Measuring parallel processor performance

Karp, Alan H.; Flatt, H.P.

doi:10.1145/78607.78614

Cited by 206 publications

(113 citation statements)

References 2 publications

Supporting

Mentioning

109

Contrasting

Unclassified

Order By: Relevance

“…Second, using serial fraction allows us to link our results with other existing similar values for deterministic algorithms like the ones explained in [13,14] and in many interesting references of the CALMA project documents [21]. Finally, as mentioned above, with f m we can identify more subtle effects (trends) in the parallel program than when only using speedup.…”

Section: Speedup and Related Performance Measuresmentioning

confidence: 75%

“…Finally, Karp and Flatt [13] have devised an interesting metric for measuring the performance of any parallel algorithm that can help us to identify much more subtle effects than using speedup alone. They call it the serial fraction of the algorithm (f m ):…”

Section: Speedup and Related Performance Measuresmentioning

confidence: 99%

“…The target island incorporates it into its population if it is better than its presently worst string. The reception of incoming strings is asynchronous for efficiency [2,13].…”

Section: Examples Of Super-linear Performance With Parallel Gasmentioning

confidence: 99%

See 2 more Smart Citations

Parallel evolutionary algorithms can achieve super-linear performance

Alba

2002

Information Processing Letters

145

View full text Add to dashboard Cite

One of the main reasons for using parallel evolutionary algorithms (PEAs) is to obtain efficient algorithms with an execution time much lower than that of their sequential counterparts in order, e.g., to tackle more complex problems. This naturally leads to measuring the speedup of the PEA. PEAs have sometimes been reported to provide super-linear performances for different problems, parameterizations, and machines. Super-linear speedup means that using "m" processors leads to an algorithm that runs more than "m" times faster than the sequential version. However, reporting super-linear speedup is controversial, especially for the "traditional" research community, since some non-orthodox practices could be thought of being the cause for this result. Therefore, we begin by offering a taxonomy for speedup, in order to clarify what is being measured. Also, we analyze the sources for such a scenario in this paper. Finally, we study an assorted set of results. Our conclusion is that super-linear performance is possible for PEAs, theoretically and in practice, both in homogeneous and in heterogeneous parallel machines. 

show abstract

Section: Speedup and Related Performance Measuresmentioning

confidence: 75%

Section: Speedup and Related Performance Measuresmentioning

confidence: 99%

See 1 more Smart Citation

Parallel evolutionary algorithms can achieve super-linear performance

Alba

2002

Information Processing Letters

145

View full text Add to dashboard Cite

show abstract

“…Given a parallel algorithm like the complete exchange algorithm, it is important to define a feasible mapping of the algorithm steps to the logical configuration. The feasibility of a mapping is defined by the extent to which it supports the performance quality attributes of speedup with respect to serial computing and efficiency [7]. Speedup S p is defined by the following formula:…”

Section: Fig 1 Pseudo Code For Complete Exchange Algorithmmentioning

confidence: 99%

Model-Driven Approach for Supporting the Mapping of Parallel Algorithms to Parallel Computing Platforms

Arkin

Tekinerdoğan

İmre

2013

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. The trend from single processor to parallel computer architectures has increased the importance of parallel computing. To support parallel computing it is important to map parallel algorithms to a computing platform that consists of multiple parallel processing nodes. In general different alternative mappings can be defined that perform differently with respect to the quality requirements for power consumption, efficiency and memory usage. The mapping process can be carried out manually for platforms with a limited number of processing nodes. However, for exascale computing in which hundreds of thousands of processing nodes are applied, the mapping process soon becomes intractable. To assist the parallel computing engineer we provide a model-driven approach to analyze, model, and select feasible mappings. We describe the developed toolset that implements the corresponding approach together with the required metamodels and model transformations. We illustrate our approach for the well-known complete exchange algorithm in parallel computing.

show abstract

“…Using the Karp-Flatt [18] metric we can estimate the serial fraction e of every tested system assuming that we know the level of parallelism p (which is equal to N ) and the speedup, ψ, achieved when the maximum level of parallelism is utilized. …”

Section: Scaling Up With Column-storesmentioning

confidence: 99%

Scaling up analytical queries with column-stores

Alagiannis

Athanassoulis

Ailamaki

2013

Proceedings of the Sixth International Workshop on Testing Database Systems

View full text Add to dashboard Cite

As data analytics is used by an increasing number of applications, data analytics engines are required to execute workloads with increased concurrency, i.e., an increasing number of clients submitting queries. Data management systems designed for data analytics -a market dominated by column-stores -however, were initially optimized for single query execution, minimizing its response time. Hence, they do not treat concurrency as a first class citizen.In this paper, we experiment with one open-source and two commercial column-stores using the TPC-H and SSB benchmarks in a setup with an increasing number of concurrent clients submitting queries, focusing on whether the tested systems can scale up in a single node instance. The tested systems for in-memory workloads scale up, to some degree; however, when the server is saturated they fail to fully exploit the available parallelism. Further, we highlight the unpredictable response times for high concurrency.

show abstract

Measuring parallel processor performance

Cited by 206 publications

References 2 publications

Parallel evolutionary algorithms can achieve super-linear performance

Parallel evolutionary algorithms can achieve super-linear performance

Model-Driven Approach for Supporting the Mapping of Parallel Algorithms to Parallel Computing Platforms

Scaling up analytical queries with column-stores

Contact Info

Product

Resources

About