Distributed Computing

FIT3142 Distributed Computing
Assignment Part 2:
Performance Modelling in a Distributed System
1 Assignment Task Instructions (110 Marks/15 percent) 2
1 Assignment Task Instructions (110 Marks/15 percent)
Length: No more than 1,350 words
Learning Objectives (LO): The Intended LOs achieved by completing this task include LO
4, LO 5 and LO 8 (please see unit guide for details).
The assignment provides an opportunity for you to test your understanding of basic
ideas in Distributed Computing against real-world examples by measuring latency,
summarising and contrasting the di erences in performance models, and testing
these models against a requirement for a real world application.
Answer the following questions 1 through 4, and submit an electronic copy of this work by
23:59 hrs Friday, Week 11 via Moodle.
Consider the following scenario: you are a software development team leader working in
industry, and your project manager approaches you with the following request:
\We have been awarded a contract to develop a distributed app to address a critical need
from one of our major clients. The system will employ a user client application hosted on a
home user desktop, notebook or pad, it will connect over a network to a web server front end
system, that will access a compute server running on a cloud platform, and a big data storage
archive running on a cluster.
The client is absolutely paranoid about response time performance of the production system,
if it is too slow their subscribers will complain or cancel their services. So before we write
any code, we have to look at critical performance issues. The client has insisted we deliver a
response time no worse than 1 second for 95 percent of user operations. They have not supplied
us with any breakdown of user transaction types so we will have to do this with estimates.
They want us to explain what methods we will use for modelling application performance and
sizing the servers.
I will have to provide the performance analysis and explanation report to the client’s program
manager who is extremely pedantic about meeting specs – her job is on the line if we mess
this up, and she has no patience for fools.
I need your latency performance estimates and recommendations on performance modelling in
three weeks for review, before I splice it into the project de nition document I am preparing
for the client’s program manager.”
Given this scenario, how should you keep your project manager happy? If you fudge the answers
the company might spend two million dollars developing the code and then it will fail and your
c 2020, Faculty of IT, Monash University
1 Assignment Task Instructions (110 Marks/15 percent) 3
Figure 1: The client’s system architecture for the distributed application.
employer gets sued, and your team will lose its jobs. Do it properly and you might get a bonus
or a promotion.
Since your manager is not a fool, he will want to see a proper analysis, and some explanation
of how and why you reached the conclusions you did the way you did. He will not have
the time to read many pages of description, and knows more than you do about designing
distributed systems, so describing things he already knows will simply annoy or o end him –
your explanations have to be short and focussed only on the key points.
The following de nes the ve questions in your assignment. Please comply with the word
counts speci ed, there will be no additional marks for overlength submissions:
1. Latency Estimation for the Distributed Application (no more than 300 words):
(a) Using Ookla Speedtest on your desktop (or laptop), measure bandwidth and latency
to ve di erent servers in the Melbourne area and record the results. Perform this
test at 11 AM, 2 PM, 5 PM and 8PM. Tabulate the results and compute the mean
() and standard deviation () of the latency at each of the four times of the day
(tabulated results);
(b) Using Ookla Speedtest on your smartphone (turn o its WiFi), measure bandwidth
and latency to ve di erent servers in the Melbourne area and record the results.
Perform this test at 11 AM, 2 PM, 5 PM and 8PM. Tabulate the results and
c 2020, Faculty of IT, Monash University
1 Assignment Task Instructions (110 Marks/15 percent) 4
compute the mean () and standard deviation () of the latency at each of the
four times of the day (tabulated results);
(c) Compare the measured mean latencies for the desktop (or laptop) and the smart-
phone, and explain any observed di erences in terms of the service and protocols
employed (no more than 100 words);
(d) Using the VMs on the Nectar cloud, measure the mean and standard deviation of
latency with iperf -u -e -s / iperf -u -c server ip;
(e) Estimate the worst case latencies for the desktop (or laptop) and the smartphone
for a 95 percent con dence interval using the  + 2 approximation, and explain
how you would apply this to the system architecture depicted in Figure 1 (no more
than 100 words);
(f) Estimate the worst case latencies between data centre front end, compute and
storage servers for a 95 percent con dence interval using the  + 2 approximation
and your Nectar cloud measurements, and then explain how you would apply this
to the system architecture depicted in Figure 1 (no more than 100 words);
(g) If the client insists that 95 percent of user transactions occur inside of 1 second
response time, what is the required worst case aggregated response time of the data
centre components of the distributed system, assuming that in every transaction
the client device accesses the front end server, the front end server accesses the
compute server, and the compute server accesses the storage server? Explain
whether the system is feasible assuming the use of the home broadband services
you measured latency on? (no more than 100 words);
2. Performance Estimation Methods for the Servers (no more than 450 words):
The client has asked for a performance analysis for the application, and expects an ex-
planation of what modelling methods will be used and exactly why they will be used.
Summarise and compare the suitability of the Amdahl, Harmonic Mean and MRM per-
formance models for modelling the performance of large servers:
(a) Compare the di erences, especially in how they capture performance (no more than
150 words);
(b) Compare the weaknesses of the three models (no more than 150 words);
(c) Compare the strengths of the three models (no more than 150 words);
3. Front End Server (no more than 200 words): The Front End Server in Figure 1 will
run a web services application on a Linux cluster platform. The application will accept
queries from the user client application, and then launch jobs on the Compute Server
that will access datasets on the Storage Server.
(a) What is the most critical performance requirement for this server and why? (no
more than 100 words);
c 2020, Faculty of IT, Monash University
1 Assignment Task Instructions (110 Marks/15 percent) 5
(b) Which of the Amdahl, Harmonic Mean and MRM performance models is the best
t for this server? (name the model);
(c) Why is the model you chose the best t for this server? (no more than 100 words);
4. Compute Server (no more than 200 words): The Compute Server in Figure 1 will
run a compute bound application on a Linux cloud platform. The application will accept
queries from the Front End Server, and then launch jobs that will access datasets on
the Storage Server. The query requests may be highly variable in the size of the data
processed, and the computational complexity of the request.
(a) What is the most critical performance requirement for this server and why? (no
more than 100 words);
(b) Which of the Amdahl, Harmonic Mean and MRM performance models is the best
t for this server? (name the model);
(c) Why is the model you chose the best t for this server? (no more than 100 words);
5. Storage Server (no more than 200 words): The Storage Server in Figure 1 will run
a distributed le system application such as Hadoop on a Linux cluster platform. The
application will accept queries from the Compute Server, and then respond with large
blocks of data it nds in the datasets it stores.
(a) What is the most critical performance requirement for this server and why? (no
more than 100 words);
(b) Which of the Amdahl, Harmonic Mean and MRM performance models is the best
t for this server? (name the model);
(c) Why is the model you chose the best t for this server? (no more than 100 words);
Helpful Tip #1: Students that are successful in this task source and contextualise information
found in published books, professional and industry websites, and peer-reviewed journal and
conference papers. Students that rely on summarising information from the FIT3142 lecture
slides, Wikipedia, popular media, blogs and other unvalidated sources will attract minimal
marks. Remember the golden rule: \Garbage-in Garbage-out” – if you regurgitate
nonsense and mistakes published by others, you will lose marks!
Helpful Tip #2: Note that \summarising and contrasting the di erences” between two items
requires that you identify how the items di er, and then describe the di erences and why they
matter in the shortest possible manner. Simply describing and/or explaining the two
items will attract ZERO marks as these are completely di erent tasks!
Helpful Tip #3: Learning resources on Moodle include a paper that is useful for this assign-
ment.
c 2020, Faculty of IT, Monash University
1 Assignment Task Instructions (110 Marks/15 percent) 6
Helpful Tip #4: Students who need a refresh on the normal distribution might nd this
tutorial useful: Khan Academy: Deep de nition of the normal distribution.
Monash university takes academic integrity seriously: It is an academic requirement that
your submitted work be original. Zero marks will be awarded for the whole submission if there
is any evidence of copying, collaboration, pasting from websites, or copying from textbooks.
All work will be checked using the Turnitin system which has a 100% success rate in detecting
copying o the web and between students. Previously, students who plagiarised have been
referred to the disciplinary committee.
Submission details: Submit an electronic copy of this work by 16:00 hrs Friday, Week 11
via Moodle. Email your lecturer, if you require any further information regarding this task or
have suitable grounds for seeking an extension.
Do Do Not
Contrast key di erences between items Describe and/or explain the items
Use peer reviewed papers, industry
white papers, standards documents
and other professional sources
Use Wikipedia, blogs, popular media,
opinion articles, amateur websites,
discussion groups as sources
Cite 3-4 sources per question Cite a single or no source
Use citations in IEEE format (numerical) Use any other citation style
Submit the report in PDF format.
The format is A4, with1.5 cm left and
right margins, 10 point sanf serif font
(eg Helvetica, Calibri or Arial).
Submit as an MS Word, text le
or scanned images.
Submit with the following naming
convention:
MyStudentNumber-3142-Ass-1.pdf
Make up your own naming convention
(as marks will be deducted)
Submit before or/ on the due date.
Submit late as penalties will apply,
10 marks per day or part thereof.
Apply for extension using the correct
Monash paperwork, if an unforeseen
circumstance arises, such as illness.
(be prepared to submit supporting
documentation and/or medical
certi cate)
Ask for an extension on the due date
or after the due date, as Monash
penalties will apply.
Additional Notes:
1. The quality of English language grammar used in the assignment will NOT be marked.
The structure, content, understanding, and referencing WILL be marked.
c 2020, Faculty of IT, Monash University
1 Assignment Task Instructions (110 Marks/15 percent) 7
2. IEEE Transactions templates for MS Word and LATEX are available here, not mandatory
but convenient: https://www.ieee.org/publications_standards/publications/
authors/author_templates.html
3. No additional marks will be given for assignments that are over the speci ed length. If
the assignment is more than 50% longer than speci ed, the total mark will be decreased
by 10% for every excess page in the report. Reports that are shorter than the minimum
will lose marks in proportion to reduced content.
4. Time budgeting is important. Plan your time carefully and start early if possible on the
assignment.
5. Learning outcomes – this assignment is intended to provide you with experience in in-
dependently collecting and understanding research materials, as you will be expected to
do in industry jobs.
c 2020, Faculty of IT, Monash University
2 Marking Guide 8
2 Marking Guide
Tables 1 and 2 explain exacty how the assignment will be marked. Please take the time to
study the marking criteria and the marking breakdown.
Most often marks are lost due to the following causes:
1. Students described the items being compared instead of comparing them.
2. Students explained the items being compared instead of comparing them.
3. Students described the items being explained instead of explaining them.
4. Students used poor quality references that did not provide the required content to
make proper comparisons.
5. Students used poor quality references that were in error and students lost marks due
to these errors.
6. Students did not spend sucient time analysing the references and missed com-
parison points or introduced errors.
c 2020, Faculty of IT, Monash University
2 Marking Guide 9
N Fail Pass Credit Distinction High Distinction
None Limited Some Most All
0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Compare from Not enough detail Satisfactory detail Almost full detail Comprehensive
the lecture slides in the comparison, in the comparison, in the comparison, and critical
or poor sources, some errors of fact few errors of fact no errors of fact comparison,
explains and/or or ambiguity or ambiguity or ambiguity demonstrating full
describes instead understanding,
of comparison, no errors of fact
many errors
of fact, frequent
ambiguity
Table 1: Marking criteria for assignment answers. Each subtask will be marked against these criteria. Please note that marks are
awarded only for real answers and descriptions or explanations instead of comparisons attract ZERO marks.
c 2020, Faculty of IT, Monash University
2 Marking Guide 10
Task Subtask Total
N
Fail
N
Fail
Pass Credit Distinction
High
Distinction
115 No Attempt None Limited Some Most All
Q1.Network Latency (a) desktop/notebook 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q1.Network Latency (b) mobile device 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q1.Network Latency (c) compare latency 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q1.Network Latency (d) server latency 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q1.Network Latency (e) worst case estimate 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q1.Network Latency (f) worst case estimate 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q1.Network Latency (g) worst case estimate 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q2.Estimation Methods (a) di erences 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q2.Estimation Methods (b) strengths 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q2.Estimation Methods (c) weaknesses 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q2.Estimation Methods References 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q3.Front End Server (a) critical reqt 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q3.Front End Server (b) best t 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q3.Front End Server (c) why best t 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q3..Front End Server References 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q4.Compute Server (a) critical reqt 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q4.Compute Server (b) best t 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q4.Compute Server (c) why best t 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q4.Compute Server References 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q5.Storage Server (a) critical reqt 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q5.Storage Server (b) best t 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q5.Storage Server (c) why best t 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Q5.Storage Server References 5 0 0-2.5 2.5-3.0 3.0-3.5 3.5-4.0 4.0-5.0
Table 2: Marking breakdown for assignment answers.
c 2020, Faculty of IT, Monash University
3 Submission Checklist 11
3 Submission Checklist
1. Submit an electronic copy of this work by 23:59 hrs Friday, Week 11 via Moodle.
2. The report must be in PDF format, MS Word will not be accepted.
3. All les must be named using the convention MyStudentNumber-3142-Ass-2.pdf.
Marks will be deducted for bad formatting if naming conventions are not followed.
4. Late submission without a good reason agreed in advance of submission due date will
result in a penalty deduction of 10 marks per calendar day or part thereof. If you believe
that your assignment will be delayed because of circumstances beyond your control such
as illness you should apply for an extension before the due date. Medical certi cates or
certi cation supporting your application may be required.
5. It is an academic requirement that your submitted work be original. Zero marks will
be awarded for the whole submission if there is any evidence of copying, collabora-
tion, pasting from websites, or copying from textbooks. All work will be checked using
the Turnitin system which has a 100% success rate in detecting copying o the web
and between students. Previously, students who plagiarised have been referred to the
disciplinary committee.
6. One PDF report as detailed above.
Order Now

Calculate a fair price for your paper

Such a cheap price for your free time and healthy sleep

1650 words
-
-
Place an order within a couple of minutes.
Get guaranteed assistance and 100% confidentiality.
Total price: $78
WeCreativez WhatsApp Support
Our customer support team is here to answer your questions. Ask us anything!
👋 Hi, how can I help?