CMSC711 Spring 2009 Final Exam: Instructions and Questions - Prof. Neil Spring, Exams of Computer Systems Networking and Telecommunications

Instructions and questions for the final exam of the cmsc711 spring 2009 course. The exam covers topics such as bittorrent, facebook connect, end-to-end redundancy suppression, and dns queries. Students are expected to write concise answers, cite sources if used, and avoid collaboration. The exam consists of five questions, each taking approximately 30-45 minutes to complete. The document also includes information on page limits, internet use, and exam duration.

Typology: Exams

Pre 2010

Uploaded on 07/30/2009

koofers-user-g2m-1
koofers-user-g2m-1 🇺🇸

10 documents

1 / 1

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
CMSC711 Spring 2009 Final Exam
This exam is due 11am Thursday May 14, in pdf form uploaded to the course blog.
On length: No question should require more than a half page to answer. Answers longer than a page will
not be graded. You understand from the blog comments that I appreciate brevity and I expect you to show
confidence in your answer. The latex \clearpage command may be useful. Diagrams don’t count against
the page limit.
On internet / wikipedia use: Sources are legal if cited. Sources are forbidden if not cited. If you find
papers that state arguments or answer the questions, you may quote them, but please develop your own
answer. To cite papers from the reading list: just write the first author’s name in square brackets, no need
for a real reference.
Don’t collaborate, this is an exam.
I expect each question to take 30-45 minutes for roughly three to four hours.
Think before you write: aim to write your answer in the first sentence, and support if needed.
1. Design an approach to determine whether a remote BitTorrent client is using Levin’s PropShare al-
gorithm, and if so, what upload rate to use when exchanging data. Note that the remote client may
participate in other swarms and may exchange with other peers at the same time.
2. Apply facebook’s “Facebook Connect” service to solve a problem in any of the papers we read this
semester. Broken down:
(a) Choose a paper;
(b) state a relevant problem (it need not have been raised by the paper’s authors);
(c) suggest how using friend information might simplify addressing that problem.
(d) explain why friend information (not of friends-of-friends) would be sufficient, or describe how they
can be made so.
3. To provide end-to-end (host-to-host) redundancy suppression, of the sort suggested by Akella, answer
the following. Assume that you’re compressing only one direction.
(a) what protocol layer(s) would (necessarily) be involved? (Use OSI reference layers; You may invent
layers between other layers.)
(b) how would dictionary misses be reported?
(c) would retransmissions be suppressed? How or how not?
(d) in the event of lossy links in the end-to-end path, how might the dictionaries be efficiently syn-
chronized?
4. RED violates the end to end argument. True or false; and why?
5. DNS queries to root servers are predominately junk. Design a scheme to eliminate junk queries to
the root servers (from compliant name servers). You may modify the protocol to add resource records
or response types. (Caveat: yes, most of the bogus queries are from buggy resolvers or buggy name
servers. Another layer of bogus-query-prevention is the goal.)
1

Partial preview of the text

Download CMSC711 Spring 2009 Final Exam: Instructions and Questions - Prof. Neil Spring and more Exams Computer Systems Networking and Telecommunications in PDF only on Docsity!

CMSC711 Spring 2009 Final Exam

This exam is due 11am Thursday May 14, in pdf form uploaded to the course blog. On length: No question should require more than a half page to answer. Answers longer than a page will not be graded. You understand from the blog comments that I appreciate brevity and I expect you to show confidence in your answer. The latex \clearpage command may be useful. Diagrams don’t count against the page limit. On internet / wikipedia use: Sources are legal if cited. Sources are forbidden if not cited. If you find papers that state arguments or answer the questions, you may quote them, but please develop your own answer. To cite papers from the reading list: just write the first author’s name in square brackets, no need for a real reference. Don’t collaborate, this is an exam. I expect each question to take 30-45 minutes for roughly three to four hours. Think before you write: aim to write your answer in the first sentence, and support if needed.

  1. Design an approach to determine whether a remote BitTorrent client is using Levin’s PropShare al- gorithm, and if so, what upload rate to use when exchanging data. Note that the remote client may participate in other swarms and may exchange with other peers at the same time.
  2. Apply facebook’s “Facebook Connect” service to solve a problem in any of the papers we read this semester. Broken down:

(a) Choose a paper; (b) state a relevant problem (it need not have been raised by the paper’s authors); (c) suggest how using friend information might simplify addressing that problem. (d) explain why friend information (not of friends-of-friends) would be sufficient, or describe how they can be made so.

  1. To provide end-to-end (host-to-host) redundancy suppression, of the sort suggested by Akella, answer the following. Assume that you’re compressing only one direction.

(a) what protocol layer(s) would (necessarily) be involved? (Use OSI reference layers; You may invent layers between other layers.) (b) how would dictionary misses be reported? (c) would retransmissions be suppressed? How or how not? (d) in the event of lossy links in the end-to-end path, how might the dictionaries be efficiently syn- chronized?

  1. RED violates the end to end argument. True or false; and why?
  2. DNS queries to root servers are predominately junk. Design a scheme to eliminate junk queries to the root servers (from compliant name servers). You may modify the protocol to add resource records or response types. (Caveat: yes, most of the bogus queries are from buggy resolvers or buggy name servers. Another layer of bogus-query-prevention is the goal.)