CPS 214 Written Homework 1 Solutions    	    	Fall, 1996

We assume 1K=10^3, 1M=10^6 and 1G=10^9 to simplify 
the calculation.
 
1. The width of a bit is defined as how long it takes to transmit one
bit. In this case, the bit width = 1bit/(1Gbps)=1ns = 10**(-9) seconds.

2. (xKB)*(8b/B)/[yMbps]=.008x/ysec

3. We use "ping -s host packetsize count" of which packetsize should
be as small as possible and we take average value as a "typical"
RTT. Typical RTTs for

a) triangle sites: < 10ms
b) Southeast sites: 20-60ms
c) West coast sites: 80-120ms

4. To transfer the file by modem, it takes (roughly):
       (20MB)*(8b/B)/[28.8b/s] = 5556sec
   To download it from school, it takes (roughly)
       (15min+15min+TimeToDownload)
   Going to school will be faster if:
   
   (15min+15min+TimeToDownload ) < 5556sec

   or TimeToDownload < 5556 sec * (1/60) min/sec - 15min -15min
   or TimeToDownload <  (92.6-15-15) min
   or TimeToDownload <  62.6 min

   This will almost certainly be the case!

5. 
   (a) Data link layer
   (b) Transportation layer
   (c) Data link layer or transport layer
   (d) Network layer
   (e) Physical layer
   (f) Presentation layer

6a) Opening a file is delay-sensitive since the amount of data
exchanged during the open (e.g., the file name) is very small.

b) Reading the contents of a file can be either delay-sensitive or
bandwidth-sensitive, depending on how large the file is.  (According
to profiles of actual systems, most Unix files are small.)

c) Listing the contents of a directory can be either delay-sensitive
or bandwidth-sensitive, depending on the size of the directory.
Generally speaking, the number of files and subdirectories contained
within a directory is small, so the operation is delay sensitive.

d) Displaying the attributes of a file is delay-sensitive.

7. The princple difference is that with a connection-oriented subnet,
a connection must first be established before the communication can
begin. The connection must also be closed when communication ends. 

8a) Acknowledging individual packets is the preferred approach when
the error rate of the physical medium is high enough that the
probability of an error occuring sometime during the transfer becomes
significant. That is, the cost of retransmitting the entire file
(assuming it takes hundreds of packets to send it) is so large, that
you never want to have to do this.
 
b) The second approach is prefered when the error rate (due to link
errors, congestion, etc.) of the path is so low that one never sees
errors. In such cases, the packet headers could be smaller in size if
checksums were omitted, increasing the throughput.

9.  IPv6 addresses are 128 bits long.