CS-739: Reading Questions

Recall that answers to reading questions should be sent via email to remzi AT cs before 9am on the day of the class.

IMPORTANT: Subject line should look exactly like this follow this format exactly: 739 Reading Question 1 (09/04/15) . Of course, the number of the question, as well as the date, should change each time.

ALSO IMPORTANT: Please no attachments of any kind; this should be an email I can quickly read in gmail without having to run a PDF viewer, Microsoft Word, or any other kind of special viewing tool.

Tip #0: Keep it short, say 2-3 paragraphs.

Tip #1: Don't spend time regurgitating obvious stuff from the papers! The point of these questions is to think , and then to write down what it is that you thought about. I read a lot of write-ups, so the more interesting you are, the better! If you find yourself just repeating a lot of details from the paper, you are going down the wrong path.

Tip #2: Don't spend time criticizing how the authors wrote the paper. That is, I don't need to know whether you thought the paper was well written or not. We'll have plenty of time to talk about those types of things later in the class. Focus on technical aspects for these questions.


Questions

No question for wed 12/09: Just read the paper.

No question for fri 12/04: Just read the papers.

Q12, 12/02: What is the most important idea in the CFS paper that builders of modern distributed systems should know?

Q11, 11/17: Which types of attacks are mentioned in the paper? Which of these attacks do you think are the biggest concern within a modern datacenter? Across datacenters?

Q10, 11/04: What is the trickiest (hardest to understand) part of the Raft protocol, in your opinion? What makes it hard for you?

No question for fri 10/30: Just read the paper and come ready to talk about how quorums can be useful.

No question for wed 10/28: Just read the papers and come ready to talk about 2pc, and typical 2pc costs given differing storage/network technologies.

Q9, 10/21: Discuss how Remus handles failures. What kind of failures in hardware/software can it deal with? What kind can't it deal with? What happens when an unexpected failure arises?

No question for friday 10/16: Just read the papers and come ready to think about the advantages/disadvantages of primary/backup vs. replicated state machines.

Q8, 10/14: What did you learn by reading about Grapevine? By reading about leases?

No question for friday 10/9: Just read the papers.

Q7, 10/07: NFS cache consistency can lead to some funny user-observable behaviors. Can you describe an example where a user sees the system do something that is odd or undesirable?

No question for friday 10/2: Just read the papers.

Q6, 09/29: Two parts today. First: What did you learn from Haryadi Gunawi's talk on SAMC? Second: Name one result from each of the papers (on memory, flash, networks) that you think is least important for system designers to know about.

Q5, 09/23: Pick one figure or table from each paper that you think is most interesting. Describe what you think is most interesting about it, in a sentence or two.

Q4, 09/18: First, re-read the instructions above. Most of you are following them carefully; a few are not. Then, answer this: in the two studies, what was the most interesting thing you learned? Describe.

Q3, 09/16: If you are building a datacenter-based replicated storage system, what are the most important one or two lessons you learned from the two TCP papers you just read?

Q2, 09/09: The U-net paper discusses the notion of zero copy vs. true zero copy. What is the point being made here? Is this important for high performance applications and services? ( we'll do TCP questions for the following class, don't worry, but do read DCTCP sooner rather than later )

Q1, 09/04: The RPC paper presents a number of trade-offs in its implementation. What is the worst trade-off made by the authors, in your opinion?