Routers

 
Routers themselves come in several forms that are determined by the function of the device in
the network. Core routers are huge, costly devices placed at major Internet access points and
hubs and handle the enormous volumes of long-distance Internet traffic. Edge routers are
much smaller devices used in metropolitan access networks that constitute “the edge” from
the perspective of the long-distance service providers. Yet smaller routers are located within
enterprises, and the smallest of all are located in residences and small businesses. Most wireless
broadband base station controllers made today happen to incorporate edge routers. To
see how routers fit into a public network, refer to Figure 6-1. Routers at base stations are switches, but they frequently perform much more complex
functions than traditional class 5 and class 4 circuit telephone switches. Whereas circuit
switches simply follow instructions and send messages to the next node in a predetermined
network path, routers may be capable of making instantaneous decisions as to an overall route
through the Internet based on communications from other routers. Thus, one packet may take
a different route from another, and the packets may arrive at their final destination out of
sequence, whereupon they will be held in a queue until all the packets have arrived. Information
in the packet header will enable the destination router to assemble them in correct order.
Routers make traffic decisions based on routing tables that are continuously and automatically
updated in large core routers and in the edge routers utilized in metro networks. These
tables list locations of core routers across the Internet and the regions served by them, and they
function somewhat analogously to the numbering system used in the PSTN.
Dynamic routing of this sort customarily takes place in edge routers and core routers but
frequently not in enterprise and residential routers where static routing schemes tend to prevail.
The presumption is that the configuration of the local network will be a given and that
network traffic patterns will be entirely under the control of the local area network (LAN)
administrator. Routers were initially designed to switch text data and still-frame graphics only (in other
words, information that could tolerate the delays inherent in packet transmissions). Today
routers are conveying all sorts of delay-sensitive traffic including but not limited to voice;
video, including high-definition video; high-fidelity multichannel audio; interactive entertainments;
video surveillance; video conferences; and business voice-over applications such
“white boarding” and real-time, online, agent-assisted sales presentations.
For a router to provide good presentation quality with such delay-sensitive content, two
things generally have to happen. First, the content to be transmitted has to be cached as close
as possible to its final destination. Second, the router itself has to begin to operate more like
a switch.
The latter generally involves a new packet-switching protocol that I have mentioned
before known as Multiprotocol Label Switching (MPLS). Whole books have been written about
MPLS and the various aspects of the protocol, and I will not attempt to explain it in detail here.
My aim is more modest, simply to indicate what MPLS means to the broadband wireless
operator.
MPLS is based on a concept initially known as tag switching that has been around for
almost a decade and was first associated with Cisco Systems and Toshiba, both of which developed
prestandards router/switches with tag-switching capabilities. The idea was that the basic
IP router functionality would remain intact—no separate and divergent packet-switching protocol
such as frame relay was envisioned. Instead, a switching function would be built on top
of an IP such that at least certain classes of traffic would take determinate paths through the
network—paths that could involve reserved bandwidth. These paths would be designated by
the tag or label associated with the traffic in question, and that traffic, instead of being routed
through first one path and then another according to network conditions, would take but a
single path. Each router/switch would strip the label from the designated stream of data and
rewrite a new label, indicating its next destination. There would be no consulting of routing
tables, no lookup, and no path determination procedure. In the case of the label-handling
function, the router would instead function essentially as a dumb switch.
According to MPLS, labels are assigned to traffic on the basis of common parameters such
as permissible latency, committed bit rate (if any), maximum jitter or timing errors, and so on.
Flexibility in provisioning bandwidth to meet service requirements far exceeds that associated
with asynchronous transfer mode (ATM) or frame relay, the legacy packet standards for handling
diverse types of network traffic.
MPLS itself is not nearly as comprehensive as ATM, the protocol it most resembles, however.
MPLS is basically a technique for segregating and shaping traffic according to its need for
controlled bandwidth, but unlike ATM, which performs much the same function, MPLS lacks
internal mechanisms for reserving bandwidth or enforcing network performance to ensure
that service levels are maintained. These are provided by other protocols such as the Resource
Reservation Protocol (RSVP), used for bandwidth reservation, and DiffServ, which, as the name
implies, posits instructions for differentiating various services. MPLS merely segregates the differentiated
traffic into streams and attaches the appropriate labels for handling that traffic
in transit.
MPLS is embedded in most edge routers sold today and in nearly all new core routers. It
is not common in the combined base station controllers/routers used in broadband wireless
networks, though. True, all equipment that is 802.16 compliant will have quality of service (QoS) mechanisms that are inherent in the standard, but these should not be considered a
complete substitute for MPLS. The routing provisions specified in 802.16 are nowhere near as
comprehensive and flexible as those in MPLS, and, moreover, they are not intended to operate
end to end over WANs. At best, they are a substitute for older ancillary standards for QoS such
as DiffServ and RSVP built around wireline networks.
The question still remains as to how quickly and how extensively MPLS will penetrate
themetropolitan area network (MAN). Like ATM before it, MPLS ultimately rests upon the
assumption that rich multimedia offerings will play an ever-increasing role in broadband
access systems and that IP will become the preferred transport protocol for delivering such services.
Given the slow but steady growth of animated graphics and streaming audio and video
over the Internet, it does seem reasonable to suppose that multimedia will assume much
greater prominence in the future. But what is absent is any clear indication that the telecommunications
industry is approaching a turning point. Internet television is still largely confined
to short news clips, movie previews, and music videos, and, despite the efforts of companies
such as Intertainer to introduce comprehensive video services to broadband operators of IP
networks, such offerings remain curiosities. Internet radio has done better, if for no other reason
than it is easier to achieve acceptable presentation quality with audio-only programming
over medium-speed connections—which is what most so-called broadband services are today.
Even so, Internet radio has not proven to be a major moneymaker for the service providers that
carry it.
I fail to see a clear path ahead for IP multimedia services; the growth of such services is
almost inevitable, but the lack of distinctive or successful program offerings to date is troubling.
Much of the problem obviously has to do with bandwidth and resolution. Even with
themost up-to-date compression technologies, minimally about a half megabit per second
throughput is required for good video quality. When the IP video transmission has to compete
with channelized high-definition television programming over cable, satellite, and terrestrial
broadcast, a 120-kilobit stream full of drop-outs and motion artifacts probably is not going to
attract much of an audience. Recently, a number of companies such as Microsoft, Myrio, and
Minerva have demonstrated high-definition video transmissions over IP networks, and such
demonstrations are encouraging. In the near term, however, only the DSL service providers are
likely to offer high-definition IP television.
Nevertheless, I see MPLS-switching capabilities as an investment in the future. By acquiring
an edge router that can handle MPLS traffic, network operators will be ready for the new
services when they emerge. They will also be better able to ensure QoS end to end in LAN
extension applications across WANs.