and to minimize latency, each processor is given a private cache. Measurements of sharedmemory multiprocessors running database workloads show that loading and flushing these caches considerably degrades processor performance [8]. As parallelism increases, interference on shared resources limits performance. Multiprocessor systems often use an affinity scheduling mechanism to reduce this interference, allowing favoritism by the processors. This is data partitioning, representing a step towards shared nothing design. ....

THAKKAR, S., SWEIGER, M. Performance of an OLTP Application on Symmetry Multiprocessor System. In Proceedings of the Seventeenth Annual International Symposium on Computer Architecture , (Seattle, May 1990)

....computed very fast by maintaining this table of page LSNs in shared memory. 7 Related Work Many studies have demonstrated the performance advantages of using SM implementation platforms for database systems. Based on a TP1 benchmark performed on a Sequent Symmetry shared memory multiprocessor, [25] conclude that an SM database system can deliver very high performance. In [26] an analytical and simulation study compares SN (shared nothing) SD, and SIM (shared intermediate memory 7 ) This comparison concludes that the data sharing architectures, especially SIM, are more resilient to ....

S. Thakkar and M. Sweiger. Performance of an OLTP Application on Symmetry Multiprocessor System. Proceedings of the 17th International Symposium on Computer Architecture, pages 228--238, May 1990.

....issues and workload choice can have a large impact on the measured performance of affinity scheduling, but the synthetic nature of the workloads means that the complex facets of real applications were missed. One study that did measure real applications on a real system, by Thakkar and Sweiger [14], looked only at the performance of a database system under an extreme form of affinity. The authors studied a database application running in a 12 to 24 processor machine. They found a significant amount of cache state lost due to cache migration, and simply attaching processes to processors ....

S. S. Thakkar and M. Sweiger. Performance of an OLTP Application on Symmetry Multiprocessor System. In Proceedings of the 17th Annual International Symposium on Computer Architecture, pages 228--238, May 1990.

No context found.

Thakkar, S.S. and M. Sweiger, "Performance of an OLTP Application on Symmetry Multiprocessor System," Proceedings of the 17th Annual International Symposium on Computer Architecture, Seattle, WA., May, 1990.

No context found.

S. S. Thakkar and M. Sweiger. Performance of an OLTP application on Symmetry multiprocessor system. In 17th Annual International Symposium on Computer Architecture, pages 228--238, May 1990.

No context found.

S. Thakkar and M. Sweiger. Performance of an OLTP Application on Symmetry Multiprocessor System. In Proceedings of the 18th International Symposium on Computer Architecture, pages 228--238, May 1990.

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