Applying Performance Models to Understand Data-Intensive Computing Efficiency
Published 1 May 2010
Elie Krevat, Tomer Shiran, Eric Anderson, Joseph Tucek, Jay J. Wylie, Gregory R. Ganger
Citations14
Generate an AI Snapshot to get a quick, structured summary of this paper.
Study Snapshot
ObjectiveStudy objective
MethodsResearch methodology
PopulationPopulation studied
Sample sizeSample sizes
OutcomesStudy outcomes here
ResultsStudy results comes here
LimitationsResearch study limitations comes here
A concise AI-generated summary of the paper will appear here once you click Generate AI Snapshot.
TL;DR
A simple analytical model is described that predicts the optimal performance of a parallel dataflow system and achieves performance close to the analytic optimal, showing that the model is realistic and that large improvements in the efficiency of parallel analytics are possible.
Abstract
Abstract : New programming frameworks for scale-out parallel analysis, such as MapReduce and Hadoop, have become a cornerstone for exploiting large datasets. However, there has been little analysis of how these systems perform relative to the capabilities of the hardware on which they run. This paper describes a simple analytical model that predicts the optimal performance of a parallel dataflow system. The model exposes the inefficiency of popular scale-out systems, which take 3-13x longer to complete jobs than the hardware should allow, even in well-tuned systems used to achieve record-breaking benchmark results. To validate the sanity of our model, we present small-scale experiments with Hadoop and a simplified dataflow processing tool called Parallel DataSeries. Parallel DataSeries achieves performance close to the analytic optimal, showing that the model is realistic and that large improvements in the efficiency of parallel analytics are possible.
Keywords
Computer Science
Communications of the ACMMapReduce
18,538 Citations2008Jay B. Dean, Sanjay Ghemawat
This presentation explains how the underlying runtime system automatically parallelizes the computation across large-scale clusters of machines, handles machine failures, and schedules inter-machine communication to make efficient use of the network and disks.
ACM SIGOPS Operating Systems ReviewThe Google file system
5,003 Citations2003Sanjay Ghemawat, Howard Gobioff +1 more
Dryad
2,460 Citations2007Michael Isard, Mihai Budiu +3 more
The Dryad execution engine handles all the difficult problems of creating a large distributed, concurrent application: scheduling the use of computers and their CPUs, recovering from communication or computer failures, and transporting data between vertices.
ACM SIGOPS Operating Systems ReviewCUBIC
2,222 Citations2008Sangtae Ha, Injong Rhee +1 more
The CUBIC protocol modifies the linear window growth function of existing TCP standards to be a cubic function in order to improve the scalability of TCP over fast and long distance networks.
Communications of the ACMParallel database systems
1,281 Citations1992David J. DeWitt, Jim Gray
Over the last decade 'Eradata, Tandem, and a host of startup companies have successfully developed and marketed highly parallel machines that refutes a 1983 paper predicting the demise of database machines.
Evaluating MapReduce for Multi-core and Multiprocessor Systems
968 Citations2007Colby Ranger, R. Raghuraman +3 more
It is established that, given a careful implementation, MapReduce is a promising model for scalable performance on shared-memory systems with simple parallel code.
ComputerAn introduction to disk drive modeling
829 Citations1994Chris Ruemmler, John Wilkes
A calibrated, high-quality disk drive model is demonstrated in which the overall error factor is 14 times smaller than that of a simple first-order model, which enables an informed trade-off between effort and accuracy.
Communications of the ACMMapReduce and parallel DBMSs
427 Citations2009Michael Stonebraker, Daniel J. Abadi +5 more
MapReduce complements DBMSs since databases are not designed for extract-transform-load tasks, a MapReduce specialty.
Flux: an adaptive partitioning operator for continuous query systems
317 Citations2004Mehul A. Shah, Joseph M. Hellerstein +2 more
A dataflow operator called flux is introduced that encapsulates adaptive state partitioning and dataflow routing that can be used for CQ operators under shifting processing and memory loads and can provide several factors improvement in throughput and orders of magnitude improvement in average latency over the static case.
Argon: performance insulation for shared storage servers
208 Citations2007Matthew Wachs, Michael Abd-El-Malek +2 more
Experiments show that, combined, Argon's automaticallyconfigured prefetch/write-back sizes and Argon's automatic configuration of each achieve the insulation goal.
A simulation approach to evaluating design decisions in MapReduce setups
206 Citations2009Guanying Wang, Ali R. Butt +2 more
The resulting simulator, MRPerf, captures such aspects of MapReduce setups as node, rack and network configurations, disk parameters and performance, data layout and application I/O characteristics, among others, and uses this information to predict expected application performance and can serve as a tool for optimizing existing MapReduces setups as well as designing new ones.
Cluster I/O with River
175 Citations1999Remzi H. Arpaci-Dusseau, Eric Anderson +5 more
This work introduces River, a data-flow programming environment and I/O substrate for clusters of computers based on two simple design features: a high-performance distributed queue, and a storage redundancy mechanism called graduated declustering.
FigshareData-Intensive Supercomputing: The Case for DISC
163 Citations2018Randal E. Bryant
By engaging the academic research community in DISC issues, the community can more systematically and in a more open forum explore fundamental aspects of a societally important style of computing.
Track-aligned extents: matching access patterns to disk drive characteristics
143 Citations2002Jiří Schindler, John Linwood Griffin +2 more
For large-file workloads, a version of FreeBSD’s FFS implementation that exploits traxtents reduces application run times by up to 20% compared to the original version and 44% lower overall write cost for track-sized segments can be achieved.
Tashi
56 Citations2009Michael A. Kozuch, Michael Ryan +8 more
It is demonstrated that location-aware applications can outperform those that are not location aware by factors of 3-11 and two general services developed for Tashi are described to provide location-awareness independently of the storage system.
RAPID: Shrinking the Congestion-Control Timescale
48 Citations2009Vishnu Konda, Jasleen Kaur
A new approach for congestion-control is designed that allows TCP connections to boldly search for, and adapt to, the available bandwidth within a single RTT, and the rate-based design allows RAPID to be truly RTT-fair.
ACM SIGOPS Operating Systems ReviewDataSeries
31 Citations2009Eric Anderson, Martin Arlitt +2 more
This paper introduces DataSeries, an on-disk format, run-time library and set of tools for storing and analyzing structured serial data, and identifies six key properties of a system to store and analyze this type of data.