Вы находитесь на странице: 1из 4

Exam #1 for Computer Networks

>>> SOLUTIONS <<<

Spring 1999

Name: _____________________________________ SSN: _____________________________ Welcome to exam #1 in Computer Networks. Read each problem carefully. There are eight required problems (each worth 12.5 points) and one extra credit problem worth 5 points. Please write your answers on a separate sheet of paper, start each problem on a new page. For problems asking for a derivation or numerical solution, the majority of points are in the set-up and derivation, the least in the statement of the final answer. You may have one 8.5 x 11 inch sheet of paper with you. On this sheet you may have anything you want (definitions, formulas, etc.) in handwriting only, you may use both sides of the sheet. You have 80 minutes to complete this exam. Good luck!!! Problem #1: (computer communications architecture) a) What is the open systems concept (in the context of the OSI model)? The ability of any two systems conforming to the reference model and associated standards to communicate. b) What are the two key differences between a protocol and an interface? A protocol is between common layers at different sites, an interface is between different (adjacent) layers at a comon site. c) What are the names and functions of the three lowest layers in the OSI model? Physical - a bit pipe (point-to-point), Data Link Control - an error free bit-pipe, and Network - error-free link (end-to-end). d) What are some of the differences between the IETF and the ISO in how they create standards? The IETF is completely open in its standards development process and encourage early implementations as part of the process. The OSI is very closed and exclusive and does not encourage early implementations (simply by not making available to in-progress standards). Problem #2: (packet switching) Create a table with circuit switching, datagram packet switching, and virtual circuit packet switching as categories to be compared. Compare these three types of switching for the following attributes (label each as Yes of No): Dedicated path - - Y, N, N Prob[block] > 0 - - Y, N, N Connection set-up required - - Y, N, N Overheads bits required in data transmission - - N, Y,Y Data arrives in sequence - - Y, N, Y Bandwidth usage is fixed - - Y, N, N Fault tolerant - - N, Y, N Delay is small and fixed - - Y, N, N b) Briefly describe the three generations of packet switches Generation #1 - queues in system memory, central CPU does all moves. Generation #2 - line cards have buffers and CPUs for moves on a high-speed bus. Generation #3 - A switching fabric does moves, line cards have no CPUs. c) What is the advantage of having only input buffering and no output buffering in a packet switch? Output buffering needs to run at N (N = number input ports) times the link speed, input buffering need only run at link speed. 1

d) What is the key challenge (the problem) with having only input buffering and no output buffering in a packet switch? An input buffered only switch suffers from head-of-line blocking. Problem #3: (Data transmission and protocol basics) Assume an asynchronous communications protocol with 1 start bit, 2 stop bits, 1 parity bit, and 7 data bits. Assume that the start and stop bits are never in error, but that the remaining bits have Pr[bit error] = 0.05 (bit errors are independent). Compute the following probabilities a) Pr[a received word has no errors or no detectable errors] Sum(I = 0 to 8) (8 choose i)*(0.05^i)*(1 - 0.05)^(8-i) = 0.715234 b) Pr[received word has a detectable error] 1 - Pr[a received word has no errors on no detectable errors] = 0.284766 c) Pr[a received word has bit #1 in error] 0.05 * 0.05^7 = 3.906*10^-11 d) Pr[a received word has exactly 1 bit in error] (8 choose 1) * (0.05^1)*(1-0.05)^7 = 0.279335 Problem #4 (Data transmission and protocol basics) a) For Tr = 10011 (received frame) and P = 11 (CRC generator polynomial), determine if Tr has a detectable error. Modulo-2 division of 10011 with 11 results in a remainder of 1, hence Tr has a detectable error b) Why is modulo-2 arithmetic used in calculating CRC's? There are probably some good coding theoretic reasons for this, but I would suggest that the answer is that modulo-2 arithmetic is very easy to implement in low-cost and high-speed hardware. c) What is the advantage of having a NACK message in a SAW protocol (over just having time-out at the sender)? Unless the time-out is tuned just perfectly, a NACK message can improve efficiency by informing the sender of a bad message as soon as possible. Problem #5: (Data link control protocol) For a stop-and-wait protocol let Nr be the average number of attempts to successfully transmit a frame (i.e., for a transmitted frame to be successfully received by a receiver). Derive Nr where p = Pr[bit error] (all bit errors are independent). Assume that ACK frames are never in error and that data frames are all a constant length of M bytes. Give the final answer in terms of p and M. Define p f = Pr[frame in error] = 1 (1 p ) (8M ) . The probability, Pr[k sends] =

Pr[k sends] = p f Nr =

k 1

(1 p f ) .

Then,

k (p
k =1

k 1 f

1 p f = (1 p f )

))

k p
k =1

k 1 f 2

= (1 p f )

d dp f

p
k =1

k f

= (1 p f )

d dp f

p
k =0

k f

d = (1 p f ) dp f

1 1 p f

= (1 p f ) 1 1 p f

1 = 1 = 1 p f 1 ( p )(8M ) 1
2

Problem #6: (Fundamentals of queueing theory) a) State the Markov property. The probability of transition to the next step from a current state is independent of any previous states. b) Derive L (mean number of customers in the system) for an M/M/1 queue. From the Markov chain and local balance we can write

n = 0
Since the sum of the marginal probabilities is 1 we can solve for 0

0 = 1 where =

Thus, n = n ( ) . 1

Solving now for L

L=

n =0

n n = ( ) 1

n =0

n n = ( ) 1

n =0

n n 1 = ( ) 1

d d

n =0

1 = ( )

d 1 d 1

1 = ( ) 1 ( )2 1

= 1

Problem #7: (More queueing) a) What is a memoryless distribution (a one line mathematical statement is best)? Pr[x < t + T | x > T] = Pr[x < t] b) Show that the interarrival times of Poisson arrivals are exponentially distributed. Recall that the probability function for the Poisson distribution is,

(t )x e t f ( x ) = Pr[ X = x ] = x!
where, is the mean rate of arrivals and t is a period of time. time of an event (a random variable), we have (by definition), Defining T as the

F ( t ) = Pr[T t ] .
This is equal to,

F ( t ) = Pr[T t ] = 1 Pr[T > t ]


where,

(t )0 e t = e t Pr[T > t ] = Pr[zero events occur in time 0 to t] = Pr[X = 0] = 0!


Now, we plug in our result for Pr[T > t] into our equation (3) and we get,

F ( t ) = 1 e t (which is the exponential distribution)


3

c) What are some of the key properties of a Poisson distribution? Merging of Poisson streams results in a Poisson stream and probabilitistic splitting of a Poisson stream results also in Poisson streams. Problem #8: (More queueing) a) What are the four ways (in increasing order of ugliness) of studying a system as stated by Kleinrock? 1) Mathematical analysis with closed-form (explicit) solutions, 2) mathematical analysis which yields an algorithmic or numerical solution, 3) write and run a simulation, and 4) build the system and measure it. b) State the Kleinrock independence assumption and describe its significance. The independence assumption assumes that the length of a message is chosen independently from the exponential distribution each time it enters a switching node in the network. c) How is a Gigabit network very different from a kilobits per second network and what are the consequences of this difference? A Gigabit network can have more bits in transit in the network that any practival amount of buffering in the receiving node. The consequence of this is that closed-loop feedback for flowcontrol becomes problematic. Extra Credit: a) What is research (one key word is important)? Research is the creation of new knowledge. b) What are the standard six chapters found in many research papers and thesis? Introduction, background, problem, method, proof, and summary. c) As a graduate student, your ambition is to do what (one key word is important)? Make a contribution, publish a paper.

Вам также может понравиться