Date post: | 02-Jan-2016 |
Category: |
Documents |
Upload: | nola-watson |
View: | 32 times |
Download: | 4 times |
Leonard Kleinrock 2010
Leonard KleinrockLeonard KleinrockDistinguished ProfessorDistinguished Professor
Computer Science DepartmentComputer Science DepartmentUCLA UCLA
Leonard Kleinrock 2010
• Many youth feel there is no great work left to Many youth feel there is no great work left to be done.be done.
• Not true – the golden era continues.Not true – the golden era continues.• There are lots of challenging problems left.There are lots of challenging problems left.• And their impact will be great inAnd their impact will be great in
• today’s global economytoday’s global economy• and in tomorrow’s unimaginable and in tomorrow’s unimaginable world.world.
• These are times of enormous change.These are times of enormous change.• And times of enormous opportunity.And times of enormous opportunity.
Golden OpportunitiesGolden Opportunities
Leonard Kleinrock 2010
Don’t ForgetDon’t Forget• But as we rush headlong into the future, we must not forget where all this came from – we must not ignore the lessons from history.• Most of today’s youngsters cannot
remember a time when there was no Internet. • They cannot imagine that their parents are “of this era”• Don’t underestimate those seniors.
Leonard Kleinrock 2010
My Five Golden Guidelines to My Five Golden Guidelines to ResearchResearch
1.1. Conduct the 100-year test.Conduct the 100-year test.2.2. Don’t fall in love with your model. Don’t fall in love with your model. 3.3. Beware of mindless simulation.Beware of mindless simulation.4.4. Understand your own results.Understand your own results.5.5. Look for “Gee, that’s funny!”Look for “Gee, that’s funny!”
Leonard Kleinrock 2010
Richard HammingRichard Hamming"Why do so few scientists make significant contributions and so many are forgotten in the long run?"
Richard W. Hamming, “You and Your Research”, March 7, 1986.
“If you don't work on important problems, it's not likely that you'll do important work.”
Leonard Kleinrock 2010
1. The 100 Year Test1. The 100 Year Test
• Hamming once asked me, Hamming once asked me,
““What progress of today will be remembered What progress of today will be remembered 1000 years from now ?”1000 years from now ?”
Will your work be remembered 100 years Will your work be remembered 100 years from today?from today?
Leonard Kleinrock 2010
2. The Modeling Process2. The Modeling Process
TheThe RealReal
WorlWorl
dd
MathematicaMathematicall
Model ofModel ofThe Real The Real
WorldWorld
Solution to Solution to thethe
MathematicMathematicalal
ModelModel
Leonard Kleinrock 2010
MathematicaMathematicall
Model ofModel ofThe Real The Real
WorldWorld
Solution to Solution to thethe
MathematicMathematicalal
ModelModel
2. But Don’t Fall in Love With 2. But Don’t Fall in Love With Your ModelYour Model
Approximation
TheThe RealReal
WorlWorl
dd
Leonard Kleinrock 2010
Michael FaradayMichael Faraday
• "I could not imagine much progress by "I could not imagine much progress by reading only, without experimental facts and reading only, without experimental facts and trials … I was never able to make a fact my trials … I was never able to make a fact my own without seeing it. “ (Faraday: 1827)own without seeing it. “ (Faraday: 1827)
• Faraday performed in his brain the work of a Faraday performed in his brain the work of a great mathematician without using a single great mathematician without using a single mathematical formula” (Hermann von mathematical formula” (Hermann von Helmholtz : April 5, 1881 )Helmholtz : April 5, 1881 )
Leonard Kleinrock 2010
3. Beware of Mindless3. Beware of MindlessSimulationSimulation
Leonard Kleinrock 2010
3. Beware of Mindless3. Beware of MindlessSimulationSimulation
Leonard Kleinrock 2010
3. Beware of Mindless3. Beware of MindlessSimulationSimulation
Leonard Kleinrock 2010
3. Beware of Mindless3. Beware of MindlessSimulationSimulation
Leonard Kleinrock 2010
3. Beware of Mindless3. Beware of MindlessSimulationSimulation
Leonard Kleinrock 2010
3. Beware of Mindless3. Beware of MindlessSimulationSimulation
Leonard Kleinrock 2010
3. Beware of Mindless3. Beware of MindlessSimulationSimulation
Leonard Kleinrock 2010
Ask the Obvious QuestionsAsk the Obvious Questions
Slope
3. Beware of Mindless3. Beware of MindlessSimulationSimulation
Intercept
Breakpoint
Asymptote
Leonard Kleinrock 2010
4. Understand Your Own Results4. Understand Your Own Results
Leonard Kleinrock 2010 Network CloudNetwork Cloud
InputInput
Response TimeResponse Time
TTRESPONSE TIMERESPONSE TIME
0
CAPACITYCAPACITY
Loss
LOSSLOSS
Leonard Kleinrock 2010
Response Time vs Throughput
0Throughput
ResponseResponse TimeTime
( )
T( )
Do you wantDo you wantto operateto operate
here?here?Or here?Or here?
Now let’s ask a good question:
Leonard Kleinrock 2010
0Throughput
ResponseResponse TimeTime
( )
T( ) P = T( )
1/P
*
Max Power Point
Kleinrock, L., "On Flow Control in Computer Networks", Conference Record, Proceedings of the International Conference on Communications, Vol. II, Toronto, Ontario, pp. 27.2.1 to 27.2.5, June 1978.
Let me define a new metric of performance:
POWER = Throughput
Response Time
The main result
Response Time vs Throughput4. Understand Your Own Results
dT( )/d = T( )/dT( )/d = T( )/
Power is max when
Leonard Kleinrock 2010
0Throughput
ResponseResponse TimeTime
( )
T( ) P = T( )
1/P
*
Max Power Point
Kleinrock, L., "On Flow Control in Computer Networks", Conference Record, Proceedings of the International Conference on Communications, Vol. II, Toronto, Ontario, pp. 27.2.1 to 27.2.5, June 1978.
We need a new metric of performance:
POWER = Throughput
Response Time
Response Time vs Throughput4. Understand Your Own Results
Leonard Kleinrock 2010
0Throughput
ResponseResponse TimeTime
( )
T( )
*
Max Power Point
Kleinrock, L., "On Flow Control in Computer Networks", Conference Record, Proceedings of the International Conference on Communications, Vol. II, Toronto, Ontario, pp. 27.2.1 to 27.2.5, June 1978.
Response Time vs Throughput4. Understand Your Own Results
For M/M/1For M/M/1this givesthis gives
max Power at max Power atNN** = 1 = 1
Let’s Dig Deeper on Understanding
Why ?Why ?
Leonard Kleinrock 2010
4. Understand Your Own ResultsUse Your IntuitionUse Your Intuition
TTTT= Min= Min
Eff = MaxEff = Max
Leonard Kleinrock 2010
4. Understand Your Own Results
Only 1 customerOnly 1 customerin the systemin the system
TT= Min= MinEff = MaxEff = Max
Use Your IntuitionUse Your Intuition
Insight:Just keep
the pipe full!
TT
Leonard Kleinrock 2010
• Our intuition says put Our intuition says put exactlyexactly one person one person in the queueing system in the queueing system • This was from “deterministic” reasoning.This was from “deterministic” reasoning.
• We can’t actually do that in generalWe can’t actually do that in general• BUT our earlier result said that we should BUT our earlier result said that we should
adjust the system to achieve an adjust the system to achieve an averageaverage of of one person in the queueing system, i.e., one person in the queueing system, i.e.,
4. Understand Your Own Results
At Max PowerAt Max PowerNN** = 1 = 1
for M/M/1for M/M/1
Leonard Kleinrock 2010
5. Gee, that’s funny!5. Gee, that’s funny!
Leonard Kleinrock 2010
M/M/1M/M/1
M/G/1M/G/1
N*=1N*=1
0
T( )
5. Gee, that’s funny!What can we say for
M/G/1 ?
Leonard Kleinrock 2010
More on ModelingMore on Modeling
•Moving the frontier is tough Moving the frontier is tough (we mislead our students)(we mislead our students)
•Once you do it, you will be able to repeat it Once you do it, you will be able to repeat it (students don’t believe us)(students don’t believe us)•Teach your students to understand their results!Teach your students to understand their results!•Generalization usually comes when you can Generalization usually comes when you can see the simplicity of a solutionsee the simplicity of a solution•Keep your interest in related areas - areas Keep your interest in related areas - areas where where
something might happen.something might happen.
Leonard Kleinrock 2010
Thank YouThank Youwww.lk.cs.ucla.edu