In the fall semester of 2002 I will teach a course on
``Quality of Service in the Internet''.

This will be a PhD level course, also accessible to advanced MSc 
students. (MSc students only with permission of the instructor).

Title:      QoS (Quality of Service) and the Internet.
Instructor: Prof Ott.

In the fall of 2002 this will be a special topics course.
CIS786 .  
To be exact:  CIS786-003 , Thu 2:30 - 5:25 pm.

Outline:

Until recently, all or almost all traffic in the Internet was 
``Best Effort'': ISPs and other carriers promise to do their best, 
no further promises are made. 

In that situation, the main ``QoS'' mechanisms were mechanisms that under 
overload cause sources to throttle back, thus alleviating the overload
(or, depending on ones point of view, moving it outside the network).
Examples are the TCP feedback mechanism and other mechanisms that utilize 
the TCP feedback mechanism e.g. RED (Random Early Detection), SRED, BLUE,
FRED, etc. These mechanisms are commonly called ``Active Queue Management''
mechanisms. They rely on sources obeying the TCP feedback rules.

With traffic being more and more commercially important, with more
traffic carried as UDP, with less reason to trust everytbody, and with
special contracts (SLAs: Service Level Agreements), the focus is
shifting.
An SLA is an agreement between a customer and an ISP, which states 
that as long as the customer's traffic remains within an agreed upon 
profile or envelope (usually defined in terms of Leaky Buckets), the QoS 
also will remain within (above) certain bounds.

The shift in focus mentioned above is toward mechanisms that protect
customers who stay within their profile from badly behaved customers. 
Among such mechanisms are WFQ (Weighted Fair Queueing), Self-Clocked WFQ,
Class-Based WFQ, etc.

With these developments, simulation of telecommunication networks has 
become quite important. There is one specific simulation tool, NS2
(freeware from ISI) that has become a de-facto standard.

In this class:

1. Students will be introduced to NS2.
2. Students will learn about various Active Queue Management mechanism.
3. Students will learn about WFQ and variations
4. Students will be exposed to the delay bounds that can be given for 
   conforming traffic that is protected by WFQ type mechanisms.
   (Popularly called ``Parekh bounds'').

With ``high priority traffic'' being protected by SLAs and WFQ
mechanisms, the remaining best effort traffic must change its response
to congestion, in order to quickly back off when required, and then
quickly speed up when bandwidth becomes available again. ECN (Explicit
Congestion Notification) potentially is such a mechanism. The students
will be exposed to ECN.

There is no single good book for any of those topics. A reading list 
will be produced and updated.

PREREQUISITE:
Permission of the instructor. 
A B+ in CIS656 or in ECE637 is enough to get permission from Prof Ott. 
Other cases will be considered individually.