CS计算机代考程序代写 chain capacity planning COMP9334: Capacity Planning of Computer Systems and Networks

COMP9334: Capacity Planning of Computer Systems and Networks
Course Review and Exam Info
A/Prof Chun Tung Chou CSE, UNSW

System performance is important
Performance metrics: response time, waiting time, throughput, availability
Performance is determined by:
Workload
System parameters
You can estimate system performance, without building the actual system, by using queueing models
Page 1

Performance analysis techniques (1)
Operational analysis (1B,2A)
Measurements on the systems Operational laws, in particular Little’s Law Key concept: Bottleneck
Upper bound on system performance
Page 2

Performance analysis techniques (2)
Need to identify the inter-arrival and service time distributions
Queues with Poisson arrival
Exponential service time (2B,3A)
Single or multiple servers: M/M/1 versus M/M/m Infinite buffer or finite buffer: M/M/m versus M/M/m/m+k
General service time distribution (4A,7A) M/G/1. Key concept: residual service time Priority queueing
Page 3

Performance analysis techniques (3)
Closed queueing networks with exponential service time
Markov chain analysis (3B)
Recipe: Identify state, transition probability, solve steady state probability, determine performance
Mean value analysis (Week 7B)
Iterative method
n = 0 jobs → n = 1 job → n = 2 jobs → …
Page 4

Performance analysis techniques: some key points
No universal analytical methods
Analytical solutions are only available for specific classes of queues
Upper bounds are only available for some general classes of queues
Simulation can be used to determine general queueing problems
Page 5

Simulation (1)
How to do discrete event simulation?
How to generate random events according to the specific inter- arrival time and service time distributions
Generating uniformly distributed pseudo-random numbers Inverse transformation method
Page 6

Simulation (2)
Simulation is not just about writing correct simulation code (though it is important), it is also important to do sound statistical analysis on the simulation results obtained
Transient removal
Independent replications
Confidence interval
How to decide whether one system is better than the other using confidence interval?
Paired observations: Paired-t confidence interval Approximate visual test
Variance reduction method
Page 7

Capacity planning and performance analysis
Solve the capacity planning problem by solving a number of performance analysis problems
Example: Revision Problem: Week 3B, Question2
Page 8

Applications of queueing (7A 2,8A)
Web services Fork-join queues
Other applications
Determining a good multi-programming level Power allocation in server farm
Server farm with set-up cost
Page 9

Integer programming (1)
Linear programming (LP)
Real values for decision variables, linear in objective function, linear in constraints
Integer programming (IP)
Some decision variables can only take integer values
Some decision variables can only take binary values, e.g. for making yes-or-no decisions
Page 10

Integer programming (2)
Applications of integer programming in network flow problems
Flow conservation constraints to ensure a unique path between two nodes in the network
Example applications
Traffic engineering Network design
Page 11

Integer programming (3)
Applications of integer programming in placement problems
Placement of wireless access points
Placement of controllers in software-defined networks (revision problem)
Power of binary variables
Restricted range of values Either-or constraints Piecewise linear functions
Page 12

Summary
What you have learned through this course are fundamental techniques that can be applied to designing computer systems and networks to have good performance
We hope you have gained some skills from this course
We hope you have been trained to be performance-minded
Due to the limited time and scope of this course, we cannot cover all techniques that have been developed in this field
However, with the knowledge you have acquired from this course, you should have the foundation to learn more …
Mathematical methods can be applied to many different areas
Page 13

Final exam
Wed 5 May 2021, 1:45-4pm (Sydney time)
2 hours and 15 minutes
If you are in a different location, make sure you know the corresponding local time
Browser based exam (more on this later)
Access via the COMP9334 Moodle page
Exam syllabus:
No programming questions
No extensive computation
Not examined: Simulation (Weeks 4B, 5A, 5B). Week 7A 2. All other topics are examinable
Page 14

Final exam (cont.)
Open book and open web
Unproctored
UNSW student code, which covers exam conducts, applies Must be your own work.
Question aims to test understanding, not memorization
Style similar to revision problems, assignment and sample exam
Show your steps in your answer
If you just write the final answer, you won’t receive full marks even if it is correct
You can receive marks for your steps
Page 15

Inspera – a browser based exam platform
The exam will use Inspera (screen shot on next slide) Should try to familiarise yourselves with the platform
Watch the webinar recordings on Inspera https://unsw. sharepoint.com/sites/Assessment-Platform-Pilot Practice using the sample exam
Know what you should do if your Internet connection is lost
Answer is auto-saved. Your answers will be automatically submitted at the end of the exam unless your Internet connection is lost.
Page 16

Screen shot of Inspera
Page 17

Answer box and answer formatting
For COMP9334, all questions use the same type of Inspera answer box.
This applies to the sample exam and the actual exam.
I am more interested to know whether you understand the subject matter or not
Write and format in a way to help me to understand your idea
Although Inspera has a good editor, some maths formatting is optional, e.g. OK to write p_{123} instead of p123, 10ˆ6 instead of 106, lambda instead of λ etc.
Page 18

Important documents and information
Screen shot from Moodle. Do read and understand the documents. You should read them before the exam.
Page 19

Preparations for final exam
Make sure you understand all the concepts, techniques and examples discussed in the lectures
Work through all the sample problems, assignment questions, etc for practice
Sample exam on Moodle (more on this later)
Misconception: Open-book exam means no preparation required Consultations: See course website under Timetable
Further questions
Post on Forum (Try to avoid last minute questions.)
Page 20

Sample exam on Moodle
Accessible via the COMP9334 Moodle page
Most questions are based on past exams
Settings
Available from 2pm on 22 April 2021
You can access it multiple times
No time limit set – you can time yourselves if you wish
Do not need to submit
A sample solution will be provided
The cover page is the same as what you will see in the exam This is so that you know the instructions in advance
Page 21

Exam policy
Exam policy
If you attend the exam, we assume that you are functioning well on that day and will not offer you a supplementary exam.
In case you are not well on the day of the exam, get a medical certificate and do not attend the exam.
If you fall ill during the exam, follow the steps in Student Declaration point #4.
If you have queries on the exam, you can ask on the forum.
Page 22

Parting messages
Please complete the myExperience survey
Good/bad/more of this/less of that/what can be done better
This course is different from many CSE courses ..
Analytical and simulation methods are useful for many disciplines
This world needs people with multiple skills (hard and soft). Important to find your talents and passions, but try to explore and learn as many different areas as you can.
Page 23