© Copyright 2007 Fraunhofer Center MD 1 Life at Fraunhofer Center - Maryland Dr. Forrest Shull...

11
1 opyright 2007 Fraunhofer Center MD Life at Fraunhofer Center - Maryland Dr. Forrest Shull Division Director, Measurement & Knowledge Management

Transcript of © Copyright 2007 Fraunhofer Center MD 1 Life at Fraunhofer Center - Maryland Dr. Forrest Shull...

Page 1: © Copyright 2007 Fraunhofer Center MD 1 Life at Fraunhofer Center - Maryland Dr. Forrest Shull Division Director, Measurement & Knowledge Management.

1

© Copyright 2007 Fraunhofer Center MD

Life atFraunhofer Center - Maryland

Dr. Forrest ShullDivision Director,

Measurement & Knowledge Management

Page 2: © Copyright 2007 Fraunhofer Center MD 1 Life at Fraunhofer Center - Maryland Dr. Forrest Shull Division Director, Measurement & Knowledge Management.

2

© Copyright 2007 Fraunhofer Center MD

Fraunhofer CenterFraunhofer Centerfor Empirical Software Engineeringfor Empirical Software Engineering

(College Park, MD)(College Park, MD)

• Not-for-profit, university-affiliated

• Mandate is applied research and technology transfer

• Via measurement, evaluation and improvement

• We bridge research and application, working with:– Government organizations (DoD, NASA, …) – Large industrial companies (Boeing, Motorola, …)– Small/medium-sized organizations (KeyMind,…)– National labs (Los Alamos Labs, ….)– Universities and research centers (CMU, USC, MIT, …)– …and others (SEI, MITRE, JHU/APL, ...)

Page 3: © Copyright 2007 Fraunhofer Center MD 1 Life at Fraunhofer Center - Maryland Dr. Forrest Shull Division Director, Measurement & Knowledge Management.

3

© Copyright 2007 Fraunhofer Center MD

An example project: HPCSAn example project: HPCS

• US Govt. wants to buy new HPC systems for national labs• Criteria for choosing machines: minimize time to solution (TTS)

• Not all labs are the same– Programmers may have different levels of experience

• One system may be more productive for experts, another for novices

– Problems being solved are different• One system may be better for climate modeling, another for

protein folding

• Research question: – Which systems / development approaches will result in higher

productivity for given users, problem domains, etc.?

TTS = Development time + Computing time

Page 4: © Copyright 2007 Fraunhofer Center MD 1 Life at Fraunhofer Center - Maryland Dr. Forrest Shull Division Director, Measurement & Knowledge Management.

4

© Copyright 2007 Fraunhofer Center MD

An example project: HPCSAn example project: HPCS% Effort saved using OpenMP instead of MPI

BuffonMPI

BuffonOpenMP

MatvecMPI

MatvecOpenMP

ResistorsMPI

ResistorsOpenMP

LifeMPI

LifeOpenMP

SharksMPI

SharksOpenMP

SharksCAF

SWIMOpenMP

SortingMPI

0.0

0.2

0.4

0.6

0.8

1.0

Problem type

Effic

ien

cy

Performance by problem/HPC model

-100

-50

050

100

% e

ffo

rt r

edu

ctio

n

Buffon Matvec Resistors Life Sharks

Effect of model & problem on

Performance

Effort

Page 5: © Copyright 2007 Fraunhofer Center MD 1 Life at Fraunhofer Center - Maryland Dr. Forrest Shull Division Director, Measurement & Knowledge Management.

5

© Copyright 2007 Fraunhofer Center MD

A day in the life…A day in the life…

• Responsibilities– Project management– Customer deliverables– Business development– Publications & Conference presentations

• Rewards– Publications & Conference presentations– Work with…

• Many different people • In varied contexts • Doing interesting things

– Independence

Page 6: © Copyright 2007 Fraunhofer Center MD 1 Life at Fraunhofer Center - Maryland Dr. Forrest Shull Division Director, Measurement & Knowledge Management.

6

© Copyright 2007 Fraunhofer Center MD

How to prepare for a job like thisHow to prepare for a job like this

• Read Fred Brooks, “The Computer Scientist as Toolsmith II”– Communications of the ACM, March 1996, 39(3): 61-68.

• Be interested in many different fields and how they use computers and software to accomplish things

• Read Victor Basili et al., “SEL’s Software Process-Improvement Program”– IEEE Software, Nov. 1995, 12(6): 83-87.

• Be interested in research that makes real changes

• Be interested in how to reason about the effect of new methods

• Read Barry Boehm, “Software Engineering Economics”– Prentice Hall, 1981.

• Be interested in choices, decisions and tradeoffs

Page 7: © Copyright 2007 Fraunhofer Center MD 1 Life at Fraunhofer Center - Maryland Dr. Forrest Shull Division Director, Measurement & Knowledge Management.

7

© Copyright 2007 Fraunhofer Center MD

How to prepare for a job like thisHow to prepare for a job like this

• Spend some time as a software developer and understand the frustrations

• Be interested in how humans use technology– And have some background in stats to reason about this

• Be attracted to both university and industry

Page 8: © Copyright 2007 Fraunhofer Center MD 1 Life at Fraunhofer Center - Maryland Dr. Forrest Shull Division Director, Measurement & Knowledge Management.

8

© Copyright 2007 Fraunhofer Center MD

Backups

Page 9: © Copyright 2007 Fraunhofer Center MD 1 Life at Fraunhofer Center - Maryland Dr. Forrest Shull Division Director, Measurement & Knowledge Management.

9

© Copyright 2007 Fraunhofer Center MD

High Dependability Computing Project Goals

• Improve NASA’s ability to build dependable software• Define dependability (system & stakeholder perspectives)• Investigate, foster, and transfer to practice technologies • Use testbeds for technology assessment: scaled-down or preliminary

versions of systems in NASA relevant domains which require high dependability, e.g.– Autonomous rover– Air traffic control systems

Rocky 7 rover

Page 10: © Copyright 2007 Fraunhofer Center MD 1 Life at Fraunhofer Center - Maryland Dr. Forrest Shull Division Director, Measurement & Knowledge Management.

10

© Copyright 2007 Fraunhofer Center MD

Using the Experience Base: The Best Practices Clearinghouse

Page 11: © Copyright 2007 Fraunhofer Center MD 1 Life at Fraunhofer Center - Maryland Dr. Forrest Shull Division Director, Measurement & Knowledge Management.

11

© Copyright 2007 Fraunhofer Center MD

Using the Experience Base: The Best Practices Clearinghouse

Flexiblefront-end

I’m on a security-critical project in an industrial organization. Let me browse practices that help me reach CMMI KPAs.

I’m working in the DoD and read about Fagan inspections. Is anyone at a similar organization using it?

I’m working on an warfighter at an industry contractor organization. Show me practices that can help reduce schedule.