Joseph A. Luchette,
Gregory K. Nelson,
Charles F. McLane III,
Liliana I. Cecan.
McLane Environmental, LLC.,
Princeton, NJ
Unlimited Virtual Computing Capacity
using the Cloud for Automated
Parameter Estimation
Why?
Where?
How?
Outline
Problem Description
Cloud Computing
Introduction
Setup and Use
Flexibility/Limitations/Reliability
Costs
Example Runs
Conclusions
Problem Description
Limited number of processors
Problem Description
Limited number of processors
Inadequate computing power
Problem Description
Limited times of availability
Problem Description
Limited times of availability
Cost of hardware
Cloud Computing
Introduction
Leveraging 3rd party computing capability
through the internet as a service.
Cloud Computing
Types of Cloud Services
Auto scaling applications.
Infrastructure-as-a-Service
Setup and Use
Flexibility/Limitations/Reliability
Costs
Infrastructure-as-a-Service
Setup and Use
Point and Click Interfaces or APIs
Setup and Use
Point and Click Interfaces or APIs
Root Access
Infrastructure-as-a-Service
Infrastructure-as-a-Service
Your Server
Setup and Use
Point and Click Interfaces or APIs
Root Access
Server “Images” – Save and Deploy
Infrastructure-as-a-Service
Setup and Use
Point and Click Interfaces or APIs
Root Access
Server “Images” – Save and Deploy
Networking
Network Interfaces (NICs)
Public/Private IP addresses
Infrastructure-as-a-Service
Flexibility/Limitations/Reliability
Operating System
• Emulation (Qemu)
• Server 2003/XP
• Server 2008/Vista
Infrastructure-as-a-Service
Flexibility/Limitations/Reliability
Operating System
CPU
Equivalencies – i.e. One EC2 ”Compute Unit”
provides the equivalent CPU capacity of a 1.0-1.2
GHz 2007 Opteron or 2007 Xeon processor.
Up to 20 units in a High-CPU
instance on EC2.
(8 virtual cores with 2.5
Compute units each)
Infrastructure-as-a-Service
Flexibility/Limitations/Reliability
Operating System
CPU
RAM & Disk
More RAM and Disk?
Infrastructure-as-a-Service
Flexibility/Limitations/Reliability
Operating System
CPU
RAM & Disk
Data transfer and SLA
Infrastructure-as-a-Service
Costs
Metered pricing
RAM or CPU capacity by the hour
$.08/hr to $1.20/hr
Data transfer in/out
$0.10 to $0.50 per GB
BandwidthRAM CPU
Infrastructure-as-a-Service
Costs
Metered pricing
RAM or CPU capacity by the hour
$.08/hr to $1.20/hr
Data transfer in/out
$0.10 to $0.50 per GB
Extras?
Operating system and software
BandwidthRAM CPU
Example Runs
Serial – Pumping Test
One virtual machine on GoGrid
Example Runs
Serial – Pumping Test
WTAQ model (USGS)
Example Runs
Serial – Pumping Test
WTAQ model (USGS)
Parameter Initial Estimated
Horizontal Hydraulic Conductivity [ft/day] 37 10 37
Vertical Hydraulic Conductivity [ft/day] 0.81 0.1 0.81
45.7 100 45.7
Specific Storage [ft-1] 0 0 0
Original
Model
Vertical Hydraulic Conductivity anisotropy
[Kh/Kv]
Example Runs
Serial – Pumping Test
10-5
10-4
10-3
10-2
10-1
100
101
Time
10-3
10
-2
10-1
100
10
1
102
103
Drawdown
10-5
10-4
10-3
10-2
10-1
100
101
Time
Initial Estimated
0 100 200 300 400 500
Initial Estimated
0
100
200
300
400
500
OriginalWTAQModel
Model Vs Initial
Real Vs Initial
45 degree line
0 4 8 12 16
Optimized Estimated
0
4
8
12
16
Model Vs Estimated
Real Vs Optimized
45 degree line
Example Runs
Parallel – Regional Model
4 virtual machines on GoGrid
Master Slaves
Example Runs
Parallel – Regional Model
MODFLOW 2000
Layer 1 Layer 2 Layer 3
Layer 4 Layer 5
Example Runs
Parallel – Regional Model
MODFLOW 2000
700
710
720
730
740
750
760
770
780
790
800
700 710 720 730 740 750 760 770 780 790 800
Modelled Water Level Elevation
OriginalModelWaterLevelElevation
700
710
720
730
740
750
760
770
780
790
800
700 710 720 730 740 750 760 770 780 790 800
Estimated Water Level Elevation
OriginalWaterLevelElevation
Original Model vs Estimate
45° Line
Conclusions
PEST parameter estimation can be performed
on a cloud infrastructure.
Conclusions
PEST parameter estimation can be performed
on a cloud infrastructure.
Eliminate the limiting factors of small-mid size
business settings for running computationally
intensive modeling.
Utilize scalable, low cost, on demand
computational resources
Questions or Comments?
Contact:
jluchette@mclaneenv.com
greg_k_nelson@yahoo.com
cmclane@mclaneenv.com
lcecan@mclaneenv.com

Unlimited Virtual Computing Capacity using the Cloud for Automated Parameter Estimation