computer networking
0 likes24 views
This document is an examination paper for a module on Virtualisation and Cloud Computing. It consists of 5 questions, and students must answer any 3 questions. The questions cover topics like defining cloud computing characteristics and business models, defining and comparing types of virtualization, analyzing performance of parallel programs using Amdahl's law and Gustafson's law, explaining big data and how Hadoop works, and calculating availability and failure cost for a cluster. Students have 2 hours to complete 3 questions from this exam.
![2 of 4 /continued
VIRTUALISATION AND CLOUD COMPUTING
(CCE4370)
April/May 2018 (main) 2 Hours
Answer any THREE questions out of FIVE.
Each question carries 20 marks.
Any University approved calculator is permitted.
1. Cloud Computing is a relatively new technology, which is generally described
with reference to its essential characteristics, the three business models, and the
three or more deployment methods.
a) Provide a clear and concise definition of Cloud Computing by listing and
illustrating the essential characteristics that distinguish it from other
computing technologies for distributed computing.
[10 marks]
b) Provide a clear and concise definition of the three business models
(Infrastructure as a Service - IaaS, Platform as a Service – PaaS, Sofware
as a Service - SaaS) characterising Cloud Computing.
[10 marks]
2. Virtualisation is one of the technologies that made possible the development of
Cloud Computing.
a) Provide a clear and concise definition of virtualisation by mentioning what
the term virtualisation refers to, and why virtualisation is so important for the
development of Cloud Computing.
[10 marks]
b) Explain the difference between Type I and Type II virtualisation.
[10 marks]](https://image.slidesharecdn.com/cce4370may18-191213061228/85/computer-networking-2-320.jpg)
![3 of 4 /continued
3. Consider a program for updating the nodes’ activation state of a relatively large
Convolution Neural Network with K nodes. The sequential updating time on a
single server is T1 = cK3
minutes, where c is a constant determined by the server
used. An MPI-code parallel program requires Tn = cK3
/n + dK2
/n0.5
minutes to
complete execution on an n-server cluster system, where d is a constant
determined by the MPI version used. Assume the program has a sequential
bottleneck equal to α=0.2. Answer the following questions for a given cluster
configuration with n=64 servers, c=0.8, and d=0.1.
a) Using Amdahl’s law, calculate the speedup of the n-server cluster
configuration for running a fixed workload corresponding to a Convolution
Neural Network with K=200 nodes.
[10 marks]
b) Using Gustafson’s law, calculate the speedup of the n-server cluster
configuration for running a scaled workload corresponding to a Convolution
Neural Network with K=300 nodes.
[10 marks]
4. Assuming you are an IT consultant providing companies solutions for the analysis
of big data. You know that Hadoop framework, thanks to MapReduce can allow
users to process and extract various different type of information from very large
text files. In order to convince the owner of a medium size company to install
Hadoop into the company cluster:
a) Briefly explain what the term “big data” refers to.
[10 marks]
b) Provide a brief definition of the Hadoop Distributed File System and of
MapReduce, and briefly explain how Hadoop works by listing using bullet-
points the main steps for processing a single large data file.
[10 marks]
5. The availability of a cluster, that is the percentage of time that a system is
available to the user (the percentage of system uptime), is generally analysed
using a formalism that expresses Availability in terms of Mean-Time-to-Failure
(MTTF), and Mean-Time-to-Recovery (MTTR).
a) For a given cluster, each node fails on average every 100 hours, and it
takes 10 hours to repair the failed node. Given this scenario, what is the
Availability of this cluster?
[10 marks]](https://image.slidesharecdn.com/cce4370may18-191213061228/85/computer-networking-3-320.jpg)
![4 of 4
b) Now, assuming that time is expressed in hours, and that there are 8760
(i.e., 24 x 365) hours in a year, compute the yearly failure cost of the above
mentioned cluster assuming that each one-hour downtime costs £15?
[10 marks]](https://image.slidesharecdn.com/cce4370may18-191213061228/85/computer-networking-4-320.jpg)
computer networking
- 1. Examination Paper Faculty of Science and Technology MIDDLESEX UNIVERSITY EXAMINATION PAPER Academic Year 2017/2018 (May) MODULE NUMBER: CCE4370 MODULE TITLE: Virtualisation and Cloud Computing MODULE TUTOR’S NAME: Elio Tuci Time allowed: 2 hours Total number of questions: Instructions to candidates: Materials provided: Equipment permitted: Total number of pages: 5 Answer any 3 questions. Each question is worth a maximum of 20 marks. The total (out of 100) will be converted to a percentage after marking None Any University Approved Calculator is permitted 4 (including front cover) EXAM PAPER CAN BE REMOVED FROM THE EXAM ROOM No books, paper or electronic devices are permitted to be brought into the examination room other than those specified above. Candidates are warned that credit cannot be given for work that is illegible.
- 2. 2 of 4 /continued VIRTUALISATION AND CLOUD COMPUTING (CCE4370) April/May 2018 (main) 2 Hours Answer any THREE questions out of FIVE. Each question carries 20 marks. Any University approved calculator is permitted. 1. Cloud Computing is a relatively new technology, which is generally described with reference to its essential characteristics, the three business models, and the three or more deployment methods. a) Provide a clear and concise definition of Cloud Computing by listing and illustrating the essential characteristics that distinguish it from other computing technologies for distributed computing. [10 marks] b) Provide a clear and concise definition of the three business models (Infrastructure as a Service - IaaS, Platform as a Service – PaaS, Sofware as a Service - SaaS) characterising Cloud Computing. [10 marks] 2. Virtualisation is one of the technologies that made possible the development of Cloud Computing. a) Provide a clear and concise definition of virtualisation by mentioning what the term virtualisation refers to, and why virtualisation is so important for the development of Cloud Computing. [10 marks] b) Explain the difference between Type I and Type II virtualisation. [10 marks]
- 3. 3 of 4 /continued 3. Consider a program for updating the nodes’ activation state of a relatively large Convolution Neural Network with K nodes. The sequential updating time on a single server is T1 = cK3 minutes, where c is a constant determined by the server used. An MPI-code parallel program requires Tn = cK3 /n + dK2 /n0.5 minutes to complete execution on an n-server cluster system, where d is a constant determined by the MPI version used. Assume the program has a sequential bottleneck equal to α=0.2. Answer the following questions for a given cluster configuration with n=64 servers, c=0.8, and d=0.1. a) Using Amdahl’s law, calculate the speedup of the n-server cluster configuration for running a fixed workload corresponding to a Convolution Neural Network with K=200 nodes. [10 marks] b) Using Gustafson’s law, calculate the speedup of the n-server cluster configuration for running a scaled workload corresponding to a Convolution Neural Network with K=300 nodes. [10 marks] 4. Assuming you are an IT consultant providing companies solutions for the analysis of big data. You know that Hadoop framework, thanks to MapReduce can allow users to process and extract various different type of information from very large text files. In order to convince the owner of a medium size company to install Hadoop into the company cluster: a) Briefly explain what the term “big data” refers to. [10 marks] b) Provide a brief definition of the Hadoop Distributed File System and of MapReduce, and briefly explain how Hadoop works by listing using bullet- points the main steps for processing a single large data file. [10 marks] 5. The availability of a cluster, that is the percentage of time that a system is available to the user (the percentage of system uptime), is generally analysed using a formalism that expresses Availability in terms of Mean-Time-to-Failure (MTTF), and Mean-Time-to-Recovery (MTTR). a) For a given cluster, each node fails on average every 100 hours, and it takes 10 hours to repair the failed node. Given this scenario, what is the Availability of this cluster? [10 marks]
- 4. 4 of 4 b) Now, assuming that time is expressed in hours, and that there are 8760 (i.e., 24 x 365) hours in a year, compute the yearly failure cost of the above mentioned cluster assuming that each one-hour downtime costs £15? [10 marks]