ch3

Question 1

Streaming vs. elastic traffic

Know what it is and understand the difference

Answer 1

Streaming traffic:

• Require fixed amount of bandwidth
• For instance, calls, streaming audio/video
• Is there enough capacity yes/no
• No real possibility to slow down
• blocking probability big thing

Elastic traffic:

• Data applications
• Don’t require fixed throughput as they share bandwidth
• If it gets crowded, it slows down.
• no real notion of blocking

Question 2

GSM planning problem

Understand the problem, how it leads to the Erlang-B model

Answer 2

How many base stations are needed to get a sufficiently low blocking probability.
We use the Erlang-B model to find the blocking probability

To boost capacity:

• Denser frequency reuse
• Smaller, more cells - Cellular densification

Question 3

Evolution in mobile networks: from CSD to HS-CSD

Understand the translation into single- vs. multi-rate model

Answer 3

single-rate model uses 1 type of job/call/class

CSD (Circuit switched data):
- Uses a single channel

HS-CSD (High Speed - Circuit switched data):

• Uses multiple connections in parallel
• Large-file transfer, video-conferencing, etc.

Question 4

Multi-rate model, product-form solution
Know what it is, understand the idea of the PF solution,
understand the scalability problem

Answer 4

K types of traffic
We try to get the blocking probabilities for each class
We use the multi-rate model to get these.
The classes may have different:
- mean call duration
- arrival times
- required capacity

We make a 2-dimensional markov chain will all the states in the state space
We can then get the state probabilities by making the balance equations where “rate out” = “rate in”
Solve the system of linear balance equations

Question 5

Kaufman-Roberts recursion

Understand how it solves the scalability problem of the PF

Answer 5

When you have a really large state space, the product form solution becomes really large

Kaufman-Roberts recursion calculates the blocking probabilities
Looks at the total number of channels occupied

Question 6

Elastic traffic
Understand what it is, how it relates to TCP and how it
naturally leads to Processor Sharing (PS) models

Answer 6

Mainly used for data applications
Not very delay-sensitive
No fixed throughput required: the flows share bandwidth

TCP traffic shares capacity in a similar way

Processor sharing models shares capacity equally

Question 7

M/G/1 PS model

Know what it is, and what the insensitivity property means

Answer 7

If k customers in the system, then each of them gets processing speed 1/k

Insensitivity property:
- Result also holds for non-exponential service-time distributions