Computer Science 280: Homework 8

Homework 8:
4/2/07 (due 4/9/07) Read 6.4-6.6. Think about (but don't hand in):


Section         Number             Points      Comments
6.3 	        9(b),(d)             4         DAM2: 6.3, 11(b),(d)
6.4 		7(b),(d),(f)         3         DAM2, 6.4, 17(b),(d),(f)
                8(b) 		     3 	       DAM2: 6.4, 4(b);  Explain your answer
                25                   4         DAM2, 6.4, 26
                28(b)                3         DAM2, 6.4, 30(b)
                38(b)                3         DAM2, 6.4, 40
                40                   4         DAM2, 6.4, 38
6.5 		7 		     4         DAM2, 6.5, 6
		18 		     3         DAM2, 6.5, 18
		34 		     4 	       DAM2, 6.5, 34.  Write this up carefully
Supp. 		22(b),(d),(f) 	     6         DAM2, Supp. 13(b,)(d),(f)

More problems from Jon Kleinberg:
(1) [5 points] Suppose that we have n identical computing servers, and n jobs that need to be performed. However, because we need to route the jobs quickly, with limited knowledge of the global state of the system, we simply assign each job independently and uniformly at random to one of the servers. That is, the jobs arrive in sequence, and each is assigned uniformly at random to one of the servers, independent of the assignment of previous jobs. Determine the probability that each server is assigned exactly one job, by giving a formula for this quantity in terms of n. In particular, describe the underlying sample space, the event in question, and its probability. (You do not need to write out the entire sample space or the event; it is enough to give a precise mathematical formulation of it.) Also, show that this probability converges to 0 as n increases. (2) [5 points] A biomedical lab runs a cluster of machines that measure and analyze data, and due to the sensitivity of the data, they need to back it up every night at a secure offsite location. They've set up the following automated system for doing this, involving four computers: a Data Aggregator, a Relay Server, a Firewall Server, and an Offsite Backup. By 11:00 PM every night, the results from all the day's experiments reside on the Data Aggregator as a single large data file. At 11:00 PM the Data Aggregator opens a connection to the Relay Server to send it this data. At 11:15 PM, the Relay Server opens a connection to the Firewall Server, and sends the data to it. At 11:30 PM, the Data Aggregator connects to the Firewall Server and sends a duplicate copy of the data. (This is designed to provide some extra redundancy in the backup, just in case.) Finally, at 11:45 PM, the Firewall Server connects to the Offsite Backup and sends the data to it (either the original or backup copy, if it has both -- it doesn't matter which). If successful, each of these individual transmissions lasts at most 15 minutes (and so the final transmission is done by midnight). Now, there's some probability that each of these connections fails, in which case the data doesn't get transmitted across that particular connection. Suppose that the connection from the Data Aggregator to the Relay Server fails with probability .1, the connection from the Relay Server to the Firewall Server fails with probability .1, the connection from the Data Aggregator to the Firewall Server fails with probability .05, and the connection from the Firewall Server to the Offsite Backup fails with probability .02. Assume that the successes of each of these network connections are mutually independent as events. What is the probability that a copy of the data file containing the day's results successfully ends up at the Offsite Backup at midnight? Provide an explanation for your answer.