DELAY MANAGEMENT 73
4.2.1 Cross-Layer QoS Framework
The upper layer adaptation relies on the cross-layer protocol stack that includes five major components
as shown in Figure 4.1: the MAC/network cross-layered differentiation scheduler, based on distributed
Waiting Time Priority (WTP) packet scheduling policy, the middleware-based components such as
the delay monitor, classifier, priority adaptor, and the application-specific requirement adaptor.
Proportional delay differentiation scheduler. This scheduler implements the proportional delay
QoS service differentiation model which was first time introduced as a per-hop behavior (PHB) for
Diffserv (Differentiated Services) in the context of wireline Internet networks [7]. The basic idea of
proportional differentiation is that even though the actual quality level (e.g., delay) of each service
class will vary with traffic loads, the quality ratio between classes will remain constant in various-
sized timescales. Further, such a quality ratio can be controlled by setting service differentiation
parameters, which provide great flexibility in class provisioning. We will discuss details of the pro-
portional delay service differentiation model in Section 4.2.2. Suffice to say at this point that one
of the packet scheduling algorithms that can enforce the proportional delay differentation in short
FIGURE 4.1: Cross-layer framework with upper layer adaptation [29].
74 QUALITY OF SERVICE IN WIRELESS NETWORKS OVER UNLICENSED SPECTRUM
timescales is the waiting time priority (WTP) scheduler [23]. In our design of the differentation
scheduler, we present the distributed cross-layer WTP scheduler between MAC and network layers for
our overall delay management. Details of the scheduler design will be provided in Section 4.2.2.
Delay monitor
: The delay monitor resides in the middleware layer, and it measures the average
round trip delay incurred to deliver multimedia packets for each application. The measured end-to-
end latency contains the delay introduced by traversing the entire protocol stack at the end nodes.
Note that we assume wireless reachability and connectivity between two end nodes when they per-
form real-time multimedia communication. The delay manager is placed in the middleware layer
to measure a true end-to-end delay measurement, and it anticipates that lower protocol stack layers
ensure connectivity and reachability even if nodes move. The sender node attaches timestamps when
sending packets to the destination. When ACK arrives from the destination, the sender retrieves
the sending timestamp, compares it with the current timestamp to obtain the round-trip delay d
i
for packet p. We take M round-trip delay measurements (d
1
, d
2
, …, d
M
) and compute average end-
to-end delay d,
d d
1
1
2
M
i
i
M
=
=
to estimate end-to-end delay from the round trip delay measure-
ments. It will be then used in the Priority Adaptor to update the service priority appropriately. The
priority adaptor will be discussed in Section 4.2.3.
Classifier
. This component is also embedded in the middleware layer, and it determines the
service classes for sending packets according to their priorities. The goal of the classifier is to map
Q priorities to K network service classes with Q > K. Note that applications may have access to a
large priority space, but as we move down in the protocol stack, the priority space gets smaller for
performance reasons. It means, large priority space in the lower network layers introduces service
overhead, which impacts throughput and timing performance of applications and their underlying
supporting system. This mapping is very similar to the priority mapping between network and MAC
layers, which we will discuss in Section 4.2.2. The classifier knows each network service class’s pa-
rameter, a number obtained by rounding up the application packet priority in a certain range. The
application packet priorities will be generated by the Priority Adaptor, and the classifier divides the
application packet priorities into K ranges R
1
, R
2
, …, R
K
. If the priority attached to the application
packet by the Priority Adaptor is in the range R
i
, then the packet will be assigned to the network
service class i with its highest priority within the class i.
Adaptors
. We will consider two adaptors within the framework, the requirement adaptor and
the priority adaptor. We only briefly discuss the interrelations between the two adaptors and their
impact on the other components of the framework. The details of the adaptors will be discussed in
Section 4.2.3. The requirement adaptor gets quantifiable metrics, end-to-end latency, and its variance
(
jitter) from the application. For example, in telephony, one-way delay requirement ranges from 25
to 400 ms, and jitter is set to be 20 ms. The requirement adaptor takes the range of requirements, di-

Get Quality of Service in Wireless Networks Over Unlicensed Spectrum now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.