Windows Cluster

1
High Performance Spreadsheet Simulation on a
Desktop Grid
Supasit Kajkamhaeng Juta Pichitlamken Putchong
Uthayopas
Faculty of Engineering, Kasetsart University,
Bangkok, Thailand
2
Agenda

• Introduction
• Problem
• Motivation
• Objective
• Methodology
• Result
• Conclusion

3
Introduction

• Simulation 1
• A sampling experiment that is done on the
  computer
• The core of simulation is a model that involves
  quantities whose values are unpredictable
• The model is represented using a computer program
• The program actually samples the random
  variables, performs the computations of the model
  and reports the outcome usually in the form of
  one or more numerical values

4
Introduction (2)

• Spreadsheet
• A widely used computer application because of its
  versatility and ease of use in calculation or
  modeling a problem
• The most popular spreadsheet program is Microsoft
  Excel on Windows
• Excel provides a very easy-to-use system, which
  any user can utilize to perform complex analysis
  on data without having to learn programming

5
Introduction (3)

• Spreadsheet simulation 1
• The use of a spreadsheet as a platform for
  representing simulation models and performing
  simulation experiments

Picture from http//www.palisade.com/images3/produ
ct/risk/RunSimulation.jpg
6
Problem

• When a problem size of the spreadsheet simulation
  is large, computation power of a single computer
  might not be enough

7
Motivation

• Solve the problem by using a parallel/distributed
  computing technique
• Provide a user-friendly, yet computationally
  powerful simulation environment for end users
• Design spreadsheet simulation infrastructure
  which is inexpensive and easy to maintain by any
  company

8
Objective

• Build a Distributed Spreadsheet Simulation
  Infrastructure (DSSI) that is easy to use and
  computationally fast by utilizing power of
  parallel computing on a Windows-based desktop
  grid.

Picture from http//www.grid.phys.uvic.ca/assets/
sun_figures/Grid-2.jpg
9
Methodology

• Parallel Algorithm of Spreadsheet Simulation
• How to do parallel computing of spreadsheet
  simulation ?
• How to hide the complexity of parallelism from
  users ?
• Design Distributed Spreadsheet Simulation
  Infrastructure (DSSI)

10
Methodology (2)

• Spreadsheet Simulation Model

Decision Variable
Random Input
Output
Model
11
Methodology (3)
Mix and Location isnt fixed
12
Methodology (4)

• Parallel Algorithm of Spreadsheet Simulation
• Provide a well-defined computation template

13
Methodology (5)
SimRun Worksheet
Model Worksheet
14
Methodology (6)
15
Methodology (7)

• Distributed Spreadsheet Simulation Infrastructure
  (DSSI)

16
Methodology (8)
17
Result

• Value-at-Risk Test Problem 5
• A risk metric that probabilistically describe
  market risks

SimRun Worksheet
Model Worksheet
18
Result (2)
Output Worksheet
19
Result (3)

• Experimental Setups

20
Result (4)

• Runtime

21
Result (5)

• Speedup

22
Result (6)

• Efficiency

23
Conclusion

• Propose an architecture that allows spreadsheet
  simulation to use Windows-based desktop grids to
  accelerate the execution speed
• The approach is almost all computations on Excel
  spreadsheets Thus, no complex programming is
  needed
• Propose that the complexity of parallel computing
  can be mostly hidden from users by providing
  well-designed computation templates in Excel

24
Acknowledgement

• This research is supported by the Thailand
  Research Fund Grant Number MRG4980121

25
Reference

• Andrew F. Seila., Spreadsheet Simulation., In
  Proceedings of the 2006 Winter Simulation
  Conference, Monterey, CA, USA
• Abramson, D., J. Dongarra, E. Meek, P. Roe, and
  Z. Shi. 2004., Simplified grid computing through
  spreadsheets and NetSolve., In Proceedings of the
  Seventh International Conference on High
  Performance Computing and Grid in Asia Pacific
  Region (HPCAsia04)
• Abramson, D., P. Roe, L. Kotler, and D. Mather.
  2001., ActiveSheets Super-computing with
  spreadsheets. In Proceedings of the High
  Performance Computing Symposium (HPC01),
  Advanced Simulation Technologies Conference,
  110115. San Diego, California

26
Reference

• Nadiminti, K., Y.F. Chiu, N. Teoh, A. Luter, S.
  Venugopal, and R. Buyya., 2004., ExcelGrid A
  .NET plug-in for outsourcing Excel spreadsheet
  workload to enterprise and global grids., In
  Proceedings of the 12th International Conference
  on Advanced Computing and Communication (ADCOM
  2004)
• Jorion, P. 2001., Value at risk The new
  benchmark for managing financial risk., 2nd ed.
  McGraw-Hill Trade

27
Question and Suggestion
Thank you