Grids Clouds Computation Steven Armentrout

Grids, Clouds and
Computation:
Getting to Ground Truth under Mostly Cloudy
Conditions

Presented by:
Steven Armentrout, PhD
President & CEO

© 2008 Parabon Computation, Inc. All rights reserved. | 1

History: Parabon Computation

Founded in 1999
Launched the first COTS grid platform –
Frontier®
Operate the only brokered computation service
Customers in the government, academic,
non-profit, and private sectors
Sponsor of Compute Against Cancer®
Privately-held small business based in Reston,
VA

Overview

Motivation and History
How Parabon does ―cloud‖
Applications
Considerations


Is Cloud Computing Different than Grid
Computing?


Computing?

No


Cloud Computing’s Pedigree
(Syllables - Letters)

Distributed (4 - 11)
Utility (3 - 7)
Internet (3 - 8)
Peer-to-Peer (3 - 3) “P2P”
Cluster (2 - 7)
Grid (1 - 4)
Cloud (1 - 5)
Net (1 - 3)<- Prediction


Computing?
Distinguish Requirements and Implementation
Key User Requirement:
Resources delivered as on-demand utility
services
 Applications
 Information
 Computation
Implementation:
Driven by other requirements
 Security
 Economics
 Technical constraints © 2008 Parabon Computation, Inc. All rights reserved. | 7

Computation as a Utility

“If computers of the kind I have
advocated become the computers of
the future, then computing may
someday be organized
as a public utility just as the telephone
system is a public utility... The computer
utility could become the basis of a new
and important industry.”
—John McCarthy, MIT Centennial in 1961

The Big Switch

Recall the world before the Internet
Imagine the world before electricity
―The Big Switch,‖ by Nicholas Carr
(Jan 2008) draws strong parallels
between the maturation of the
electricity and computing industries

The implications for IT are profound


The Big Switch

IT is racing toward a new paradigm:
Computing delivered as an on-demand
service
The transformation is like that of the electric
industry during the 20th century
It’s the biggest IT transformation since
the introduction of the World Wide Web


Within the DoD, why now?
Consumer market successes Awareness
Unprecedented capability Potential
Technical advances Feasibility
 Virtualization
 Security
Economic pressures Necessity
Adoption by adversaries Competition
―Retirement of the Boomers‖ Openness to
change


Frontier Online Service



1. Security


4. Scalability


5. Usability

Frontier Enterprise


Frontier Enterprise

6. Unobtrusiveness


Frontier Enterprise

7. Ease of Admin.


Considerations: Locality

1. Remote or Local
 Security
 Cost
 Data


Locksmith Decryption Service
NIJ An online decryption service
NATIONAL
INSTITUTE
to serve law enforcement
of JUSTICE developed with AccessData®

DNA Frontier

Qu i ckTi me ™ an d a
TIFF (Un co mp re s se d) de co mp re s so r
a re ne ed ed to se e thi s p i ctu re .

Only file headers need to be transmitted
so Locksmith is suitable for a public grid.


Considerations: User Interface

Who are the end users?
 Unix wizards
 Support staff
 Warfighter


Frontier Dashboard


Watchman™

QuickTime™ and a
TIFF (LZW) decompressor
are neede d to see this picture.


Watchman™

QuickTime™ and a


Considerations: Types of Applications

What are the computational demands
of the applications?

Web apps load = transactions / s
Grid apps load = calculations / s
Data apps load = throughput / s


NASA Origin
The Ultimate Source
from Which to Evolve TM

Automated design of spacecraft antenna

Non-branching: Branching:
ST5-4W-03 ST5-3-10


NASA Origin
The Ultimate Source
from Which to Evolve TM

Automated design of spacecraft antenna
Evolved antenna flew
on three nanosats (20‖)
for ST5 mission


InSēquio™

Synthetic DNA can be used as a nanoscale
construction material and woven into designs

QuickTime™ and a

See CAD View of DNA


InSēquio™

Using its complementary binding properties, DNA can
be ―programmed‖ to self-assemble into target designs

9 DNA strands
9 strands

1 DNA motif 4x4 grid 8x8 grid


The limit of microtechnology:
1 micron box shown in blue
on a smooth muscle cell

100 nm


Using theof microtechnology:of DNA,
The limit binding properties
1 micron box shown in blue to be
inSēquio enables molecules
on a smooth muscle cell
engineered at the nanoscale.

QuickTime™ and a

80 nm

100 nm


InSēquio™

The resultant structures have the important property
of being uniquely addressable.


Applicable Domains for the D0D
Does your organization do any of the following?

Assembly planning Evolutionary computation Optimization
Biodefense Financial forecasting Parameter studies
Bioinformatics Genomic sequencing Photorealistic Rendering
Biometrics (multi-modal) Logistical planning Proteomics
Combinatorial opt. Machine learning Psychometric profiling
Data mining Modeling and Robotic motion planning
Decision support simulation Search
Decryption Molecular modeling Statistical analysis
Drug discovery Monte Carlo simulations Steganography
Environmental modeling Nanotechnology Weather Prediction
Operations research


Considerations: Feature Set

Business reporting capabilities
Ease of administration
Fault tolerance (―How many 9s?‖)


Considerations: Cost of Dedicated
Hardware
100 node (200-400 core) cluster ~ $200,000 /
yr
(includes cost of power, space, cooling, labor &
support)

Frontier Online
 $0.10 - $0.30 / cap-hour (Ch)
 100 C @ $0.15 / Ch for 10 hrs = $150

Frontier Enterprise
 A 10 : 1 price-performance advantage

Final Considerations
The DoD has ~5M desktops
At 20% utilization, rougly $10B of capacity
is wasted annually
At the same time, there are few examples
of high-value SOA mashups that demonstrate
the value of this latent capacity
This ―Mashup Gap‖ has resulted in
requirements
that may not live up to the demands of the
applications that will ultimately create the most
value for the DoD

Q&A


Grids Clouds Computation Steven Armentrout

Recommandé

Recommandé

Contenu connexe

En vedette

En vedette (6)

Similaire à Grids Clouds Computation Steven Armentrout

Similaire à Grids Clouds Computation Steven Armentrout (20)

Plus de GovCloud Network

Plus de GovCloud Network (20)

Dernier

Dernier (20)

Grids Clouds Computation Steven Armentrout