The 3GPP all IP architecture relies on mobile IP (MIP)
infrastructure for roaming between different access gateways. Thus, MIP home
agent and foreign agent are introduced in reference architecture.A wireless IP
network consist of two components: a wireless access network and a fixed
corenetwork. There are important issues that should be addressed in wireless IP
networks in order toprovide a seamless service in both fixed and mobile
environments. Perhaps the most challenging issue is resource management and
quality of service provisioning. This is even more difficult considering that
there is no native resource management or quality of service control function
in traditional IP networks. Current wired IP only offers the best effort
service model which treats all packets from all users equally. In wireless
environments due to specific characteristics of wireless channel, QoS
provisioning is even more challenging. Furthermore, each wireless access
network, has potentially its own wireless technology and administrative
policies which makes it more difficult to have a uniform resource management
mechanism. The goal of this section is to discuss resource management
architectures for integrating different wireless networks in order to provide
seamless connectivity. The proposed architectures adopt IP as the common
network layer protocol. To minimize the amount of change inside the network, we
use the existing resource management mechanisms proposed for IP networks as
much as possible. In all proposed architectures, the core network is based on
the Diffserv model.
DiffServ-Based Architecture
In DiffServ-Based architecture, not only the core network is
DiffServ-capable but also the access networks are DiffServ-capable. In this
architecture acellular network overlaid by DiffServ domains operates as the
radio access network.In the previous subsection, we discussed how the static nature
of DiffServ can degrade the radio resource utilization in wireless access
networks. Therefore, a more fine-grained architecture based on IntServ was
proposed. It was also mentioned that because of limited radio resources, the number
of flows and consequently the amount of state information required for
IntServ/RSVP7 operation is quite reasonable with respect to the scalability
requirement. These assumptions are reasonable for low-bandwidth systems, e.g. a
3G networks. However, when the access network operates on a high-bandwidth
IP-based wireless technology, e.g. a wireless LAN, these assumptions do not
stand for the following reasons: Typically, such technologies have high
capacities in order of several Mbps. Therefore, it is possible to have a large
number of flows simultaneously in the network. Considering that future cellular
technologies such as 4G will expand the available radio resources to the same
orders, then this will be problematic even in those environments.
• Due to the inherent IP-based architecture of these
technologies, traffic flows have different characteristics and requirements than
those in conventional cellular networks. The applications intended for such
environments are delay-tolerant and do not require strict QoS guarantees (web
browsing compared to voice calls for instance). Also, their generated traffic is
bursty in nature and hence it is difficult to describe their bandwidth
requirements accurately a priori. The types of applications supported by conventional
cellular networks are limited which facilitate the classification of their
requirements. This is not true in wireless LAN environments.
• Mobility patterns are different in WLAN-based hot spot
environments compared to those in conventional cellular networks. Hot spot
traffic is more chaotic and hence more difficult to predict. As a result, it is
not possible in practice to reserve appropriate amount of resources beforehand
for each individual connection which may handoff to the hot spot.In contrast,
traffic aggregates are usually more smooth and predictable thanks to the law of
large numbers. This suggest that class-based resource management is more
feasible in wireless environments.
• The wireless environment is rapidly changing. Wireless channel
capacity fluctuates over time with interferences. So, it is difficult to achieve
strict QoS guarantees similar to those in wireline networks with fairly stable
channel quality. In this case coarse grained QoS guarantees like those offered
by DiffServ are sufficient and in fact more appropriate for the target
application types. For all above reasons we believe that the DSB architecture
is a more appropriate candidate for future all-IP wireless networks than the
ISB architecture.
