Acceleration of XML Parsing through Prefetching

35
Acceleration of XML Parsing through Prefetching IEEE TRANSACTIONS ON COMPUTERS, VOL. 62, NO. 8, AUGUST 2013 Authors: Jie Tang, Student Member, IEEE, Shaoshan Liu, Chen Liu, Member, IEEE, Zhimin Gu, and Jean- Luc Gaudiot, Fellow, IEEE Presented By : Rohit Deshpande Mtech CSE 2nd sem 130913016 06/18/2022 1 Department Of Computer Science and Engg.

Transcript of Acceleration of XML Parsing through Prefetching

Page 1: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

1

Acceleration of XMLParsing through Prefetching

IEEE TRANSACTIONS ON COMPUTERS, VOL. 62, NO. 8, AUGUST 2013Authors: Jie Tang, Student Member, IEEE, Shaoshan Liu, Chen Liu, Member, IEEE, Zhimin Gu, and Jean-Luc Gaudiot, Fellow, IEEE

Presented By :Rohit Deshpande

Mtech CSE 2nd sem130913016

Page 2: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

2

Contents• Introduction• Motivation• The XML Parsing Process• Prefetching Techniques• Methodology• Performance Analysis of XML Parsing• Memory-Side Acceleration• Implementation Feasibility• A Exemplary Hardware Implementation• Conclusion• Future Work

Page 3: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

3

Introduction

• Extensible Markup Language (XML) has become a widely adopted standard for data representation and exchange.

• Neutrality, Application independency and Flexibility. • Morgan Stanley’s Financial Services system, spends

40% execution time on processing XML documents.• This paper proposes to accelerate XML parsing from

the memory side.

Page 4: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

4

Motivation

• XML is widely adopted standard for data representation and exchange.

• Parsing of XML data introduce significant overhead.

Page 5: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

5

The XML Parsing Process

Page 6: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

6

• XML data parsing consists of three steps: • Character conversion • Lexical analysis• Syntactic analysis.

Page 7: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

7

XML Parsing Modeling

Page 8: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

8

• Event Driven Parser

• Tree-Based Parser

Page 9: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

9

Prefetching Techniques

• Data prefetching has been proposed as a speculative technique to bridge the speed gap between CPU and memory subsystem.

• Classic Prefetching Techniques:-• Sequential Prefetching• Stride Prefetching• Strem Prefetching• Correlating Prefetching

Page 10: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

10

Overview of Hardware Prefetching• System Integration of Hardware Prefetcher

Page 11: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

11

Architecture of Hardware Prefetcher

Page 12: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

12

Methodology

• XML Parsers and Benchmarks

Page 13: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

13

• Performance and memory Modeling

Page 14: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

14

Native versus Managed Execution

Page 15: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

15

Performance Analysis Of XML Parsing

• Data Flow of XML Parsing

Page 16: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

16

• Disk Data Loading

Page 17: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

17

Data Loading from Memory Side

• Data flow starting from main memory, going through each cache layer and finally fetched in to CPU.

• The CPI of speed test is 0.80. • Using the SAX parser, the CPI of standard is

1.27, which introduces 58.75 % of overhead.• Using the DOM parser, the CPI of standard

becomes 1.42, which introduces 77.5% of overhead.

Page 18: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

18

Memory-Side Acceleration

• Prefetchers

Page 19: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

19

• Performance Analysis

Page 20: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

20

• Performance Improvement for SAX Parsing

Page 21: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

21

• Performance Improvement for DOM Parsing

Page 22: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

22

Implementation Feasibility

• Bandwidth Utilization

Page 23: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

23

• Hardware Cost and Energy Consumption

Page 24: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

24

Page 25: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

25

• Energy Consumption of SAX Parsing

Page 26: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

26

• Energy Consumption of DOM Parsing

Page 27: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

27

A Exemplary Hardware Implementation

• Power consumption of the prefetcher on FPGA

Page 28: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

28

• Hardware Cost and Power Consumption Comparison between eMIPS and Prefetcher

Page 29: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

29

Performance comparison of L2-only versus two-level prefetcher

Page 30: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

30

consumption comparison of L2-only versus two-level prefetcher

Page 31: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

31

Conclusion

• Authors Proposed to make acceleration for XML parsing from memory side by improving its data loading performance.

• Reduce cache misses by up to 80 percent, which translates into up to 20 percent of performance improvement.

• 12.77 percent of energy saving.

Page 32: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

32

Future Work

• The ultimate goal is to built a many core chip in which one core perform conventional general computing workloads and the other cores do XML parsing and Garbage Collection.

Page 33: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

33

References• K. Chiu, M. Govindaraju, and R. Bramley, “Investigating the Limits of Soap Performance for Scientific

Computing,” Proc. IEEE 11th Int’l Symp. High Performance Distributed Computing (HPDC-11), 2002. • M.R. Head, M. Govindaraju, R. van Engelen, and W. Zhang, “Grid Scheduling and Protocols-Benchmarking

xml Processors for Applications in Grid Web Services,” Proc. ACM/IEEE Conf. Supercomputing (SC ’06), p. 121, 2006.

• P. Apparao et al., “Architectural Characterization of an XML-Centric Commercial Server Workload,” Proc.

33rd Int’l Conf. Parallel Processing, 2004. • P. Apparao and M. Bhat, “A Detailed Look at the Characteristics of xml Parsing,” Proc. First Workshop

Building Block Engine Architectures for Computers and Networks (BEACON ’04), 2004. • M. Nicola and J. John, “XML Parsing: A Threat to Database Performance,” Proc. 12th Int’l Conf.

Information and Knowledge Management, 2003. • Int’l HapMap Project: http://hapmap.ncbi.nlm.nih.gov/, 2013. SAX Parsing Model: http://sax.sourceforge.net, 2013.

Page 34: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

34

• W3C, “Document Object Model (DOM) Level 2 Core Specifica-tion,” http://www.w3.org/TR/DOM-Level-2-Core, 2013.

• K. Chiu, T. Devadithya, W. Lu, and A. Slominski, “A Binary XML for Scientific Applications,” Proc. First Int’l Conf. e-

Science and Grid Computing, 2005. • XimpleWare, “VTD-XML: The Future of XML Processing,” http://vtdxml. sourceforge.net, Accessed 10, Mar. 2007. W. Lu, K. Chiu, and Y. Pan, “A Parallel Approach to XML Parsing,” Proc. IEEE/ACM Seventh Int’l Conf. Grid Computing,

Sept. 2006. • M.R. Head and M. Govindaraju, “Approaching a Parallelized XML Parser Optimized for Multi-Core Processor,” Proc.

Workshop Service-Oriented Computing Performance: Aspects, Issues, and Ap-proaches (SOCP ’07), June 2007. • R.D. Cameron, K.S. Herdy, and D. Lin, “High Performance XML Parsing Using Parallel Bit Stream Technology,” Proc.

Conf. Center for Advanced Studies on Collaborative Research, Oct. 2008. • L. Zhao and L. Bhuyan, “Performance Evaluation and Accelera-tion for XML Data Parsing,” Proc. Ninth Workshop

Computer Architecture Evaluation Using Commercial Workloads, 2006. • J. Moscola and J.W. Lockwood, “Reconfigurable Content-Based Router Using Hardware-Accelerated Language

Parser,” ACM Trans. Design Automation of Electronic Systems, vol. 13, article 28, 2008.

Page 35: Acceleration of XML  Parsing through Prefetching

04/13/2023 Department Of Computer Science and Engg.

35

THANK YOU