Download A1-solutions

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Solutions to Assignment 1
ECE 6610
Wireless Networks
Notes and Instructions:
1. Five questions, each carrying 10 points
2. Assignment worth 3.75% of final grade
3. Individual work only. Any collaboration will result in an allocation of 0 points for all collaborating students.
4. Due date: January 25th , 11.59pm
5. Submission instructions: Only email submissions accepted. All submissions must be in .doc format.
Send your submissions to TA
Question 1:
Why is fluid fair queuing idealistic and cannot be achieved practically? Why is it
necessary to keep the scheduling algorithm’s complexity low? Explain what the key
advantage of DRR is over WRR.
Sol: FFQ is idealistic because it is impossible to divide packets infinitesimally. The
scheduling algorithm has to be simple to reduce computations at the router and hence the
switching delays. The key advantage of DRR is that it can handle variable size packets.
Question 2:
What is the key difference between the bit-map protocol and the variants of CSMA and
ALOHA? When is the bit-map protocol better than variants of CSMA/ALOHA? When is
it worse? Why?
Sol: The bit map protocol is a contention free MAC protocol. There is an initial
reservation period where each node gets to say whether it has a packet to send or not,
followed by the packet transmissions of each of the nodes that wanted to transmit. CSMA
and ALOHA, on the other hand are contention based protocols.
The bit map protocol is better when there is heavy load. When a number of nodes wish to
transmit there is very minimal wastage of the time for reservation. Likewise when the
load is low the bit map protocol performs poor because a number of slots of reservation
period are wasted and still very few nodes wish to transmit packets.
Question 3:
Why does TCP use the LIMD congestion control paradigm? What happens if another
paradigm such as LILD is used? Explain using the vector representation of a two-user
case.
The LIMD congestion paradigm allows efficiency as well as fairness. The LILD on the
other hand allows efficiency but is not fair. The figures below show the vector
representation of the throughputs for the two flow case using the LIMD and the LILD
paradigms. LIMD approaches the intersection of the efficiency and the fairness lines,
whereas the LILD paradigm just oscillates around an initial point on the efficiency line.
For a detailed description of the vector representation refer to the paper by jain1 et al
given as a reading paper (paper 0 in the reading list).
1
Analysis of the Increase/Decrease Algorithms for Congestion Avoidance in Computer Networks, jain et al
Question 4:
Why are loss recovery mechanisms required both at the link layer and the transport layer
of the protocol stack? Explain with detailed arguments for why such mechanisms are
required at both layers.
Sol: Link layer loss recovery mechanisms recover from errors and losses caused by
interference, collisions and random bit errors. It is a point to point protocol and does not
care about any losses in the higher layers.
Transport layer recovers from congestion related loses. Congestion occurs because of
buffer overflows at routers. Transport layer works as an end to end loss recovery
mechanism.
Question 5:
Explain the motivation behind address classes in IP addresses. What are the drawbacks of
using address classes? How are the drawbacks addressed in the current Internet. Explain
with clear examples.
Sol: Motivation behind the address classes is ease of routing. Class based routing
introduces the concept of hierarchical routes. This means that routers need not know the
addresses of each and every host on the network. They can route packets based on the
network part of the IP address.
Drawbacks: class based addressing leads to a lot of wastage of address space. For
example if a company has a class C address space allocated but has only 150 hosts the
remaining ~106 addresses are wasted. Further if a company that has a class C address
space wishes to expand then a new chunk of addresses have to be given and these might
not be related to the original class C addresses.
Subnetting and Supernetting are principles employed in the internet to over come these
drawbacks. Refer to the lecture slides on IP layer for examples.