Name:  Marjory Johnson
       Senior Scientist

White Paper Topic:  Some Quality of Service Issues

Affiliation:  Research Institute for Advanced Computer Science (RIACS)
              NASA Ames Research Center

Postal address:  Mail Stop 19-39
                 NASA Ames Research Center
                 Moffett Field, CA 94035 - 1000

e-mail address:  mjj@riacs.edu

Telephone:  1-415-604-6922

Fax:  1-415-962-7772


SOME QUALITY OF SERVICE ISSUES

The Next Generation Internet will support diverse applications
with different Quality of Service (QoS) requirements.  In
addition to requiring guaranteed bandwidth, applications might
require other QoS support.  For example, video-conferencing
applications might require jitter control and synchronization
mechanisms and distributed-simulation applications might require
tight constraints on delay.

Traditionally the Internet has provided reliable transmission,
with little concern regarding delay.  This is not acceptable for
the new classes of multi-media and multicast applications that
are emerging.


Traffic Modeling

There have been several analyses of Internet traffic in the past,
but the nature of applications is changing and the newer
applications are demanding of network resources.  It is important to
determine appropriate traffic characterizations for applications
that are representative of these newer applications, e.g.,
multi-media, real-time, interactive, and multicast.  Traffic models
for some of these applications are appearing, but further modeling
is required.  Once developed, these models could be incorporated
into simulations to analyze proposed protocols for providing quality
of service and to evaluate tradeoffs between various approaches.


Interactions between Applications

In the Next Generation Internet new types of applications will
interact with each other.  While TCP traffic dominates the current
Internet, bandwidth-intensive multimedia applications and multicast
applications are based on UDP.  As these newer types of applications
become more prevalent, a large percentage of the traffic on the Next
Generation Internet may be UDP traffic.

It is important to understand what constitutes a realistic traffic mix
for the Next Generation Internet, e.g., the percentage of traffic that
is likely to be characteristic of yesterday's Internet (i.e., primarily
TCP traffic using window flow control and a relatively small volume of
UDP traffic) and the percentage of traffic that is likely to be
multimedia and multicast UDP-based traffic.  Furthermore, it is
important to understand what this new traffic mix will mean regarding
such issues as congestion control, providing fair access to network
resources, and providing other QoS guarantees to the various
applications.

In addition to different types of applications, different protocol
mechanisms will be interacting with each other in the Next
Generation Internet.  It is important to understand how
different protocol mechanisms will interact with each other, e.g.,
best-effort resource utilization versus admission control and
resource reservation.

Simulation modeling and testbed experimentation will both be useful
in determining the nature of the above interactions.  However some
problems in achieving satisfactory QoS might be apparent only in
large-scale systems.  Hence, it will be important to ensure that the
analyses are scalable.


Interoperability

Different administrative domains within the Next Generation Internet
will utilize different mechanisms to provide QoS.  These mechanisms
need to be able to interoperate, so that QoS can be provided from
end to end.  For example, ATM networks can provide QoS at the link
layer, while the Internet community is developing the concept of
flow within IPv6, along with the use of RSVP for resource
reservation.  The two approaches need to be interoperable.

A more general issue to address is how the Next Generation Internet
can be made to utilize link-layer mechanisms for providing QoS,
whether the underlying network is based on ATM or some other
technology.

Another interoperability consideration in the development of QoS
mechanisms is the different processing capabilities of different hosts.
QoS mechanisms may need to be adaptable to the capabilities of the
receivers.  For example, some receivers will be able to process higher
quality video than others.

An additional problem to address is how to provide adequate QoS
while the Internet is in the process of being upgraded.  Protocols
incompatibilities are likely to exist for some time.